The Transparency Paradox: A Role for Privacy in ... · PDF fileWORKING PAPER E. Bernstein, The...

The Transparency Paradox: A Role for Privacy in Organizational Learning and Operational Control

WORKING PAPER

(Please do not cite without permission of the author.)

Ethan S. Bernstein Harvard Business School

Boston, MA 02163 Tel: 617 453 3290

[email protected]

May 19, 2011 I thank Amy Edmondson, Nitin Nohria, Richard Hackman, Bradley R. Staats, Leslie Perlow, and Robin Ely for extensive feedback on the research design for this study and several versions of this paper, and participants in the QUIET Group, WOM Seminar, and Groups Group for their help with various stages of this study. I am indebted to Professor Willy Shih and thousands of “Precision” employees for extensive support in making this field study possible. I gratefully acknowledge the Division of Research at the Harvard Business School, the Kauffman Foundation, and the 2010 Susan G. Cohen Doctoral Research Award for providing financial support for this research.

WORKING PAPER E. Bernstein, The Transparency Paradox

The Transparency Paradox: A Role for Privacy in Organizational Learning and Operational Control

ABSTRACT:

Using data from embedded participant-observers and a field experiment at the second largest mobile phone

factory in the world, located in China, this article theorizes and tests the implications of transparent organizational design on worker productivity and organizational performance. I introduce the notion of a transparency paradox and, drawing from theory and research on learning and control, examine the impact of reduced transparency, in the form of increased group-level privacy, on factory-floor performance. Empirical evidence shows even a modest increase in group-level privacy sustainably and significantly improves line performance, while qualitative evidence suggests the importance of privacy in supporting productive deviance, localized experimentation, distraction avoidance, and continuous improvement. I discuss implications of these results for theory on learning and control and suggest directions for future research.


P a g e | 1

“Man, we are repeatedly told is a social animal, and yet he constantly seeks to achieve a state of privacy.” R. Ingham (1978:45)

Organizations’ quest for worker productivity and continuous improvement is fueling a

gospel of transparency in organizations (Böhm, 2005; AMA, 2007; Bennis et. al., 2008).

Transparency, or accurate observability, of an organization’s low-level activities, routines,

behaviors, output, and performance provides the foundation for both organizational learning and

operational control, two key components of productivity (Deming, 1986). Transparency enables

organizational learning by improving one unit’s access to the expertise, experience, and stored

knowledge of another (Hansen, 1999), thereby creating the potential to increase the quantity and

quality of knowledge transfer (Argote et. al., 2000), knowledge reuse (Murray and O’Mahony,

2007), and shared understanding (Bechky, 2003), accelerate organizational learning curves

(Adler and Clark, 1991), proliferate network ties for exchange of learning-before-doing

knowledge (Pisano, 1994; Fleming, Sorenson, and Rivkin, 2006), and reduce the risk that

localized problem solving fails to contribute to organization-wide learning (Tucker, Edmondson,

and Spear, 2003). Transparency concurrently enables operational control by ensuring access to

richer, more extensive, more accurate, more disaggregated, and more real-time data by managers

and employees, thus improving both hierarchical control (Taylor, 1911; Adler and Borys, 1996;

Sewell, 1998) and peer control (Barker, 1993). Senior management, as architects of their

organizations, is therefore redesigning organizations to make more work more visible more of

the time, embracing innovations including advancements in surveillance and knowledge search

technologies (Sewell, 1998; Levinson, 2009), open workspace design (Zalesny and Farace, 1987;

Brennan, Chugh, and Kline, 2002), and “naked” communication of real-time data via advanced

information technology tools (Tapscott and Ticoll, 2003).


P a g e | 2

Nowhere has this trend towards transparency been more evident than in the design of the

world’s factories, where visual factory implementations have been “spreading through the

factories of the world like a trail of gunpowder” (Grief, 1991). It is difficult, if not impossible, to

find a modern-day facility without a design intended to provide near-perfect observability of the

actions and performance of every employee, line, and function. This observability serves as an

important foundation for all aspects of the Toyota Production System DNA (Spear and Bowen,

1999) and is a necessary antecedent behind the Toyota truism, “you can’t improve a process that

hasn’t been standardized” (Adler, 1993). Factory managers and employees need to see in order to

improve. Observability also provides the basis for many of the widely accepted practices in Total

Quality Management (TQM) implementations (Hackman and Wageman, 1995), which target

simultaneous improvements in both learning and control (Sitkin, Sutcliffe, and Schroeder, 1994).

An emergent logic about the relationship between observability and performance has thus

become dominant in theory and practice: organizations “that are more open [will] perform better”

(Tapscott and Ticoll, 2003).

Nonetheless, the organizational performance implications of increased transparency

remain surprisingly unstudied, both in factories and more broadly. No existing study provides

rigorous, field-based evidence of a causal relationship between observability and performance.

There are, indeed, reasons to be skeptical: detailed field research from the long tradition

of factory floor research in management studies has documented multiple instances where

observability has encouraged hiding behavior among organization members (Roy, 1952; Dalton,

1959; Burawoy, 1979; Hamper, 1986), producing only the appearance of enhanced learning and

control without real benefits to organizational productivity, continuous improvement, and

performance. Dalton describes how managers, mandated by their superiors to conduct “surprise


P a g e | 3

inspections,” instead chose to “telephone various heads before a given inspection telling them the

starting point, time, and route that would be followed” so that each inspection would simply

“appear to catch the chiefs off-guard” (Dalton, 1959). Roy and Burawoy, in reconfirmations of

the “restriction of output” observations in the Bank Wiring Observation Room at the Hawthorne

Works (Mayo, 1933; Roethlisberger and Dickson, 1939), provide substantial detail on the “quota

restriction” and “goldbricking” activities in the Greer machine shop (Roy, 1952), which only

became worse when managers were within sight (Roy, 1952; Burawoy, 1979). Subsequent

insider tales from one of General Motors’ largest and most open plants in the world portrayed

management’s stance on various workarounds like “doubling up” as “a simple matter of see no

evil, hear no evil,” leaving workers with the challenge of hiding their self-defined “scams”

within the context of an observable factory floor—the more observable the factory floor, the

more effort ‘wasted’ on hiding them (Hamper, 1986).

Each of these facilities was designed to be extremely transparent. Yet in these specific

examples, organization design with high observability resulted in only an “illusion of

transparency” (Gilovich, Savitsky, and Medvec, 1998)—a myth of control and learning—

maintained through careful group-level behavioral responses by those being observed. While

observability was achieved through the removal of physical barriers like walls, accurate

observability (transparency) was not. Goffman (1959) originally suggested that increasing the

size and salience of an “audience” has the tendency to reduce sincerity, and increase acting, in

any “performance”. Analogously, increasing observability in a factory may reduce transparency,

displaced by illusory transparency and a myth of learning and control, by triggering increasingly

hard-to-detect hiding behavior—a result I term the transparency paradox.


P a g e | 4

To untangle the transparency paradox, this paper presents a behavioral model of

observability in organizational design, based on both ethnographic and field experiment data, in

an empirical setting that uniquely allows me to investigate deeply and instrument precisely

transparency within the locus of organizational experience and performance. In spite of the

widespread calls for transparency, these field data are the first to rigorously explore the impact of

observability on organizational performance while holding constant such variables as human

capital and organizational culture. Given the lack of prior research, this work begins inductively

and concludes with a field experiment, informed by the initial findings, to build theory and an

emergent model of the organizational performance implications of observability. In doing so, this

paper seeks to make contributions to both the learning literature and the literature on operational

control.

I chose to study workers at Precision’s1 mobile phone factory in Southern China, the

second largest mobile phone factory in the world at the time. Over the past century, factory

studies have been core to building the foundations of organizational theory (Taylor, 1911; Mayo,

1933; Roethlisberger and Dickson, 1939; Roy, 1952; Roy, 1960; Adler and Clark, 1991; Barker,

1993; Anteby, 2008); however, I chose these workers not because of the type of work they did

nor because they worked in the epicenter of Chinese outsourced manufacturing, but rather

because organizational life for them was extremely transparent, in both actions and performance.

In accordance with best practices for visual factory design (Grief, 1991) and TQM (Hackman

and Wageman, 1995), visibility was everywhere: there was a clear line of sight across factory

floors, each several football fields long, such that learning could be quickly captured, distributed,

and replicated by managers; hat color signaled organizational role, function, and rank, such that

expertise could be visibly sought when needed; and both output and quality performance were 1 Precision (name disguised for confidentiality) is one of the top three largest global contract manufacturers.


P a g e | 5

constantly monitored on very visible end-of-line whiteboards, factory floor computer terminals,

and real-time reports to management and customers worldwide. If ever there were an

organizational context where existing practice demanded transparency, this factory in Southern

China was the epitome, and management had implemented the best existing transparency tools

with great diligence and success. I, in contrast, inductively explored the workers’ behavioral

responses, at both the individual and group-level, to such stark transparency, while

simultaneously controlling for any Hawthorne Effects, defined as a circumstance where subjects

improve the aspect of their behavior being experimentally measured simply in response to the

fact that they are being studied, not in response to any experimental manipulation (Mayo, 1933;

Roethlisberger and Dickson, 1939). This paper uses the resulting ethnographic data in Study 1,

and the empirical field experiment it informed in Study 2, to challenge some of the current,

blanket assumptions about the value of transparency for productivity and organizational learning

and construct a contingent, behavioral model of the relationship between organizational

transparency, learning, control, and performance.

This paper proceeds as follows. After describing the research site, the paper details the

methods and results of Study 1, an inductive ethnography by embedded researchers who were

simultaneously operators on the factory lines and participant-observers for this study. The

subsequent discussion of Study 1 uses those results, and the underlying theory, to motivate Study

2, a field experiment. After detailing the methods and results of the field experiment, I discuss

implications and opportunities for further research.

RESEARCH SITE AND METHODS

I studied PrecisionMobile’s 14,000-person, one-million-square-foot manufacturing

facility in Southern China, situated within Precision’s 150-acre industrial park employing over


P a g e | 6

65,000 individuals in 3.1 million square feet of manufacturing space. Spread across roughly two-

dozen factory buildings, Precision employees produced everything from washing machines to the

Microsoft Xbox for original equipment manufacturer (OEM) customers. At the time of this

research, Precision was one of the three largest contract manufacturers in the world,

PrecisionMobile (their global mobile devices division) was the second largest producer of mobile

devices in the world, and the Southern China mobile plant was PrecisionMobile’s largest, with a

capacity to produce up to 2 million mobile devices per week, or roughly 1 out of every 25 mobile

phones sold in the world. As a contract manufacturer, Precision produced mobile phones and

devices to defined specifications for well-known OEM brands, including Sony Ericsson, Huawei,

Motorola, RIM, and others.

Precision management viewed transparency as absolutely essential to performance and

survival. To efficiently achieve such large-scale manufacturing, PrecisionMobile simultaneously

operated large numbers of identical production lines and implemented sophisticated systems and

processes to ensure cross-line transparency for learning and continuous improvement.

Precision’s vision targeted “limitless scale managed by a system that delivers repeatable

execution” (company materials). In the intensely competitive contract manufacturing industry, it

was not uncommon for negotiations with customers over thin manufacturing margins to result in

prices that assumed, without any evidence of feasibility, significant efficiency improvements

during the ramp-up stage before the contract manufacturer could earn any profit at all. Therefore,

organizational learning and operational control within such “limitless scale” and “repeatable

execution” was core to PrecisionMobile’s survival (company materials). Given that mobile

phone product lifecycles were steadily falling, with current models lasting only three or four

months before being replaced by the next generation, the pressure for fast learning and tight


P a g e | 7

control had only increased. When I first arrived at the facility, recent success stories included

ramping stable production of a new mobile phone model from 0 to 96,000 units per day within

four weeks for a tier 1 OEM customer (the fastest ramp ever completed by either

PrecisionMobile or its customer) and shipping 36 million units of a different new mobile phone

model (for a different OEM customer) in 36 weeks using production capacity distributed across

four sites (China, Malaysia, Brazil, and Mexico) with total ramp up time across all four sites of

only six weeks.

My contact at Precision was a board member who worked closely with me to first make

introductions to select global executives and then open the door to PrecisionMobile in China. At

the senior management levels, PrecisionMobile executives were aware of a surprising degree of

productivity variation in their 44 tracked performance metrics across otherwise identical lines;

management’s ‘red-yellow-green’ performance reports across lines resembled a Rubik’s cube.

Given the homogeneity of systems Precision had in place, management’s only explanation for

the variation was what they called “human capital considerations” which they felt they had

inadequate tools to study rigorously. My research interests and methodologies fit nicely within

those management interests, and I was invited to structure an inductive, qualitative research

project at the Southern China facility—PrecisionMobile’s largest, and therefore the one that

would provide the richest source of data. As each factory is run fairly autonomously by a local

General Manager (GM), the GM and his senior team in Southern China facilitated nearly all of

my on-site activities, allowing me to keep my direct connection to the Board and senior

management invisible. I was introduced to the GM, Head of HR, and Head of Operations as a

researcher conducting a study on transparency and human capital and was given nearly unlimited

access to the site. The two studies involved data collection over the course of the next 18 months.


P a g e | 8

STUDY 1 METHODS: Preliminary Qualitative Research—Participant Observation

In a one month preliminary visit to PrecisionMobile China, I adopted a special

methodology to avoid contaminating the environment and the behaviors I was attempting to

observe. My research team included three undergraduate students (two females and one male),

all of whom had been born and raised in China until at least the age of 10. The three students

were inconspicuously placed onto the factory lines as ordinary employees—only the GM, Head

of HR, and Head of Operations of the 14,000 person facility knew their true identities, which

were guarded carefully. Given PrecisionMobile’s 6% monthly operator turnover (average for the

manufacturing sector in Southern China), operators were constantly coming and going, and the

integration of three new recruits was not out of the ordinary. As native Chinese born in regions

that are common sources of migrant workers, the students’ personal characteristics were typical

of new recruits, and the extraordinary diversity of the migrant labor pool meant that the students’

small idiosyncrasies went unnoticed. As college freshman, the researchers’ age matched the age

of the average recruit, allowing them to blend in.

Given the open-ended nature of the behavioral questions I sought to answer concerning

the impact of observability on performance, this research phase was structured inductively, with

the research team entering the field with the open mind characteristic of grounded theory

building through ethnographic participant-observation (Glaser and Strauss, 1967; Van Maanan,

1988). Prior to arrival, the researchers were trained in how to properly conduct an ethnography

and collect ethnographic fieldnotes (Emerson, Fretz, and Shaw; 1995). Upon arrival, they were

put through the PrecisionMobile orientation and then put onto the line just like new hires, living

in the factory dorms and eating at the factory cafeteria. After several weeks of working on the


P a g e | 9

line, in some cases for as long as 12 hours per day, the team had collected several volumes of

ethnographic data, reflecting approximately 800 hours of observation in total.2 Consistent with

the tradition of participant-observation in industrial contexts, data in this context took full

advantage of the embeds’ dual role, consisting of both what was observed and the embeds’ own

observations (e.g., Ellis and Flaherty, 1992; Ellis and Bochner, 2000; Ely and Meyerson,

2010:10). In the former case, embeds reported distanced observations of factory floor activity; in

the latter, to borrow Peggy R. Sanday’s terminology, the embeds were themselves “instruments”

of the research, their reactions to factory floor interactions capturing observations that might

otherwise go unnoticed (Sanday, 1979).

As my presence on the factory floor would have contaminated the study, I limited my

factory floor visits to times when other foreigners were also on the floor. While the embedded

operators were on the line, I waited in an isolated office on a separate floor, where the embeds

would come to record their observations with digital audio recorders during breaks. To maintain

the freshness and purity of their thoughts, as well as to make the most use of the time provided

by normal short production breaks, the embeds recorded their observations verbally, and I then

transcribed all of the recordings while they were back on the production lines. After production

finished each day, we recorded any remaining observations and reviewed the day’s transcripts as

a team, giving the embeds a chance to confirm or challenge each other’s perspectives. At the end

of the month, the researchers revealed their identities in debriefs held at their daily line meetings3,

2 For adequate protection of human subjects (IRB Approval #F16225-101), only four individuals—myself and the three embeds—were authorized to access data with individual identifiers. As per the IRB approval, “a deidentified data set of factory line observer comments [was] created as soon as practicable, for analysis purposes,” prior to sharing any results with anyone (including the Board member who granted us access). Anonymity was practical only because of the size of the site: with 65,000 employees, it was quite easy to remove individual identifiers with zero possibility of later discovery by management. 3 Operator responses to the debriefs were surprisingly positive. In each case, one embed was placed near the debrief so that the embed could observe reactions of the other embed’s fellow operators. On one line, the operators applauded for the embed, as if one of their own had just been admitted into university. On another, the embed in the


P a g e | 10

permitting them to subsequently administer a survey of their end-to-end assembly lines and

conduct in-depth exit interviews with several workers with whom they had built a high level of

trust.

STUDY 1 RESULTS

During this first phase of research, it became clear almost immediately that operators

were hiding their most innovative techniques from management so as not to “get in trouble” for

doing things differently. One of the first rules in which my researchers were trained by peers was

how to act whenever a customer, manager, line leader, or any other outsider came in sight of the

line. First the embeds were quietly shown “better ways” of accomplishing tasks—a “ton of little

tricks” which “kept production going” or enabled “faster, easier, and/or safer production”; then

they were told “whenever the [customers/managers/leaders] come around, don’t do that, because

they’ll get mad.” Instead, when under observation, embeds were trained in the art of appearing to

perform the task the way it was ‘meant’ to be done, per the codified process rules posted for each

task. Because many of these performances were not as productive as the “little tricks,” I

observed line performance actually dropping when lines were actively supervised, a result I

began calling a Reverse Hawthorne Effect because productivity fell, not rose, as a result of an

observer. My embeds’ privileged role as participant-observers, along with the substantial hours

of research time on the lines, allowed me to collect numerous examples (Table 1) of these

productively deviant behaviors (Process-in-use), the performances that were used to hide them

(Process-in-theory), and their antecedents, across 14 categories of tasks. Meanwhile, suggestion

boxes on every line remained empty.

audience heard the operators whispering that they did not “believe that [the embed] was really a college student” and actually laughed at the prospect.


P a g e | 11

************** INSERT TABLE 1 ABOUT HERE **************

The gap between observed and actual behavior was intensified by the fact that these

workers, though unskilled, were by-and-large very clever, according to the embeds’ observations.

As one of my researchers proclaimed, “if they want to hide something from you, they will

succeed.” In a conversation after a line visit by a global manager, one embed seized the

opportunity to ask her fellow operator and trainer about this hiding behavior. In her response, the

operator referred to it as the “privacy” operators needed to keep production moving sufficiently

smoothly to meet ever-increasing management targets. The research team, which had not used

the term prior to that worker’s comment, adopted the word in turn.

The qualitative evidence of hiding behavior to maintain privacy was prevalent in the

transcripts. In fact, while hidden from others, workers were very open with their peers about it.

The embeds even observed lower-level line management (‘line supporters’) helping operators

maintain privacy, as the supporters often warned “hey, you’re doing [this]… don’t do that when

[so-and-so] comes around!” and would let them know when senior management or other

observers, such as 6S auditors, would be within observation range, much like Dalton’s

observations (Dalton, 1959). Because the materials team was the most mobile of the ordinary

operators, materials operators were important in providing warning signals, and one of the

embeds described her materials operator as the line’s “CNN.” The materials team seemed to

always know what was going to happen before it did.

Ironically, the extremely high level of visibility across the factory floor was perhaps the

most important enabler of this behavior, as lines could see management coming long before they

arrived. The research team called this double-sided transparency. As one line operator put it,


P a g e | 12

with what the embeds understood as well-meaning intentions, “of course we prefer that the

managers wear different colors and are easy to spot. That way, we know they are coming.

Otherwise, you can’t even prepare for their arrival.” Visual factory tenets, intended to enable

seeking behavior by operators for needed expertise (Grief, 1991), were instead enabling hiding

behavior.

Informed by this data, a bird’s-eye observer of the floor could indeed observe bubbles of

less productive behavior surrounding any outsider walking the lines. With the benefit of the

embeds’ experiences, behaviors that had previously been quite hidden were relatively easy to see

as long as you were not one of the individuals from whom the behavior was intended to be

hidden. In most cases, the hidden behavior involved doing something “better” or “faster” or to

“keep production going,” often by engaging in activities that operators claimed were “not hard”

and had been learned by “watching [others] do it”—a form of tribal knowledge on the factory

floor. What operators described as “their” [management’s] way of doing things often involved

“more procedures” and was “a lot slower,” whereas the improved, more “fluid” methods were

necessary to avoid complaints from management about the line “being so slow”. In an operator’s

words, the deviance doesn’t “cause any [quality or safety] problems and it keeps production

moving.”

Such private deviance in workplaces is common and well-documented (Roy, 1952;

Burawoy, 1979; Anteby, 2008). What made the deviance in this context so interesting is that so

much of it appeared to be productive for line performance—and that such productive deviance

existed even though the workers, who were paid a flat rate by shift and not piece-rate, had no

financial incentive to enhance performance. When my researchers casually probed operators

about their motivation to hide productive behavior from management, operators’ most common


P a g e | 13

first response involved a perceived lack of capability on the part of management, particularly

middle management: “people from above don’t really know what they are doing. They set all

these rules, but they have no idea how it actually operates down here. Sure, process engineers

come and time these things and set all of these requirements, but they have no idea how people

operate.” When the embeds pushed a bit harder, suggesting that one reason that management

might have ‘no idea how it actually operates’ was because operators were actively hiding it from

them, they received a consistent response: experimentation and the communication of new

knowledge to management was costly. As explained by a 9-year veteran of the factory, who had

worked her way from ordinary operator to line leader and now to a Training Coordinator in the

HR department,

“It’s easy for workers to find something that works better. They have very valuable input. As someone who is in close contact with the line, I know what works well. But when you tell others, they’ll say ‘how do we know how much value this has?’ We don’t have the kind of data they want, and we can’t make a case for our findings.”

For an operator, successfully transferring knowledge highly situated in his or her task first

required transformation of that knowledge into common ground using boundary objects (Bechky,

2003), namely data-driven analytical language that spanned organizational divides between roles

(management or operator), functions (engineer or operator), training (skilled or unskilled), or

tasks (line designer or line operator). But for an operator in this facility, busily doing one set of

tasks 2,400 times per day, it was far less costly to hide that knowledge than share it—keeping it

private meant, in one operator’s words, “everyone is happy: management sees what they want to

see, and we meet our production quantity and quality targets.” Line leaders, who had to manage

across this divide in understandings, described themselves as having the hardest jobs, needing to

“make the impossible possible.” But as one line leader told us, the best line leaders did so by

always “keeping one eye closed and one eye open”—maintaining a privileged position of


P a g e | 14

awareness on both sides of the privacy boundary while being careful not to pierce it, for fear of

the negative performance implications of doing so. When line leaders were with supervisors,

they pretended not to see the productive deviance; when surrounded only by their team, it was

okay for them to quite obviously observe it.

STUDY 1 DISCUSSION: Privacy and the Reverse Hawthorne Effect

Participant-observer experiences at Precision were not consistent with prior theory that

transparency enables performance, but rather transparency appeared to keep operators from

getting their work done. The operators’ choice of the word “privacy” went to the core of my

observations of these behavioral responses to transparent factory design. Mechanisms for

achieving transparency not only improve the vision of the observer but also of the observed, and

increased awareness of being observed in this setting negatively impacted performance,

generating a Reverse Hawthorne Effect.

This unanticipated outcome may, in part, be explained by Zajonc’s seminal finding that

mere exposure to others affects individual behavior by activating dominant, practiced responses

over experimental, riskier, learning responses (Zajonc, 1965; Hackman, 1976; McGrath, 1976),

possibly more so in an evaluative context (Cottrell, 1972), and has been found to encourage a

number of other social facilitation dysfunctions (Bond and Titus, 1983). Similarly, at the group

level, increased observability can lead to less effective brainstorming (Paulus, Larey, and Ortega,

1995), blind conformity (Asch 1951, 1956), and groupthink (Janis, 1983). But qualitative data

collected at Precision (Table 1) suggests the Reverse Hawthorne Effect went beyond passive

social facilitation effects to something more intentional and strategic, thus necessitating a look at


P a g e | 15

the full implications of what the operators referred to as the need for and value of “privacy” on

the factory floor.

A vast interdisciplinary body of theory argues for the existence of an instrumental human

need for transparency’s antipode, privacy—the ability to limit physical, interactional,

psychological, and informational access to the self or to one’s group (Westin, 1967; Altman,

1975; Parent, 1983; Schoeman, 1984; Burgoon, Parrott, LePoire, Kelly, Walther, and Perry,

1989; Schofield and Joinson, 2008; Solove, 2008). The need for privacy exists at both individual

and group levels. Simmel states that normal human behavior involves cycles of engagement and

withdrawal from others—“directly as well as symbolically, bodily as well as spiritually, we are

continually separating our bonds and binding our separations” (Simmel, Landmann, and Susman,

1957). Boundaries providing freedom from transparency, creating a state of privacy, have been

found to enable the authenticity required for meaningful experimentation (Simmel, 1950),

generation of new ideas (Eysenck, 1995; Hargadon, 2003; Simonton, 2003), maintenance of

expertise attached to professional identity (Anteby, 2008), the capacity to trust others (Scheler,

1957), and maintenance of long-term meaningful relationships and group associations (Mill,

1859; Simmel, 1950; Schwarz, 1968; Kanter and Khurana, 2009), all behaviors associated with

effective knowledge sharing (Edmondson, 2002) and “enabling” operational control (Hackman

and Wageman, 1995; Adler and Borys, 1996).

While use of the term “privacy” has not diffused from the legal and philosophical

literatures into the management sciences, a number of pivotal studies in organizational behavior

have touched on the value of privacy without using the term itself. Rich discussions of the value

of boundaries in both the sociological and networks literatures (for reviews, see Lamont and

Molnar, 2002 and Lazer and Friedman, 2007 respectively) suggest that formation of productive


P a g e | 16

individual and group identity requires four components, the first of which is “a boundary

separating me from you or us from them” (Tilly, 2003). Although a highly productive literature

has emerged on the management of knowledge across such boundaries (Carlile, 2004;

O’Mahony and Bechky, 2006), few management scholars have focused on the strategic

placement or creation of the boundaries themselves. Yet if privacy breeds authenticity, and

authenticity enables engaging in hidden yet beneficial behavior, the creation of organizational

boundaries should be of great strategic importance. “Boundary objects” are required for defining

different social worlds (Star & Griesemer, 1989), can take the form of “material objects,

organizational forms, conceptual spaces or procedures” (Lamont and Molnar, 2002), and have

been found to be important for supporting coordinated action and common knowledge at the

group level (Chwe, 2001). There are, of course, group and individual boundary objects that do

not involve physical perimeters or “borders” (e.g., race, gender, status, etc.), but those that do are

reliant on some degree of visibility-based privacy (e.g., a wall, fence, cubicle, etc.) to demarcate

“us” from “them” (Lamont and Molnar, 2002). Where such boundaries are permitted, and how

permeable they are, has profound implications for one’s feeling of privacy and, therefore,

behavior.

Analogous similarities can be drawn to prior work on territoriality, defined as “behavioral

expression… of feelings of ownership toward a physical or social object” which includes

“behaviors for constructing, communicating, maintaining, and restoring territories” around those

objects, whether tangible or intangible, toward which one feels “proprietary attachment” (Brown,

Lawrence, and Robinson, 2005). Territoriality can increase productivity and satisfaction by

engendering feelings of belonging to social groups (Altman, 1975; Lewis, 1979) and reducing

conflict through clarification of the boundaries of social interactions (Altman and Haythorn,


P a g e | 17

1967; Rosenblatt and Budd, 1975; Brown, Lawrence, and Robinson, 2005). But transparent

design pierces territoriality markers in its attempt to improve knowledge transfer and operational

control, putting both tangible and intangible territoriality at risk. By definition, increased

operational control seeks access to tangible territory that would otherwise be protected, like an

experimental workspace or a secure file cabinet, while increased knowledge transfer seeks access

to intangible territory guarded to protect previously private ideas. A move towards transparent

design, ceteris paribus, does make it far more difficult to create a place of one’s own (Pierce,

Kostova, and Dirks, 2003) in the organization. In their review of territoriality, Brown, Lawrence,

and Robinson (2005) ask how a move from “private offices to primarily open offices with few

partitions” might affect organizational performance, given likely defensive behavioral responses

triggered by organizational intrusion on individuals’ and groups’ territory. In the case of

Precision, the answer appears to be rampant, creative hiding of deviant behavior.

While the deviant behaviors were productive, the front-line effort expended to hide them

were not, and no one described this environment as ideal for organizational learning or

operational control, given Precision’s vision and operating model. One side-effect of the front

line’s attempts to create privacy where there was none appears to have been the localization of

operators’ work collaboration networks. A sociometric network survey, with 98% participation

from 2 production lines and 82% participation from the managers and functional teams which

supported them, confirmed significantly localized knowledge communication (Table 2).


When asked with whom individuals communicated to get their work done, only 1.4% of alters

identified were not immediately proximate in role. In a different manifestation of a similar

phenomenon, suggestion boxes were prevalent but empty across the factory floor.


P a g e | 18

STUDY 2: FIELD EXPERIMENT METHODS

Viewing the lack of transparency as a failure of organizational learning and operational

control (Tucker, Edmondson, and Spear 2002), I returned to the site with the intention of

implementing a number of field experiments, drawing on existing literatures in training and

knowledge transfer (Argote et al, 2000). I again chose to study the lines representing the largest

volume product, this time 3G USB datacards, as the mix of the factory production had shifted

substantially from phones to wireless data devices over the ten months since the first visit.

Although the number of operating lines varied by day depending on production needs, on

average there were 16 lines producing nearly the same products across two shifts (day and night),

or 32 line-shifts total, representing a total production capacity of roughly half a million 3G USB

datacards per week. Two lines (four line-shifts) were randomly selected for the experimental

condition, leaving 28 line-shifts in the treatment control. Operators had been randomly assigned

to lines when the lines were initially staffed, thus ensuring no systematic differences at the outset,

and operators rarely switched lines before the end of a product’s production life cycle, thereby

containing potential diffusion of the treatment condition across lines. Through the management

information systems, I tracked detailed hourly production and quality data for all 32 lines. To

supplement the quantitative data, I assigned one embed to a treatment line and one embed to a

control line, although neither embed was told details of the treatments prior to implementation.

At its best, executing a field experiment is an inductive process that iteratively

incorporates the input of the individuals involved (Perlow, 1999:71-74). In preparation for the

experiment, I asked PrecisionMobile’s engineering department to put up a curtain between an

adjacent pair of experimental and control lines to avoid cross-treatment contamination. In this


P a g e | 19

setting, changes to the factory environment are very common, so a change such as the curtain

was not out of the ordinary. Given the repetitive nature of the work tasks, however,

environmental changes typically serve as topics of conversation, as it did in this case. The

embeds reported multiple theories circulating amongst the operators on the purpose of the curtain.

When the bar for the curtain was initially hung from the ceiling, “people joked that they could

hang their clothes on it.” When the curtain went up, an operator “made a swine flu joke out of it,

saying that people with the disease will be quarantined on the other side.” And then the operator

adjacent to one embed said, “wouldn’t it be nice if they hung up curtains all around the line, so

we can be completely closed off? We could be so much more productive if they did that.”

Although the originally planned interventions involved increased transparency, not less, I

decided to implement her suggestion in place of our own ideas. Starting several days later, for 5

consecutive months, the 4 experimental lines operated inside the equivalent of a hospital bed

curtain (Appendix A).

To ensure a clean experimental design, where the relationship between privacy and

performance was accurately instrumented, I painstakingly controlled for Hawthorne Effects. A

Hawthorne Effect, first understood through the research done at the Western Electric Hawthorne

Works in the 1920s and replicated elsewhere, refers to a circumstance where subjects improve

the aspect of their behavior being experimentally measured simply in response to the fact that

they are being studied, not in response to any experimental manipulation (Mayo, 1933;

Roethlisberger and Dickson, 1939). While the presence of a Hawthorne Effect has been called

into question (Carey, 1967; Jones, 1992), anticipating a potential Hawthorne Effect, I carefully

designed this field experiment to avoid it. Because the same space is used by the day and night

shifts, the curtains—once installed—were present for both shifts, but the night shift was under


P a g e | 20

the clear impression that the curtains were “for something that the day shift was doing,” thereby

eliminating the possibility that the night shift’s performance was a response to feeling special or

believing that they were being studied, rather than simply to the reduced transparency afforded

by the experimental treatment, i.e., the curtain. Although the research team neither created nor

confirmed that impression, I did ensure not to disprove it by only conducting surveys and

interviews with the day shift.

STUDY 2 RESULTS

Performance on each of the four lines surrounded by curtains increased by as much as 10-

15% after the first week and maintained a lead over the 28 control lines for the remaining 5

months of the experiment (see Table 3 for a graphical representation).

************** INSERT TABLE 3 & TABLE 4 ABOUT HERE **************

Table 4 gives descriptive statistics for each relevant grouping of lines. The descriptive statistics

demonstrate the degree of improvement on the experimental lines. Because accepted wisdom at

Precision held that the night shift’s performance at this site was lower, Table 4 breaks out the day

and night shifts individually. However, the operators on the line cycle from day to night shift

monthly or semi-monthly (i.e., operators on Line F60-Day rotate together to Line F60-Night and

vice-versa), so for purposes of excluding the Hawthorne effect as a potential explanation for the

performance improvement, it is necessary to follow the groups of people as they rotate back and

forth. The last two sections of Table 4 show descriptive statistics for those groups. Note that even

as lines rotate, my embeds confirmed that operators never communicate across shifts, as one shift

is sleeping while the other is working.


P a g e | 21

A difference in differences estimation model confirms this performance improvement

result and permits a more disaggregated, hour-by-hour analysis of the data. Difference in

differences estimation has become an increasingly popular way to estimate causal relationships

(Bertrand, Duflo, and Mullainathan, 2004) and consists of first identifying a specific intervention

or treatment and then comparing the difference in outcomes, before and after the intervention, for

groups affected by the intervention to the same difference for unaffected groups. Difference in

differences estimation models have three primary advantages: simplicity, the potential to

circumvent many of the endogeneity problems characteristic of cross-sectional analysis, and

robustness, assuming appropriate corrections for serial correlation are made (Bertrand et al 2004).

The basic estimation model is:

Y = β0 + β1*dExp + β2*dTime + β3*dTime*dExp + ε (1)

where Y is the units per hour produced on the line, dExp is a dummy variable expressing

whether a line is an experimental (1) or control (0) line, dTime is a dummy variable expressing

whether the time is before (0) or after (1) the intervention, and dTime*dExp is the interaction of

the two. I avoid selection endogeneity by randomly selecting treatment and control lines.


Table 5 reports the results of the difference in differences estimation models. In Model

(1) and Model (2), performance is measured as an average of the units per hour produced per day

on the experimental and control groupings of lines. Model (3) and Model (4) disaggregate the

groups into individual lines, using instead the average unit per hour production of each

individual line as the dependent variable. That disaggregation also permits controls for line fixed

effects in both Models (3) and (4) and day fixed effects in Model (4). Models (5) and (6)


P a g e | 22

disaggregate the daily production averages into actual hourly production, again controlling for

line fixed effects in both Models (5) and (6) and day fixed effects in Model (6). In each model,

the interaction term dTime*dExp is highly significant and positive, indicating a sustained and

significant improvement in production on the lines surrounded by curtains.

As difference in differences estimates have been shown to suffer from inflated standard

errors due to serial correlation with data from a large number of periods (Bertrand, Duflo, and

Mullainathan, 2004), I used Bertrand et. al.’s suggested remedy of collapsing the time series

information into a “pre” and “post” period by group, which they argue to be valid when a small

number of groups is involved. Doing so, in effect, takes into account the effective sample size

and thus avoids over-estimation of significance levels. In this study, each of the 32 lines

represents a “group” as defined by Bertrand et. al., with “pre” and “post” periods defined as the

time before and after the curtains were installed. As shown in Model (7), the dTime*dExp

variable remains significant at a 5% significance level.

The qualitative data collected by the embedded participant-observers on the line during

the first month of the experiment, supported by a series of 15 detailed interviews conducted by

the participant-observers post-experiment, offers highly detailed accounts of the mechanisms

behind this performance improvement. According to the operators on the line, three categories of

changes contributed roughly equally to the boost performance: (1) privacy to permit tweaking the

line as temporary issues arise (productive deviance); (2) privacy to permit experimenting with

new ideas prior to explaining them to management; and (3) privacy to permit avoiding

interruptions with negative consequences from outside of the line without engaging in value-

reducing hiding activities.


P a g e | 23

Tweaking: Real-Time, Temporary Adjustments to the Line

The manual assembly of 2,400 devices per shift involves the timely and intricate mixture

of three sets of ingredients: roughly a million components, approximately 100 assembly tools,

and roughly six dozen people—half line operators, half functional support (e.g., materials, QA)

or experts (e.g., process engineers (PE), industrial engineers (IE), technical engineers (TE)).

There can be great variance in each of those ingredients: components can be defective, assembly

tools can break down, and human beings can have bad days. As one supporter explained, “when

unexpected things happen, operators need to tweak the line to solve the problem, minimize its

impact, and stay on track to meet our production target. But if anyone outside of the line catches

us operating in non-TQC ways [i.e., not according to the Total Quality Control charts posted in

front of each station], they will blame the problem on the fact that we are not following the TQC,

even though that’s the solution to the problem and not the problem itself.” When problems arise,

problem-solving activities attract attention to the line, thereby making it harder for operators to

tweak their activities. Instead, they end up slowing down or even stopping the line and waiting

for experts to come and solve the problem, which “can take a long time” according to several of

the operators the embeds interviewed. One embed witnessed the emergence of a significant

bottleneck on her line which drew management attention immediately but then led to a two-hour

wait before the correct process engineer could authorize an official fix to the problem.

The curtain changed this dynamic significantly. Where privacy to tweak the line had

previously been achieved through carefully hiding the adjustments, the visibility curtain

substituted for that hiding behavior, making tweaking within the curtain far more transparent to

other operators on the line. Tweaking was particularly prevalent with respect to allocation of

tasks to workers, one traditional form of productive deviance. As temporary bottlenecks arose,


P a g e | 24

workers moved fluidly to reduce them—the line’s own form of organizational improvisation.

Having previously observed workers dedicated exclusively to one station for months or even

years, after the installation of the curtain, the embeds were surprised at how many of the

operators had cross-trained themselves on other stations, especially those immediately adjacent

to their own. The workers reported this fluidity to us as “making the line feel more like a team.”

In fact, the operator who originally suggested the curtain, when interviewed at the end of the

experiment, told us that “when I brought up that suggestion, I was thinking that the curtain could

potentially make our line look more like a team, so that the whole production can be team work.”

The curtain transformed the work from individual to collective via common knowledge (Chwe,

2001), even without changing the tasks themselves. Just as strategic alliance teams are more

successful with space to “learn to work together away from the spotlights” (Doz, 1996), team

learning at Precision benefitted from the curtain’s shield against the spotlight of the factory floor.

The line leader was clear that serious line problems, including persistent bottlenecks, would still

be resolved by escalating the problem to engineering or management, since that was “their job”

to fix. But temporary issues that could be solved or at least mitigated locally would be handled

by the line, through line-level learning, to avoid slowing down production in the short-term.

While not all tweaks were beneficial, the performance data is unequivocal in demonstrating that

they were, on net, positive for line performance, both for the day shift and the Hawthorne-

controlled night shift.

In addition to static line improvements upon installation of the curtain, the effectiveness

of tweaking also improved dynamically over the five months of the experiment. The curtain,

while shielding tweaking from immediate interruption, simultaneously provided the privacy

necessary for operators to engage in activities intended to improve their capability to tweak. For


P a g e | 25

example, one embed reported that “when people weren’t as a busy [during unavoidable

production downtime and/or lower target production days], I saw people switch roles a lot so that

they can learn multiple tasks.” This sort of cross-training would have been nearly impossible in

the visible condition, but with visible privacy, operators were able to self-train in order to make

the line more fluid for future tweaking.

Self-training extended beyond switching tasks to attempts to increase the scope of what

could be safely and effectively tweaked. For example, one of the greatest challenges for

operators was that most of the computer-based tasks, including stations designed to load software

onto the devices, test the software, and test the send/receive transmission functions, was run

using English-based software—an unavoidable circumstance for PrecisionMobile, as the

proprietary software was provided directly by an English-based engineering division of this

customer’s (OEM) organization. As a result, computer training for operators involved

memorizing a specific set of actions: when the embed was trained on the computer, she was told

“click the first button here, and the second button here, when this menu comes up, click this

button, and then pick the third one from the top.” Because she understood English, the embed

learned it very quickly. But for the average operator, the process involves rote memorization.

When the process works properly, a green “OK” pops up on the screen and is easily recognizable.

But error messages in red are typically undecipherable to operators, requiring them to stop and

wait for an engineer to come fix the problem for them—a problem they could often fix

themselves if they understood the error (often as simple as “software upload failed—please try

again”). On the first day after the curtain went up, an embed observed the supporter on one of the

curtained lines instructing

“everyone, especially those at the computers, to copy down error messages they don’t understand while they were working. So a couple of women who work with me at the


P a g e | 26

scanning stations were copying down words constantly. Anything red came up, she was writing them down and trying to figure out what they mean. They were being very diligent about that… I was really impressed.”

The supporter subsequently brought that list to a cooperative English-speaking engineering

manager, who helped him translate. He posted the translations next to the computer so that

operators could problem solve themselves, when possible. For a different computer terminal

which spit out error messages in programming code instead of English, they pursued a similar

strategy. The result: the line became better and better, over time, at tweaking the process such

that they could maintain production speed in the event of temporary issues, which were reported

to the engineering department in parallel (automatically via the computer terminals) for ultimate

resolution.

Interestingly, operators explained that while the privacy curtain permitted tweaking, it

was the resulting transparency that emerged inside the curtain that allowed it to be effective on

both the day and night shifts. The purpose of tweaking was, as before, to ensure that production

and quality targets were met. For the “team” to be effective in knowing when to tweak, progress

towards targets needed to be transparent. The operators’ solution to this was noteworthy: first,

they moved the production reporting whiteboards outside of the curtain and continued to

“smooth” the hourly reporting such that targets were consistently and smoothly met (see “Line

Reporting” in Table 1 above for more detail); then, they put the actual production, as tracked

real-time by the PrecisionFlow IT system, on several computer monitors around the line so that

members of the line would always know whether they were actually on-track to meeting targets.

Carrying that idea one step farther, they subsequently asked the engineering department to install

large monitors at the end of their lines (inside the curtain) to show simply their shift’s production

target (manually entered) and their real-time progress towards that goal (see Figure 1). Group-


P a g e | 27

level privacy was effective by creating a knowledge-centric in-group and out-group (McGrath,

1984), although the qualitative data suggest that the productivity improvement was more due to

the benefits of “privacy” than feelings of in-group and out-group alone. The curtained boundary,

by providing visible privacy, had removed the need for encryption, but only within the curtain.

Experimenting: Prototyping New Knowledge Before Sharing It

In concert with temporary tweaking, visible privacy appears to have encouraged

operators within the curtain to experiment with an increased number of permanent improvements

to the line. Upon the curtain going up, one embed commented that the supporter on her line

“walked around the line and was being very quiet, staring at every single station, pondering it for

a while, looking for any possible improvement.” Table 6 provides a selected list of innovations

which the operators experimented with during the early stages of the five-month experiment

when the embeds were still working on the lines. Based on the embeds’ experience on the line,

they reported that the innovations were a mix of pre-existing and new ideas: some of these were

ideas that were just waiting for an opportunity at experimentation, while others reflected novel

learning on the line through the increased levels of experimentation the curtain enabled.

The visible boundary provided by the curtain enabled experimentation through two

mechanisms. First, it allowed the line to collaborate on new ideas. As one line supporter

explained prior to installation of the curtain, “when we experiment with changing these things,

when we are in a small clump discussing issues, or when we help each other when we move

around and do things that we are technically not allowed to do, how do we stop the 6S people,

the QA people, and the general managers from coming around and questioning us? Sometimes

[observers] don’t even have authority over our line, but they still intervene.” Second, the curtain


P a g e | 28

provided the privacy necessary to prototype these process innovations before sharing them. As

an operator who contributed a number of new ideas explained in a post-experiment interview,

“we have all of these ideas… but how do we feel safe to try them? We’ll experiment as long as

the consequences aren’t so great. As long as the price we pay isn’t so great.” When senior

managers heard these quotes, they were surprised. Precision not only encourages learning on the

line, but they are constantly searching for productivity improvements, as every little idea

contributes to their razor-thin margins. Senior managers immediately questioned why middle

managers were not taking advantage of these ideas. But a single question made them pause: how

was middle management to evaluate whether these ideas were good or not? The only way to

evaluate these sorts of manual process innovations is to prototype them, analyze the performance

changes, and then make a decision. Trying new ideas, however, could only happen at the

expense of variation from exploiting existing best practices, and that was a tradeoff that middle

management had no incentive to make (March, 1991). The best people to make that judgment

call were, indeed, those individuals who actually completed these tasks 2,400 times per day—the

operators inside the curtain.

The visibility curtain provided operators with just enough privacy to experiment within

the bounds of acceptability, enabling on-the-line experimentation. Having produced 2,400 units

per day, often for months or even years, individual workers had developed a number of bottom-

up innovations to improve existing processes, which needed trial-and-error prototyping to prove,

disprove, or develop further. Under the observation of management, the cost to the workers of

sharing, explaining, and/or fighting for their innovations simply outweighed the benefits, a

finding consistent with organizational learning failures previously observed in nursing operations


P a g e | 29

(Tucker, Edmondson, and Spear, 2002). Some degree of privacy was required to experiment,

develop a ‘prototype’ process, and prove the concept prior to sharing.

Even with the curtain, sharing remained an important part of the process, not just for

organizational learning but also for the operators themselves. Operators were paid per hour, not

piece rate, and there were no financial incentives for them to work faster. However, the pride of

sharing an improvement, or of using an improvement to produce faster than the other lines, was

at least a part of the strong motivation driving operators to experiment. Operators from different

lines would often compare, late in the shift, various lines’ production for the day, with

“admiration” offered to the top achievers followed by demonstrations of “how they made so

much so quickly.” While the factory provided excellent working conditions for China, factory

life was not easy, and a great deal of encouragement took the form of peer-to-peer factory floor

wisdom. When one of the embeds asked why she should bother to do any more than the

minimum, one of the factory women responded, “if you quit, then they have won. If you want to

win, you stick here and you work your hardest and you prove to them that they are wrong [about

your lack of ability to make it in this environment].” Sharing a real improvement, home-grown

by an operator on the line, was one of the best forms of winning.

Even when pride was insufficient to motivate sharing, good ideas spread quickly. While

activity on the line became less visible with the curtain, line performance and quality remained

extremely transparent. It took at least a month for the analysts to notice, in their monthly report,

the performance improvements reported in Tables 3, 4, and 5, but once they noticed, there were

plenty of people going inside the curtain to see what was different. The analysts also noticed the

increase in variance on those lines, but any concerns about the increased variance were quickly

overlooked in light of the interest in the increased performance. The curtain provided sufficient


P a g e | 30

privacy of the prototyping process to keep the unwanted variance under the radar long enough to

generate a proof of the benefits.

Avoiding Interruptions with Negative Consequences: The Benefits of Managing by

Standing Still

A third source of the curtain’s performance boost came simply from removing the need

for operators to engage in many of the non-value-added activities listed in Table 1. When being

observed, the system of codes and hiding behaviors that operators adopted reduced productivity,

and those non-value-added hiding activities were both systematic and non-trivial. As one embed

explained, “the wording that the supporter used for watching out for managers was ‘fang shao.’

In Chinese, that phrase refers to the lookout person traditionally assigned to watch out for cops

during an illicit activity, or during war to watch out for enemy activity while the rest of the unit

is doing something. So the supporters think that’s part of their job—that’s part of what they are

supposed to do in their daily organizational lives.” And when the signal is given, everyone

assumes the hidden, less-productive version of their working task. From a productivity

standpoint, both the lookout and hiding activities represented a waste of time and effort. As one

operator said simply, “if we didn’t need to hide things from the management levels, we could

finish production so much faster.” That indeed appears to have been true.

Even if visits by managers with formal authority over the line were relatively infrequent,

others could provide an equivalent interruption. Categories of work interruptions include both

intrusions and disruptions (Jett and George, 2003), and the curtain enabled the removal of both,

especially those that had a negative effect on productivity. As one line leader explained, “people

who have nothing to do with this line need to get away, because materials people, random


P a g e | 31

engineers, random management… they come around and distract you. That’s the first reason—

because they come and chat with you, they chat with your neighbor, and that distracts you. And

sometimes they play with your material… it’s distracting.” As an operator added, “it’s already

hard to keep track of everything and make sure I do everything correctly. They come around,

they pick up a card, they play with it… then I have to pay attention to the card that you are

playing with, make sure they don’t break it, make sure they don’t walk away with it… it’s just

extra wasted time that I don’t have.” The privacy of the curtain thus permitted added

concentration on the tasks of the line. With fewer visual distractions and physical intrusions, it

also focused operator attention on the social unit inside of the curtain (Dunbar, 1992)—on both

the people and the tasks contained on the line—much the way psychotherapists set up therapy

environments to keep patients ‘on-story’ by shutting out extraneous stimuli (Gabbard, 2005).

This, one operator explained, is what led to a stronger sense of “team spirit” on the line. Perhaps

the best indicator of this spirit was the fact that the curtained lines quickly also became the

loudest, with the most talking inside. While one senior leader from a different line “really

dislikes people talk during production because she thinks it slows down the production,” the

curtain provided a sufficient boundary to minimize her impact on the line.

Interruptions can also have a positive impact on productivity (Jett and George, 2003;

Staats and Gino, 2010), so a blanket removal of intrusions could be a net negative for

organizational performance. Interestingly, the curtain appears to have been sufficiently

permeable to block many negative intrusions and disruptions without impeding all positive ones.

Management could still come and go as they pleased, and those interruptions—although now less

frequent—appeared to be more valuable on average. As one line supporter explained, “when

management comes into the curtain now, it is more often for good reasons. People only come to


P a g e | 32

this line when it is a purposeful destination, not when it is a convenient place to stop, look busy,

and be an imposition.”

Ironically, operators also saw a downside to this aspect of the curtain. In post-experiment

interviews, several operators mentioned that the curtain prevented them from “seeing the

management coming as quickly” and therefore occasionally put them at greater risk of being

caught doing something they were not supposed to be doing. But here, too, operators learned to

improvise. Movement of the curtain by the entrance to the line became a signal of possible

intrusion and triggered caution on the line. Operators, to avoid triggering that response

themselves, started using the back or side exit rather than the front. Beyond the curtain, nothing

had changed. Inside the curtain, however, transparency replaced the previously prevalent and

highly guarded hiding behavior.

DISCUSSION:

The research at PrecisionMobile illustrates the transparency paradox: reducing

observability on the factory floor through the introduction of physical group boundaries

paradoxically improved performance and, with it, transparency within the new boundaries.

Existing theory predicts that reduced observability will reduce both organizational learning and

operational control to a detriment to performance. At Precision, however, reducing observability

simply dissolved the myths of learning and control exposed in Study 1. Furthermore, in reducing

observability between lines, the curtains increased transparency within them. Edmondson argues

that the seeds of organizational learning are “local (focused on specific organizational tasks),

interpersonal (influenced by individuals’ perceptions of the social climate), and variegated (non-

uniform in both learning and learning goals)” (Edmondson, 2002), and group privacy at


P a g e | 33

Precision enabled such learning within lines, fueling performance gains. Group privacy

boundaries similarly blocked more coercive forms of cross-group control and replaced them with

more enabling forms of within-group control, increasing performance. The result: through

improved tweaking, experimentation, and distraction avoidance, group privacy enhanced

performance.

How is group privacy achieved? If privacy is a basic human need, it is accomplished in

society, organizations, and at Precision in only one of two ways: either by blocking visibility of

people’s actions through physical boundaries (e.g., private offices, team meeting rooms, war

rooms, phone booths) or by blocking understanding of people’s actions through encrypted

communication (e.g., the mafia (Blok and Tilly, 1975; Gambetta, 1996), finals clubs at Harvard

and eating clubs at Princeton (Karabel, 2006), and the CIA (Marchetti and Marks, 1974)). We

either close the door, window, or curtain, or we speak in code that only chosen-others can

interpret. In either case, we achieve privacy and reduce transparency. In other words, we create

either visibility or encryption boundaries (or both) to separate the private from the transparent

(Figure 2).

************** INSERT FIGURE 2 ABOUT HERE **************

Because these two forms of privacy are substitutes in most cases—few organizations find

the need to construct both visibility-based and encryption-based privacy boundaries, existing

organizational research has collapsed the two forms of privacy into one, instead focusing the

theoretical discussion almost exclusively on movement from “private” to “transparent”

organizations (c.f. Best, 2005), or from the top-right to bottom-left movement in Figure 2. But at

Precision, the movement in these two studies was along the off-diagonal, starting from the top-

left (highly visible but highly encrypted) and moving to the bottom-right (less visible but less


P a g e | 34

encrypted). Functionally, that resulted in a shift from less visibility across lines, levels, and

functions, and more visibility within.

Why would groups, such as the lines studied at PrecisionMobile, choose one form of

privacy over another? While the literature is silent on this question, this research suggests that

encryption boundaries offer one major advantage and suffer one large disadvantage in

comparison with visibility boundaries. Encryption boundaries have the advantage of being far

more versatile: while visibility boundaries are physical and thus can only be constructed where

legitimate (institutionally or otherwise), speaking in code can happen without raising the

awareness of uninformed parties. That versatility, however, comes at a cost—encryption

boundaries are far more expensive to construct and deconstruct, and operators at Precision often

complained about the energy they needed to invest in hiding efforts. Logic would therefore

dictate that encryption boundaries would be reserved primarily for illegitimate hiding—as

Simmel stated, “where [visibility-based] privacy is prohibited, man can only imagine

separateness as an act of stealth.” (Simmel, 1964) Where privacy is legitimized through physical

boundaries, the need for “stealth” encryption dissipates.

A Behavioral Logic of Transparency

Such a decision choice between the two mechanisms for achieving privacy implies that a

behavioral logic of transparency and privacy must be a contingency-based phenomenon.

Organizations decide how much visible privacy to support; then individuals and groups decide

how much code they will implement to make up for remaining, unmet privacy needs. Ultimately,

there may be agency at both senior levels (design) and lower levels (code) of the organization,

with the lower level’s code contingent upon the senior level’s design.


P a g e | 35

PrecisionMobile is no exception. Prior to the installation of the curtain, transparency

imposed through public visibility encouraged individuals to create privacy through encrypted

behavior (the top-left of Figure 2)—a code that only peers could decipher— without the

knowledge of management. With the installation of the curtain, a boundary to visibility was

created, allowing operators to unencrypt their communication. For operators, encryption was a

cost-benefit decision: with the installation of the curtains, the benefit of encryption was reduced,

while the cost of maintaining the code remained the same. Thus, the data demonstrated a

contingent model of transparency: operators determined how to satisfy their instrumental needs

for privacy depending on how transparently management designed their environment. At least at

Precision, aiming for the bottom-left (high visibility, low encryption) would simply result in an

outcome at the top-left (high visibility, high encryption), while aiming for the top-right (low

visibility, high encryption) would result in an outcome in the bottom-right (low visibility, low

encryption). As illustrated in Figure 2, only two cells were realistically achievable at Precision:

the top-left (high visibility, high encryption) or the bottom-right (low visibility, low encryption).

That result has implications for organizational theory, given much of the research to date

has focused on the movement from “Private” to “Transparent” organizations. Indeed, at least at

Precision, efforts to move to the high visibility, low encryption (as established management

precepts dictate) would result in the original status-quo: a highly transparent organizational

design with significant encryption in peer communication. Instead, a more valuable question is

which of the two off-diagonal cells—top-left (privacy by encryption) or bottom-right (privacy by

visibility)—are preferable for organizational learning, productivity, and performance.

For Precision, movement from the top-left (privacy by encryption) to the bottom-right

(privacy by visibility) by installing visible privacy, via curtains, resulted in a number of valuable


P a g e | 36

behavioral changes which made deviance far more productive. But the curtains did not offer the

individual workers any additional empowerment, autonomy, or opportunity to operate as a self-

managed team. Indeed, in a context where real-time performance is so closely and transparently

monitored, as on a large number of metrics by Precision analysts, the curtains actually offered

individual operators very little—only privacy of activity was granted, and a permeable,

temporary, and delicate privacy boundary at that. Compared with other studies that have looked

at significantly more invasive interventions in autonomy or empowerment (Mayo, 1933;

Roethlisberger and Dickson, 1939; Barker, 1993), this intervention reflects an organizational

change of minimal cost. But the impact at the group level was substantial: a sustained marginal

performance increase as large as 15%, induced from such a minor privacy intervention, is a

testament to the power and value of visible forms of group privacy.

The result is simultaneously a testament to the value of transparency. An important third

dimension is excluded from Figure 2: time. Encrypted privacy is durable—so long as the

decipher key remains in the hands of those who wish to hide, their activities remain hidden.

Visibility privacy, on the other hand, is fleeting—it vanishes as soon as the boundary is pierced.

When one walks inside the curtain, one sees all there is to see, at least until hiding behavior is re-

triggered. As the experimental lines at Precision moved from top-left (high visibility, high

encryption) to bottom-right (low visibility, low encryption), privacy became temporary: not only

was the curtain permeable, but the operators appeared motivated to share their innovations

beyond the curtain when they were ready. This, ultimately, became not only a source of local

learning but of organizational learning.

Implications


P a g e | 37

With a few notable exceptions (c.f. Bechky, 2003, Gibson and Vermeulen, 2003), prior

scholarship has assumed that visibility of action and accessibility of knowledge are logically

collinear. This research provides evidence that they can be behaviorally orthogonal. Precision,

and perhaps other organizations like it, exists far more frequently on the off-diagonal: when

visible, workers attempt to maintain the proprietary quality of their knowledge through

encryption, limiting group-level creativity (Sutton and Kelly, 1997; Waterworth and Waterworth,

2001) as well as the scope of experimentation (Thomke, von Hippel, and Franke, 1998) and

improvisation (Barrett, 1998); when private, workers openly (at least within their groups) tweak

and experiment to improve their processes. It is impossible to force creativity or the transfer of

private knowledge, and attempts to do so through transparency to authorities can encourage

organizational silence, or the “widespread withholding of information” and a “climate of silence”

(Morrison and Milliken, 2000). Transparency may be correlated to several of Morrison and

Milliken’s antecedents to organizational silence (Morrison and Milliken, 2000) and may

reinforce individuals’ implicit voice theories which limit speaking up (Detert and Edmondson,

2011), while some degree of “constructive ambiguity” may provide greater openness to the kinds

of organizational flexibility and collaboration that provide “opportunities to learn” (e.g., Best,

2005). As suggested by the experience with behavioral responses to transparency at Precision,

privacy may indeed be an important and underutilized management lever for enabling continuous

improvement and performance.

Transparency intended to improve performance through increased operational control is

also subject to such behavioral considerations. Operational control can either be enabling or

coercive: formalization either “enables employees better to master their tasks or functions as a

means by which management attempts to coerce employees’ effort and compliance” (Adler and


P a g e | 38

Borys, 1996). The ideal would be the establishment of a transparent, enabling bureaucracy. Is

that an achievable ideal? With respect to design, transparent architecture can be implemented

with either a learning, usability orientation “aimed at accumulating data in all its diversity” or

with a control, highly-focused orientation “with the express purpose of direct supervision and

monitoring of employee performance” (Sewell, 1998). Sewell hypothesizes that the first

encourages “self-disciplining” team members under “unobtrusive control” through peer scrutiny,

while the second encourages the sort of covert attempts at avoidance typical of traditional

Taylorist coercive control (Sewell, 1998). Work at Precision would suggest that starkly

transparent design, without some allowance for group-level privacy, produces a level of

surveillance that de facto encourages a coercive orientation.

Perhaps the best example of a de facto coercive design is found in Foucault’s

characterization of Jeremy Bentham’s Panopticon, a plan for a prison in which large numbers of

convicts could be kept under surveillance by very few guards (Bentham, 1787; Foucault, 1977)

and the basis for Sewell’s panopticism (Sewell, 1998). By building prison cells in a circle around

a guard post, all the prisoners in Bentham’s Panopticon would be silhouetted against light

coming into the cells from the windows on the outside of the circle. Their movements would

therefore be visible to a single guard in the center. As Bentham explained and Foucault

emphasized, the system of control works even if there is no one in the guardhouse. Foucault

argued that the very fact of general visibility—being seeable, not necessarily being seen—would

be enough to produce effective social control. As Foucault wrote, “awareness of being visible

makes people the agents of their own subjection.” (Foucault, 1977) But even in prisons, which

so interested Foucault, lack of visible privacy often gave birth to privacy through code. At the

famous Number 4 prison in South Africa, where Nelson Mandela and Mahatma Gandhi were


P a g e | 39

once held prisoner along with thousands of others in overcrowded conditions with no privacy at

all, one of the world’s most complex number-based codes for communication among inmates

was developed (South Africa Constitution Hill Museum). In Russian prisons, tattoos often served

as records of the wearer’s gang membership and personal history and were highly coded, so

much so that few were able to decode all of them (Lambert, 2003).

While the Panopticon is a far more extreme manifestation of transparency than anything

Precision even cared to consider, Adler and Borys could nonetheless have been describing

Precision’s approach to transparent design when they wrote on coercive control: “the procedure

is neither designed nor implemented as an aid to the user…. as a result, covert and inefficient

‘workarounds’ abound” (Adler and Borys, 1996). Prior to the installation of the curtain,

managers operated under a myth of transparency and control, believing that what they saw was

in fact what they were managing; instead, coded interaction was everywhere, preventing true

operational control from becoming a reality. Installation of the curtain dissolved both the need

for the code and, with it, the myth, leaving actual organization-level control relatively unchanged.

But the level of control at the group level changed considerably: line-level transparency

increased, and therefore so did operational control and performance.

Despite this heritage of transparency research, our investigations in the management

sciences, while investing substantial effort in understanding the value of autonomy, or discretion

in decisionmaking (Hackman and Oldham, 1975), have done relatively little to investigate a

potential value of privacy, or discretion for unobserved local activity (c.f. Anteby, 2008).

Unobserved and uncontrolled local group activity can reflect experimentation, improvisation,

productive deviance, learning, and improvement (Edmondson, 2002), or it can represent

dangerous practical drift (Snook, 2000) and the organizational slack assumed by traditional


P a g e | 40

economics. Any conclusions about the value of transparent organizational design must be

contingent upon an understanding of these behavioral dynamics. While generating a rigorous

theoretical framework, Sewell concludes with a need for “theoretically informed empirical

studies” to investigate further (Sewell, 1998). This work at Precision provides an initial answer

to that call.

What does this mean for management? In an age when transparency has become gospel,

and when technological advances have made it more costly to guarantee privacy than to invade it,

management theory needs to recognize privacy as an important lever: for tweaking and

productive deviance, for experimentation, and for avoidance of unproductive distractions. As

managers, we must also be aware that deprivations of privacy may not result in more

transparency, but rather a more permanent and transfer-resistant form of privacy that preserves

the myth of control without incrementally yielding any. In organizations like Precision where the

primary movement is along the off-diagonal of Figure 2 (from high visibility, high encryption to

low visibility, low encryption), providing the appropriate level of visible privacy appears to be a

far superior strategy.

As human beings think and do, invent and systematize, innovate and execute, they

engage and withdraw accordingly. In his foundational Social Psychology of Privacy article,

Schwarz (1968) finds that privacy rules in groups constitute “a common bond providing for

periodic suspensions of interaction,” a phenomenon that seems to play out at the organizational

level at this research site. Based on the data, I would argue that such group-level privacy is

required “for incubation and time to discuss and elaborate their ideas with others”

(Csikszentmihalyi and Sawyer, 1995), for exploration, experimentation, and prototyping of

improved processes prior to sharing (March, 1991; Thomke, 1998), and to avoid interruptions


P a g e | 41

with negative consequences (Jett and George, 2003). At PrecisionMobile, visible privacy turned

out to be a management lever that had been underused to increase organizational learning and

performance.

Limitations and Future Research

The field-based nature of this experimental research required focusing on a single site,

making generalizability this study’s largest potential limitation. This site is notable for both its

Chinese location and its manufacturing focus. Several recent books and articles have described in

detail the personal ambitions, challenges, successes, and failures of migrant workers in China

who have fueled the Chinese outsourced manufacturing miracle (Chang, 2008; Chen, 2008),

while a number of other authors have contributed significantly to our understanding of

management and organizational challenges specific to China (Tsui, 2007; Barney and Zhang,

2009). To the extent that the results presented here are influenced by the unique characteristics of

Chinese manufacturing and Chinese migrant workers, generalizability beyond that context will

be limited, although these features characterize a majority of manufacturing production in the

world. There are, however, reasons to believe that these results are more universal to

organizations in other contexts. This highly operational context is one of the more conservative

settings for an experiment of this kind. More so than most other areas of management theory,

operations theory would predict transparency to be valuable for learning and performance,

particularly on a factory floor, offering some indication that similar logic would hold in less

industrial settings. Furthermore, similar outcomes have already been observed at another

Precision site in Guadalajara, putting into doubt that this is a China-specific phenomenon. Future


P a g e | 42

research, in either other contexts or the laboratory, can more scientifically study how the results

here may or may not generalize to other settings and geographies.

Similarly, because this research was limited to one organization to maintain the rigor and

comparability of the field experiment, there was no variance in organizational culture and

therefore no data to investigate the influence of organizational culture on the relationship

between transparency and performance. There are reasons to believe organizational culture could

significantly affect that relationship. Research on Toyota, for example, suggests that knowledge

transfer and organizational performance benefit from transparency in the presence of a

complementary “culture of transparency” (Spear and Bowen, 1999). Given the inability of most

organizations to replicate the success of TPS (Spear, 2009) and Toyota’s own recent inability to

sustain it, organizational culture does not alone resolve the transparency paradox, but there are

likely significant opportunities for future research on the role of organizational culture in

moderating this relationship between transparency, learning, and control, on one hand, and

performance, on the other.

These findings about privacy are also specific to the group level of analysis. Privacy at

the individual level may have different performance implications. Mas and Moretti, for example,

have found through a very careful study of supermarket checkout workers that individuals

became more productive when they were within direct sight of higher productivity workers (Mas

and Moretti, 2009).

Finally, due to a limitation at the site, we were unable to take the experiment full-cycle by

removing the curtains and observing the impact to performance, a fourth limitation of this study.

August through November is the busy season for PrecisionMobile, as over one-third of annual

sales of mobile devices are estimated to occur during the winter holiday season. As cooperative


P a g e | 43

as PrecisionMobile was, they were understandably reluctant to make any major modifications to

the lines during peak season. Once the window for making changes reopened in December,

product cycles had run their course, product mix was changing rapidly, and the factory was once

again reconfiguring itself—of the 32 lines we studied, only 8 were scheduled to remain

operational by the end of the month. In January, ordinary attrition also ran its course: after the

Chinese New Year holidays when all workers traditionally visit home, fewer than half of the

operators on the studied lines had returned to the factory, with new faces taking their place. As a

result, comparability between time periods became impossible to achieve. Future research may

be able to look at the on-and-off cycling of visible privacy to more precisely understand the

relationship among forms of privacy, learning, and organization design. A particularly fertile

avenue of investigation would appear to be a study that asked not whether there should be a

curtain, but rather when and for how long particular kinds of groups should be given their time

inside the curtain.

Conclusion

The findings at PrecisionMobile carry important practical as well as theoretical

implications. Existing organizational learning research celebrates intraorganizational

transparency and reduction of knowledge boundaries (Argote and Ophir, 2002), while network

tools have increased our ability to observe, analyze, and rewire intraorganizational networks to

be more connected (Cross, 2004). The research at PrecisionMobile suggests that the very efforts

to increase transparency in organizations may, in fact, be limiting it. In an environment flooded

with calls for transparency, or accurate observability, groups of individuals may not want to be

accurately observed and may act to prevent it. Thus the removal of visible privacy boundaries


P a g e | 44

within an organization can motivate the creation of encrypted privacy, which is both more

permanent and more costly in knowledge transfer, operational control, and productivity. In effect,

managing by walking around becomes inferior to managing by standing still, because the latter

permits improvement rights to vest, if only temporarily, with those most able to use them to

enhance performance. Yet the field experiment at PrecisionMobile also indicates that the reversal

of this trend is not only possible but relatively simple to implement. For optimal levels of

learning and productivity, a pragmatic model of organizational transparency should incorporate

spaces and/or times of visible privacy.


P a g e | 45

Table 1: Selected Examples of Hidden Variance on the Lines (Quotes from Operators and/or Embeds)

Task Process-in-Theory Process-in-Use Operator Rationale Training Attach Protective Chip Covers with Identifying Unique Bar Code Stickers

“The correct way would involve using both hands to remove one label from the roll, picking up a lid, taking out a board from the rack and then setting it down and putting the cover on with both hands. Then you need to set the board down, pick up the roll and repeat that process 3 more times. Finally you have to put the board back onto the rack. After the whole rack is done, then you have to take each board out to scan into the computer.” (Embed, based on the TQC chart)

“My job and my trainer’s job was to take these stickers that were given to us, keep a record of how many stickers we got, keep track of the numbers on those stickers. We take logs of those. Then we scan every single one of those stickers into the system and... put each sticker onto these little metal casings. The metal casings come in big trays of 64. The other two women, the women from Szechuan, would stack all of the boards that they had checked in a shelf-like thing in groups of 15, and I would take those boards and put the casings with the stickers on.” (Embed)

“The short cut is significantly better because, first, all the bar codes can be applied with one hand at an amazing speed. At least 4 times faster than if you were to apply one by one. Second, it is 10 times easier to scan when all the bar codes are aligned on the roll [of stickers] than already on boards--we don’t need to flip the boards, and we save time otherwise lost to aiming the laser (UPC reader) correctly. Third, we take the boards off their racks only once instead of twice (i.e., first time to apply the covers, second time to scan).” (Embed) “If you didn’t do it their way, they say you didn’t follow the rules and you should do it their way. If we follow their rules, there would be a whole lot piled up and they would complain about us being so slow. If we tell them that this is going to pile up doing it their way, they’ll say that they will just add one more person, and if there are two people at a station, then there won’t be enough to do, so we’ll be standing around a lot, and so they’ll come and complain.” (Operator)

“Basically, she [the floor head] took me to the line and told the line leader that I was being put into their line. Nobody questioned that or said anything, they just said okay. And then she randomly picked out this one girl who was standing at the end of the line and said, ‘are you putting covers on those boards?’ and she said, ‘yes.’ And then the floor head said, ‘take off your training badge and give it to JH, and teach her everything your teacher taught you.’ And then she left and my day started.” (Embed) “This morning when I was putting the lids on, my trainer told me to do it their way, as told to me yesterday by the mean pink hat. So I put the lids on, and then put the barcodes on. I asked her why we were doing that, and she said: ‘well, she’s hovering around. If we’re not busy, we’ll do it her way so we don’t get yelled at again, but if we do get busy, then we’ll go back to doing it our way.’ And if we do it their way, it does create two more procedures and is a lot slower.” (Embed)

Manual Inspection of PCBs

“We were supposed to look at the board and the model, compare, and if it matched, put the cap on.” (Embed)

“The others were trying to teach me how to check two cell phones at the same time, which I don’t think you are supposed to do. The girl was all for it, because then I would be faster and she would have less to do, but the guy said, ‘no, she’s new, you don’t want her messing things up,’ so I didn’t do it.” (Embed)

“Faster, less taxing production” (Embed)

“Some lady just came around and said ‘oh you guys got a new person?’, they said ‘yeah,’ and she said ‘be nice to her, teach her well, be very detailed, don’t teach her all your bad habits, teach her the right things’ and everybody laughed.” (Embed)

Putting Leads on the PCBs

“I was doing it kind of slowly, carefully, and gently.” (Embed)

“She [my trainer] is a lot rougher on the boards—she takes them over and flips them around quickly, I do it slower and I use 2 hands to do it.” (Embed)

“We need to be able to work fast under pressure, when necessary.” (Operator)

“She said ‘don’t do that… right now you are fine, but when you have boards piling up, you are never going to finish it, you are going to be under lot of pressure, you need to practice now being really fast to do it. Like this: one hand here, two fingers here, flip….’” (Embed’s Trainer)

Protective Gloves

“Everyone on the line is supposed to wear gloves on both hands and/or rubber fingertip covers.” (6S Rules)

“People usually wear their gloves in their own ways—either they wear a glove on only one hand or they cut their gloves so their fingertips stick out, giving them a bare hand or bare fingertips so when they are doing little things it goes a lot faster.” (Embed)

“Not all tasks require protective gloves—the management just do that for show. And for tasks that don’t need gloves, production goes much faster and more effortlessly with bare

“Once when the customer came, people were passing around the word, so they all changed gloves. So you could then see that everybody was wearing a full pair of gloves. And then


P a g e | 46

fingertips.” (Operator) once the customer was gone, they just went back to the way they were wearing them before.” (Embed)

Backend Assembly and Packaging Task Design

“The facility is designed for 3 second takt times, meaning each task should be 3 seconds long.” (Senior Management)

“Their tasks were definitely not completed in 3 seconds. In fact, I counted – it was more like 4-5 seconds. And each task was different.; We didn’t have one task, like it takes (1…2…3…4…5…6…) 7 to 10 seconds for my trainer and longer for me.” (Embed)

“Better motion, less fatigue, faster production with the same resources” (Embed’s Trainer)

“If someone is having build-up, they would call someone from down the line to help.” (Embed)

“I have a feeling that there is one TQC for the station, each station is designed to need only one person.” (Embed)

“But for all of those things that are being tested, there are three people there and they sort of divide all the tests among themselves. It’s very fluid.” (Embed)

“Faster reduction of bottlenecks” (Embed)

“If one person is having too many, like today we were about to get up for lunch and everybody was hurrying and the girl left for the bathroom or something so I had a couple piled up on my side, and then everybody just helped so we can finish and leave.” (Embed)

“Put the cell phone in the box, scan the box, and pass it along to the next person.” (TQC Chart)

“After I stood there for 10 minutes, she started grabbing the phone, scanning it, and automatically giving that phone to the person next to her (without putting it in the box), and I don’t think she was supposed to do that. But it speeds up the process for the girl next to her, because she gets the phone earlier and she can put that in the plastic bag and just wait for her box which she puts the sticker on.” (Embed)

“Especially when they’ve worked on one station for a long time, they tweak the steps here and there and find a faster way of doing it.” (Former Line Leader, Training Manager)

Testing “I’m supposed to be testing the cell phones. I’m supposed to plug in both headphones and talk into the cell phone to test whether the headphones work – if I can hear stuff.” (Embed) “We were supposed to test for the keys, we were supposed to press them one by one and see if the correct numbers register.” (Embed)

“For convenience sake, I usually only use one earbud, and that was the way I was taught, that’s the way my trainer does it.” (Embed) “The guy that was teaching me would scrape his hand like this across the keys, rather than pressing them one by one. So I learned it that way. So I was doing it that way and then little bit later, he turned around and he was like: ‘oh don’t do it that way, that’s not the way you are supposed to do it, don’t do it like me, you’re going to get yelled that.’ He said: ‘I am going to teach you the right way, you are supposed to press them one by one and if you choose to do it my way, that’s your choice.’” (Embed)

“Because with only one earbud in, you can just unplug it, and immediately listen to the cell phone in the other ear to test the speaker on the phone itself.” “Being a QA, I see what the process is like on the line, and I see these documents that clearly state how things should be done, but they’re not followed because they want to hit the production mark before anything else. ‘We need to rush production!’ I hear that a lot.” (QA)

“Then the yellow hat came around and I was doing that [testing with only one earbud]. After the yellow hat left, my trainer said, ‘Didn’t I already tell you this?! Every time the yellow hat comes, you need to test with both ear buds. You talk into the cell phone once and listen to the left earbud, and then you talk into it again and listen to the right earbud.’ And she said, ‘when the yellow hat isn’t there, you don’t need to do it, but you should do it every time the yellow hat comes around.’ And then the guy next said, ‘oh, so you’re teaching her all of the bad habits, eh?’ And she responded, ‘so what, you taught her how to not do the keys one-by-one, you told her to just run her fingers across the board.’” (Embed)

Line Operator Allocation

“Each station is allocated 1 operator.” (Management)

“I counted today—on the backend portion of the line, we had 17 people and 13 stations on our line today.” (Embed) “There are many stations that are supposed to be managed by one person, but in fact we have to put an

“They expect us to finish so much in so little time. We need these extras to actually meet the target. Think about it, when they increase our UPH targets, the kaizen engineers, supporter, line leaders come and help

“The floater took us along with four others to the end of the floor. She explained that someone is coming to visit the lines, and because we are the new workers, we are the ‘extra people’ on the lines and need to be taken


P a g e | 47

extra operator at each of those stations in order to get work done. These stations are the engraving / accessories, boxes / sticker, and away-from-line jobs (putting instruction and cables in the boxes). All of these stations are supposed to only be operated by one person, but we assign two at each one every day. When the management come around and check, we have to get the ‘extra’ operators away from the line so not to be discovered.” (Line Supporter)

us with production. Once we reach the target, they all leave. For that UPH to be sustainable, we sometimes have to have extras. The TQC say that each station can only have one worker, but what’s wrong with people helping each other when they have extra time? But these things we need to hide from the management.” (Operator)

away. She took us to this rest area and we were supposed to stay there as long as that important someone is still around the lines. Someone asked the floater asked again why we had to leave, the floater said, ‘we really are only supposed to have 4 people on each line, but there are about 6 people on the lines, obviously more than necessary according to the rules.’” (Embed)

“We are constantly re-balancing the lines to reduce bottlenecks and cost.” (Management)

“There was the first girl, then the second girl, and then me. And the second girl said to the first girl, ‘why are you working so fast? You’ve got to wait—I’m not keeping up.’ And then the girl did slow down. Because they don’t want there to be a lot of USB cards on their table, because that looks bad for them.” (Embed)

“Because each station is only supposed to have one operator, they cannot help each other when they fall behind, at least when someone is watching—instead, they have to slow down.” (Embed)

“I had put so many boxes on my table, so it looked a little messy, and the girl right next to me poked me and told me that I need to clean it up… otherwise, the production leader is going to yell at you.’ So I started putting things back… and the production leader came over and told me, ‘don’t accumulate this many boxes on the table.’” (Embed)

Line Appearance

“Lines are always supposed to be busy.” (Floor Head)

“In reality, there is a lot of downtime--for maintenance, because of a problem, because you are waiting for materials, because you are waiting for paperwork… there are lots of possibilities.” (Embed)

“There were some foreigners today walking around, and because we were doing maintenance, there was nothing to do. So my trainer told me to take out the graphs and pretend to look at them. She told me in the morning—‘I think there are customers coming. If they come, you take out this graph, and you hold it like this, and you pretend that you are checking the graph with the boards, okay?!’ And so I responded, ‘ok.’ And I think I was leaning on the table at the time, and my trainer leaned over and whispered to me, ‘stand up straight!’ So I did what she said and stood up straight and stared at the log book, like she was doing.” (Embed)

Line Reporting “In front of each line, there is a production status indicator and a whiteboard with the hourly production (targeted and completed) for the shift. It also includes reporting for any quality failures (rejects) on the line.” (Floor Head)

“What I thought was really strange is that there was a chart in front of the line that’s supposed to say how many you have produced this hour and what your goal is, and she updates that every hour. And I thought it was very strange the numbers are perfect, everything was 100% -- it was just perfect. I asked the operator writing on the board where she gets her data from, and she said she gets it from the guy at the last post in the line. He makes the final check on the phone and scans the barcode into the computer. And he keeps track of how many boards he puts out. I have a feeling that he tampers with the numbers, because only 2 days ago, I was one station away from him. I don’t think he’s actually counting them one-by-one, I think he controls

“When I asked [the operator writing on the reporting board] about the validity of the numbers, he said, ‘if the actual doesn’t equal the target, then managers will come over to complain. But overall, we’re on time, aren’t we?! We never fail to have enough at the end of the day– we always meet our goals.’” (Embed)

“Our blue hat always writes down 240. And he never… I actually pay attention to what he does… he never actually looks at what it says on the PrecisionFlow screen. He just notices that it’s about the right time—let’s say it’s 9:56—and he will write down for the block from 9-10 that we made 240.” (Embed)


P a g e | 48

the numbers throughout the day—sometimes they produce more, sometimes they produce less. And he definitely verbally tells people when we aren’t producing enough. He would say: ‘hey, speed it up.’ Or ‘hey, slow it down.’” (Embed)

Circuit Board Automated Production (SMT)

“SMT (surface mount technology) are very complicated and involve the most expensive materials (the chips). When an SMT machine breaks down, you are supposed to call an engineer for help and wait until he arrives.” (Embed)

“This morning, in one of the machines, one of the circuit board for some reason fell out of where it was supposed to be. My first trainer just opened up the machine cover and put one foot into the machine to retrieve the circuit board. I thought that was interesting because they didn’t bother to tell the blue hats or the engineers before she did that.” (Embed)

“Keep production going” (Embed) “I’ve learned much of the stuff that engineers do by watching them do it. It's not hard.” (Operator)

“Around lunch, the SMT machine I was working with broke: one of the components (line people called it “worm”) could not be sucked by the machine correctly. So my trainer asked one of the engineers for help.” (Embed)

“While waiting for the engineer, my trainer checked the machine for a while, but he couldn’t fix it himself. In order not to slow down the production rate—because today we had a specific goal of how much we need to produce—my trainer started the machine and stopped it frequently to put the materials that the machine failed to suck in back to the reels, so that these materials didn’t get wasted. The engineer did not like the way my trainer did it, he asked her not to do it. And my trainer was annoyed, retorting back saying, ‘you can’t even solve the problem, the machine is still broken.’ The engineer and my trainer didn’t talk for a day afterwards. They used to flirt and joke with each other a lot.” (Embed)

“Once these machines are broken, we will have to delay our production and often times we can’t hit the target. However, the management above us won’t give us chance to explain the delays, so we never get permission to reduce the production goal for the day. As a result, we need to find ways catching up with the missing productivity due to machine problems all by ourselves.” (Operator)

“Since some of these engineers have only been on the floor for a few months, they in fact ask us how to fix it.” (Operator)

Automated Optical Inspection of Completed PCBs

“Boards are inspected optically by an advanced machine to make sure all of the chips are in exactly the correct place after being placed on PCBs by the SMT machines. When an error is detected, the machine sounds an alarm to the engineers and waits for their attention. If the engineer inspects the board and all appears okay, then the error is in the inspection software, and the engineer clears the error by pressing ‘Esc’ on the keyboard.” (Engineer)

“Engineers rarely come quickly enough, so operators handle it themselves. When multiple boards trigger the alarm, they inspect the first one, and if it is okay, they assume subsequent errors are also software-related and just press ‘Esc’ over and over to clear the errors and avoid a stop in production.” (Embed)

“It doesn't cause any problems, and it keeps production moving.” (Operator)

“We just do what the engineers would do--we hit Esc if everything is fine.” (Operator)

ESD (Electrostatic Discharge) Prevention

“At designated stations, we were supposed to work on this surface that reduces static electricity and/or wear a bracelet connected to the workstation that keeps the operator grounded.” (Embed)

“When I was working today, my trainer said, ‘You’re too slow. Don't work on the ESD surface. Instead just work on top of the tray that keeps all the little covers.’ So I did that. And later on, my trainer told me: ‘whenever the customers come around, don’t do that, because they’ll get mad.’ And I asked her what the customers look like, and she said, ‘they wear blue robes, and they’re the only ones that wear different color robes, so just look out.’” (Embed)

“Keep up with production speed of co-workers” (Line Leader)

”I’m not quite sure if she’s actually a trainer, but perhaps just a random person on the line. But I think that is a good idea—it was her post that she was teaching me now, so she has all these little tricks, like putting the board on the tray instead of on the table—that really does speed things up. I’m a lot faster now.” (Embed)

“What I thought was really interesting was that the pink hat [line supporter] also disobeyed rules. She needed an ESD bracelet for one of her posts, and she didn’t have one with her, so she asked some guy in the

“Better mobility around the line” (Embed)

“They gave a little sheet of paper, and I was supposed to memorize it. It’s kind of ironic—it’s a sheet of paper about static electricity and how to


P a g e | 49

line—‘hey, can I use your bracelet?’ Obviously, if the guy has a bracelet, he needs to use it, but she knows that he doesn’t use it, so she took it from him. And he said, ‘hey, I still need that’… but I don’t think he was serious about it. And she replied, ‘okay, if anyone comes to check, I’ll give it back to you, okay?’” (Embed)

avoid it, the ways of taking care of it. But at the same time, the operators are not using the ESD surface and bracelets.” (Embed) “And when I was doing the tweezers thing, the girl asked me if I had a blue wristband—the thing that takes static electricity away from you—and I said I didn’t, should I go ask for one? She said, ‘no, you don’t have to, it’s just that someone is coming over to check, and I don’t know if she’s going to yell at you or not. That’s the only worry.’” (Embed)

Materials / Kanban

“Materials are brought to the line in volumes that do not exceed 1 hour of production--or, in some cases, two. The line production pulls materials; the materials people do not push line production.” (Embed, based on training materials)

“Our job is to make sure that production is normal and the line doesn’t stop. Basically, you deliver your materials, try to do it efficiently and try to do it fast, try to get yourself on a schedule, be smart—see ahead of time how much they need. So your job is easier, and you don’t want to mess up anything, because if you mess up on something, the whole day you are behind and you’ll always be running around. So, basically get yourself on a schedule and besides that, you stand over here in front of the kanban store for a while, look like you are doing something, and then stand on the line for a while and look like you are doing something, check underneath the tables, collect garbage once in a while, and then there you go, you repeat the process again, you scan the codes, you get the paper and you get the materials and send it out.” (Materials Operator)

“Because if look at this from a materials point of view, you would want to give out as much as possible, so you don’t have to make multiple trips and keep on giving them, right? So that’s what they try to do and that’s the limit that they try to push.” (Embed) “One other reason why they don’t keep kanban, especially with smaller materials like stickers, domes, or LCD screen glass covers, is that those materials are small, and they come in packages of let’s say 1000 or 500, and if UPH requirements are 300, it makes it very difficult for them to open up the package and only send in 300.” (Embed)

“I was asking my trainer, ‘why do they even have this kanban stuff? Why do we have to add material according to UPH?’ And she replied, ‘personally, I think that the bigger materials, perhaps if they control UPH, there is a reason – so we don’t crowd out the tables, so we don’t have too many materials piled in the tiny moving area, because the lines are really cramped. You know about 6S, right? So it doesn’t look messy. But the smaller materials… they just want to create more work for us, when they control the smaller material.’” (Embed)

“The materials handlers have a ton of little tricks that they do that they are not supposed to do.” (Embed) “She also knows all of the tricks of putting the stuff into the shelves so that it doesn’t look like a lot. She always said, ‘ok, so you put the lenses in a pile in a box and fold it like this, so it doesn’t look like a lot. And if you’re trying to put the paper boxes, you stack them like this so it doesn’t look like a lot.’” (Embed) “Also, later on in the evening, just before we were supposed to get off work, my trainer made a mistake. Somehow, the UPH plan chart that she got was much different from the UPH chart that the line got. So she ended up working according to her chart, and kept on asking for materials and kept on sending it to the line. But then, pretty soon she realized that they still had a lot of materials, and she was sending way too much. She could not send it back to kanban store, because

“The materials trainer told me, ‘all the rules are dead, but people are alive. We can’t be stupid and follow those stupid rules. Because when people first get on the floor, production is usually really, really slow; in the middle of the day, it peaks; then, right when they are about to get off, things slow down again. It’s all up to the materials person to really get to know the line, and get to know the pace, and control it that way. You can’t say, ‘one hour, it must be xx number of parts.’ Because then materials will pile at the beginning and right before break, and they will definitely run out during the middle of the shift.’ And having worked on the line, I know that’s definitely true.” (Embed)

“We need to keep kanban, that one I’m not even going to say again. I know that’s not going to change anything, but I still have to say this.” (Senior Manager)


P a g e | 50

they wouldn’t take it back. And they are not allowed to keep stuff on the shelves in front of the kanban store, because people actually check that, and during the meeting the pink hat said, ‘you shouldn’t keep things there for too long – more than an hour is really ridiculous.’ So what she did is—there were empty lines around the area that weren’t being used, so she stuffed all of her extra materials into the shelves, in some corner where no one is going to notice. And said, ‘ok, we’ll just hide this stuff here for a day, and then tomorrow, we’ll use it.’ And apparently, she does that quite frequently if there is too much material.” (Embed)

“I think they do recognize that people check for kanban. That’s why they say the smaller materials is okay, we’ll stuff it in the box so nobody will see… well, that’s true. You’ve seen how close, there are so many things going on. If this tiny pack of tiny little stickers, if you stick it somewhere, nobody is going to see that.” (Embed)

“When a line changeover occurs, all materials are supposed to be returned to the kanban store for re-cataloging.” (Senior Manager)

“And in the middle of the day, his line changed model. They took all of the materials from the line, except they didn’t put it back in kanban, because apparently there’s another line still making the same product. So instead, there’s a tiny room, I think it’s a kanban store kind of place, and they put everything in that room. So it’s all half-used materials, packages are open, etc. So my trainer took everything from that room the next day until the leftover materials had been used.” (Embed)

“When materials people collect the garbage of the line, they first make sure that no materials are left in the packaging. Then, when they throw out the garbage, they are supposed to check again with a guard watching.” (Operator)

“And when we actually go dump that stuff, we are supposed to dump it one by one and check if there is no material left, and she did that the first time because there are security guards looking, but then later on the security guard wasn’t there. So she told me: ‘you are not supposed to do this, but if the security guard is not here, we don’t have to look because we already looked on the line.’” (Embed)

“When we’ve already checked it once, there’s no value to checking again. It’s purely a waste of effort.” (Operator)

6S (Sort, Straighten, Sweep, Standardize, Sustain, Safety)

“There are very strict hygiene rules for the factory floor, which are consistently enforced through training, workflow design, audits and surprise visits by 6S teams.” (Site Training Documents) “And of course on the workspace there’s tape clearly labeling that you’re supposed to keep the records there: keep the good ones there, the bad ones here, records here, etc.” (Embed)

“6S guidelines are not really followed. Operators do whatever they feel like. But when they were doing a maintenance, and there’s talk of people from above coming to look, everybody was cleaning. We had to wipe the tables, pick up crap from the floor, we did open the oven and wipe it down with alcohol.” (Embed)

“When I was in B11, one of the customers was coming, and the entire day was spent on cleaning, and it was like new years! But usually, it’s never like that. They make it such a big deal.” (Embed) “I also saw a few white hats chewing the gum we got at lunch while working. I asked the girl on my right if I could chew gum, and she said, ‘oh yeah sure, just make sure after you finish it, wrap it with the working paper (like those paper for reporting quality and quantity) and then throw it in the trash bin, this way people can’t find out that’s not actually gum.’” (Embed)

W

T

WORKING PAP

Table 2: Net

Line A

Line B

PER

twork Surv

ey on Factoory Floor In

ndividuals S

E.

elf-Identifie

LRB

CTS

. Bernstein, Th

ed as Impor

Legend: Red = IndirecBlue = Direct

Circle = SurvTriangle = IdeSquare = Line

he Transparenc

P

rtant for Wo

ct Labor t Labor

veyed Ego entified Altere Managemen

cy Paradox

P a g e | 51

ork

r nt


P a g e | 52

Explanation of Network Survey Method: All direct labor (operators) from the two, end-to-end assembly lines on which the embeds had worked in Study 1 were asked to complete a paper survey including the question: “Please identify up to twenty people who are important in terms of providing you with information to do your work or helping you think about complex problems posed by your work. These may or may not be people you communicate with on a regular basis and can come from within Flextronics or outside (e.g., customers, suppliers, family, etc.). Recent interactions in June 2008 should help you choose. You do not need to identify as many as 20 individuals--only as many as you feel are relevant.” In addition to identifying the alter’s name, they were also asked to answer questions about the tie, including physical proximity (e.g., same floor, different floor but same building, etc.), primary form of communication (e.g., e-mail, phone, face-to-face, etc.), length of acquaintance (e.g., less than 6 months, 6-12 months, etc.), relative hierarchy (e.g., senior, same, junior), frequency of interaction (e.g., times per day, week, or month), purpose (e.g., information gathering, problem-solving, cooperation), initiation (e.g., ego, alter, or mutual), and value of the interaction (e.g., agree or disagree that more interaction with yield greater work effectiveness). A similar survey was then administered, via an intranet-based online survey, to all of the indirect labor (engineers, supervisors, managers, etc.) who were responsible for each of those two lines. On both surveys, the inclusion of both English and Chinese names, along with the tie-based data referenced above, enabled a higher likelihood of accurately matching individuals across responses. Results: The survey had a total response rate of 98% (n=172) for the direct labor and 82% (n=678) for the indirect labor. The survey of Line A resulted in 1131 nodes with 1859 ties. The survey of Line B resulted in 1211 nodes with 2422 ties. The diagrams above are representations of the two self-reported sociometric networks. In each case, the layout was automatically calculated based on Netdraw’s standard algorithm (“layout with node repulsion and equal edge length bias”). The diagrams were then rotated so that their orientations matched (i.e., IDL nodes on top and DL nodes on the bottom), and the two clusters were slightly separated to show the bridging nodes most clearly. Overall, only 1.45% of the ties on Line A and 1.36% of the ties on Line B identified interactions between IDL and DL nodes.


P a g e | 53

Table 3: Line Performance—Experimental and Control Lines

Note: Performance was measured as defect-free hourly output (units per hour) during full hours of production.

170

180

190

200

210

220

230

240

250

260

270

280

5/1/2009 6/1/2009 7/1/2009 8/1/2009 9/1/2009 10/1/2009 11/1/2009

Ave

rage

UPH

(Uni

ts P

er H

our) Control Lines

(N=28)

Experimental Lines(N=4)

15 per. Mov. Avg.(Control Lines(N=28))

15 per. Mov. Avg.(ExperimentalLines (N=4))Curtain

Installed


P a g e | 54

Table 4: Line Performance Descriptive Statistics

Line Performance Descriptive Statistics, Grouped by Treatment and Control Groups Metric: Units Per Hour (UPH)1

Mean Standard Deviation Median

ALL

Control Lines, Pre-Curtain 219.071 12.679 219.359 Control Lines, Post-Curtain 218.158 16.359 217.881 Treatment Lines, Pre-Curtain 216.013 16.212 217.820 Treatment Lines, Post-Curtain 240.027 24.924 238.352

DAY


NIGHT


INITIAL DAY

(Rotating)2


INITIAL NIGHT

(Rotating)2


Notes: (1) UPH (Units Per Hour) is the average hourly zero-defect production of each line over the relevant time period. The Pre-Curtain period includes 55 operating days from May 1, 2009 through June 27, 2009. The Post-Curtain period includes 142 operating days from June 28, 2009 through November 30, 2009. There was no production on May 1, May 28, and October 1-3 due to national statutory holidays in China. There was no production on June 28-30, August 1-2, August 29, October 6-8, and November 9 due to line maintenance and inventory counts. There was no production on May 3 and July 19, both Sundays, due to insufficient production demand to drive Sunday overtime. (2) Operators rotate, as a line, between day and night shifts roughly every month. From May 1-November 30, 2009, the lines being studied shifted between day and night shifts on May 28, June 30, August 1, August 29, October 6, and November 9. The “Rotating” figures above reflect production keeping the operator group constant (i.e., it tracks the production of a specified set of people who were initially on the day or night shift at the start of this study). The “Day” and “Night” figures above track production on those shifts, regardless of the rotation of the operators.


P a g e | 55

Table 5: Difference in Differences Models

Dependent Variable: Units Per Hour (UPH)

Difference in Differences Models

Grouped Daily

Average

Grouped Daily Average w/ Night/Day

Daily Average w/ Line Fixed

Effects

Daily Average w/ Line and Day Fixed

Effects Hourly w/ Line Fixed Effects

Hourly w/ Line and Day Fixed

Effects

“Pre” vs. “Post” Average Line Performance (Bertrand et al)

(1) (2) (3) (4) (5) (6) (7)

dExperiment -3.05886 -3.003139 -13.84052*** -12.87122*** -10.88213*** -11.07775*** 1.960865

(2.011991) (2.110287) (4.19621) (4.412268) (4.260413) (4.313327) (6.733517)

dTime -0.9136098 -0.9136098 -1.03688 -1.029706 -0.5343025 -3.668409 -5.270521

(1.571069) (1.589696) (2.032045) (6.244247) (1.455715) (34.16294) (3.994127)

dExperiment * dTime

24.92832*** 24.86162*** 16.28714*** 16.73595** 14.56742*** 14.51727*** 21.24569**

(2.72689) (2.772862) (3.646308) (3.677667) (2.643314) (2.69977) (9.566616)

dNight 5.683536*** -10.2238* -8.850901 -67.12926*** -65.32735***

(1.408708) (6.130399) (6.48501) (6.2139) (6.701037)

Constant 219.0714*** 216.2296*** 217.4635*** 216.0429*** 215.8333*** 195.2289*** 203.2387***

(1.217646) (1.405507) (2.89862) (6.368186) (3.443887) (12.55065) (2.748947)

R-squared 0.2307 0.2474 0.1084 0.2186 0.0299 0.0531 0.2206

Observations 741 741 2799 2799 25040 25040 46

Notes: *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively. Models are GLS fixed-effects models with heteroskedasticity-robust standard errors. Models (3)-(6) include fixed effects for lines which are not shown. Models (4) and (6) include fixed effects for day which are not shown. Model (7) is a robustness check based on Bertrand et. al.’s (2000) suggested remedy (for a small number of groups) for potentially inflated standard errors of difference-in-differences estimates due to serial correlation with data from a large number of periods.


P a g e | 56

Table 6: Selected Examples of Experimental Innovations on the Curtained Lines 1 Lower the level of the computer screens on the line by 3 inches. "Right now they are so high that everyone has to look up to see the

screens, which takes extra time and reduces accuracy." 2 Increase working surfaces at certain stations. "Some of the working stations are so small that materials fall off too often. At some stations,

we should make the working spaces larger." 3 Optimize working surfaces. "Move the shelf next to me higher so that my materials boxes can fit there." "There are three levels of shelves,

most of which are never used. Change into two layers and integrate with the shelves holding the computers for testing. That will make the space more useable and reduce the surfaces we need to keep clean."

4 Relocate keyboards to be more convenient. "Keyboards are situated at awkward places given the motions we need to do 2400 times per day. Relocating the keyboards closer to the computers would also allow us to reclaim more workspace on the counters."

5 Connect FVMI (Final Visual Manual Inspection) and CS (software testing) stations. "Reconfigure line so that FVMI is together with CS to reduce trips FVMI operator has to take to pick up materials from CS."

6 Change order of CS and engraving stations. "Currently, after the software on the card is upgraded, the plastic cover of the card is engraved with its unique EIN and then the card is tested to make sure the software works properly. Instead, we should test the software before engraving the EIN number. This would both avoid destroying the plastic shell if card fails the CS test and would avoid build-up at CS when the engraving machine is busy or down for maintenance. In fact, we do this when the engraving machine is down, but it would make more sense to just to this all the time."

7 Integrate reports to save time. "Several of the manual reports could be integrated into one, saving time."

8 Improve location and design of accessory storage. "The accessory storage is too high, which makes it harder to reach." "The accessory parts are in containers that are not attached to the workstation, so they move around and keep me from being able to take out parts as quickly. It would be really easy and really helpful to attach the containers to the back of the workstation, even if just with little plastic ties."

9 Get better protective goggles for the engraving station. The current goggles are "heavy to wear, too big for our Chinese faces, and when you put them on, you can't read the engraved number clearly because they are not clear enough. Since it’s hard to see with the goggles on, there was once a girl who worked at the engraving station ended up engraving the bar code twice on the cover for a few cards, so all of those cards became defective.... So right now, we just don't wear our goggles unless the 6S people are around, but it would be good to have better goggles instead."

10 Add a simple venting duct below the engraving machine, with more powerful suction, to redirect the hot exhaust air and suck away the harmful particles left over after engraving the plastic. "The fans underneath the engraving machine should be sucking air more forcefully away so that particles remaining from previous engraved card don't affect the next one." "It would also help to keep us from getting unbearably hot while sitting there."

11 Improve operator comfort. "I hope to install some sort of bars underneath our table so that we can rest our feet on it once a while, it’s just simply too tiring to stand all day long."

12 Modify the U-shape of the line. "At first, the line shape was a LONG line, it was extremely difficult to manage. So the line changed into two U shapes, one for the first half of the line, the other for the second half of the line. Now, it has changed to one U shape for the whole line. I still think the shape we are adopting now is hard to manage because there’re blind spots in the corners where I can’t see what people are doing. As of right now, I haven’t thought about what shape to change it into, but I think it’s something we can work on."

13 Change the allocation of workers at certain stations. "Some stations should probably have more people, and other stations should probably have fewer. Specifically, the people who are doing packaging/stuffing, they want 1 more person, because it is too many independent operations per individual at the moment. And the scanning operators feel like maybe we can somehow rearrange the people so that we can take one person out from there."

14 Allocate a second worker to the stickers station (often identified as the bottleneck), while still maintaining team responsibility for quality control. "You know the station for putting two stickers on the folded boxes, forming/unfolding the boxes, and packaging the boxes? It used to be that one operator off the line could put the first sticker on the folded boxes, and the operator working at the station would only need to put the second sticker on each box and then package the box. But one time, someone forgot to put the second sticker on and that brought in trouble. Afterwards, the TQC changed to what we have now, which is requiring one operator alone to put on both stickers plus packaging the box. It’s impossible for one person to finish those tasks while at the same time keeping up with the pace of production, so every time I have to assign other people to help her when no one is looking."

15 Move certain activities around so that there is sufficient space to use efficient motions to accomplish the task, e.g., putting the first sticker on the box. "The operator said to me that she had to put on two stickers, and one sticker she’s supposed to put on right before I put the cable into the boxes. So the way she usually does it is—once you bring a stack of new boxes, she would put the sticker on all of them, because the boxes happen to be on the right side facing up and she can just put them on right away, one after another. And after she does that, I can plug the cable into the boxes. But she usually does the sticker thing at my station, which you will see is at the beginning of the entire line. But this time the ITQC manager came over and said that she’s not supposed to do that. She can only stick the stickers at her own station, which is 7 or something. It’s a very crowded, very narrow area. You have a whole box over there—there’s no way you can do that. Obviously, she’s been using the other station for a long time, but the ITQC manager had never noticed before. She actually defended herself by saying that this has more space over here, and if she does it this way, it goes a lot faster. And after she puts the sticker on, I can put the cable in, and it all works out well. But the ITQC didn't see the proof of the benefits to justify changing the TQC, so she said she just has to follow the TQC—she has to do it that way."


P a g e | 57

16 Adjust motion-intensive tasks to the height of the operator. "Because I’m taller than most of the guys there, and I felt like I had to bend my back so much every time I have to put a screw on the USB drives. So I requested to have a little table placed on the counter where I work to make it higher."

17 Adjust the highly modular workspaces to improve efficiency of motions.

18 Adjust the connections between stations. "An engineer heard one of the operators talking about how to improve her station (scanning after engraving), and he offered a recommendation. Because there is a distance between those two stations, it takes time to deliver the stuff. And he suggested building a small ramp, so you just slide the cards down."

19 Fix the ports at the software upgrade station. "Many devices are broken: mouse, scanner, and ports included. No one seems to be fixing them. At the very least, we should consolidate the broken devices so that we have as many working combinations of parts as possible."

20 Be given ability to service screwing machine regularly so that it doesn't break down so much. "Because the screwing machine isn't serviced by the engineers as often as it should, it breaks down a lot. If we could grease it ourselves, then we could avoid a lot of downtime. It would also help to adjust the fixtures surrounding the machine so that it is easier to target the very small screw into the delicate plastic holes."

21 Change the positioning of the label printers so that hand motions are more fluid and natural. "Because of the way the scanner and label printers are currently positioned, I need to take the card, scan it into the system, put it down on the counter, pick up the box, put the freshly printed labels onto it (triggered by scanning), put down the box, pick up the card, put it into the box, and then pick up the box and pass it to the next operator. If we changed the location of the scanner and the label printers, I could do the entire operation without having to switch hands and materials so many times. It would save me a lot of time and effort."

W

F

B

A

WORKING PAP

Figure 1: Ex

Before / Out

After / Inside

PER

xample of B

side the Cur

e the Curtain

Before and A

rtain:

n:

After Curtai

in

E.. Bernstein, Th

he Transparenc

P

A

cy Paradox

P a g e | 58

Actual 60 240 80 200 280 280 80 289 271 240


P a g e | 59

Figure 2: Forms of Privacy

Figure 2: Forms of Privacy

Private

Transparent

Encryption Boundaries

Encrypted

Open

Visibility Boundaries

Public Hidden

APPEND

DIX A: PHOTOSS OF CURTAIN

No Curtain

N

n Currtain

P a g e | 60

APPEND

DIX A: PHOTOSS OF CURTAIN

No Curtain

N

n Currtain

P a g e | 61


P a g e | 62

REFERENCES:

AMA (American Management Association) 2007 ʺElectronic monitoring & surveillance survey.ʺ Adler, P. S. 1993 ʺThe ʹlearning bureaucracyʹ: New United Motors Manufacturing Incorporated.ʺ In B. M. Staw, and L. L. Cummings (eds.), Research in Organizational Behavior: 111‐194. Greenwich, CT: JAI Press. Adler, P. S., and B. Borys 1996 ʺTwo types of bureaucracy: enabling and coercive.ʺ Administrative Science Quarterly, 41: 61‐89. Adler, P. S., and K. B. Clark 1991 ʺBehind the learning curve: A sketch of the learning process.ʺ Management Science, 37: 267‐281. Altman, I. 1975 The Environment and Social Behavior: Privacy, Personal Space, Territory, Crowding. Monterey, Calif.: Brooks/Cole Publishing Co. Altman, I., and W. W. Haythorn 1967 ʺThe ecology of isolated groups.ʺ Behavioral Science, 12: 168–182. Anteby, M. 2008 Moral Gray Zones: Side Productions, Identity, and Regulation in an Aeronautic Plant. Princeton, NJ: Princeton University Press. Argote, L., et al. 2000 ʺKnowledge transfer in organizations: learning from the experience of others.ʺ Organizational Behavior and Human Decision Processes, 82: 1‐8. Argote, L., and R. Ophir 2002 ʺIntraorganizational learning.ʺ In J. A. C. Baum (ed.), Companion to Organizations: 181–207. New York: Wiley‐Blackwell. Argyris, C., R. Putnam, and D. M. Smith 1985 Action Science. San Francisco: Jossey‐Bass. Asch, S. E. 1951 ʺEffects of group pressure upon the modification and distortion of judgments.ʺ In H. Guetzkow (ed.), Groups, Leadership, and Men. New Brunswick, NJ: Rutgers University Press. Asch, S. E. 1956 ʺStudies of independence and conformity: a minority of one against a unanimous majority.ʺ Psychological Monographs, 70. Barker, J. R. 1993 ʺTightening the iron cage. Concertive control in self‐managing teams.ʺ Administrative Science Quarterly, 38: 408‐437. Barney, J. B., and S. Zhang 2009 ʺThe future of Chinese management research: a theory of Chinese management versus a Chinese theory of management.ʺ Management and Organization Review, 5: 15‐28. Barrett, F. J., and K. Peplowski 1998 ʺMinimal structures within a song: an analysis of ʺAll of Meʺ.ʺ Organization Science, 9: 558‐560. Bechky, B. A. 2003 ʺSharing meaning across occupational communities: the transformation of understanding on a production floor.ʺ Organization Science: 312‐330. Bennis, W. G., D. Goleman, and J. OʹToole 2008 Transparency: How Leaders Create a Culture of Candor. San Francisco, CA: Jossey‐Bass. Bentham, J. 1787 ʺPanopticon.ʺ


P a g e | 63

Bertrand, M., E. Duflo, and S. Mullainathan 2004 ʺHow much should we trust differences‐in‐differences estimates?ʺ The Quarterly Journal of Economics, 119: 249‐275. Best, J. 2005 The Limits of Transparency: Ambiguity and the History of International Finance. Ithaca: Cornell University Press. Blok, A., and C. Tilly 1975 The Mafia of a Sicilian Village, 1860‐1960: A Study of Violent Peasant Entrepreneurs: Harper & Row. Böhm, S. 2005ʺTransparent organization architectures of ideology and politics.ʺ Presentation given at the Warwick Space Symposium, University of Warwick, 23 May. Bond, C. F., and L. J. Titus 1983 ʺSocial facilitation: a meta‐analysis of 241 studies.ʺ Psychological Bulletin, 94: 265‐292. Brennan, A., J. S. Chugh, and T. Kline 2002 ʺTraditional versus open office design: a longitudinal field study.ʺ Environment and Behavior, 34: 279‐299. Brown, G., T. B. Lawrence, and S. L. Robinson 2005 ʺTerritoriality in organizations.ʺ Academy of Management Review, 30: 577‐594. Burawoy, M. 1979 Manufacturing Consent: Changes in the Labor Process Under Monopoly Capitalism. Chicago: University of Chicago Press. Burgoon, J., et al. 1989 ʺMaintaining and restoring privacy through communication in different types of relationships.ʺ Journal of Social and Personal Relationships, 6: 131‐158. Carey, A. 1967 ʺThe Hawthorne studies: a radical criticism.ʺ American Sociological Review, 32: 403‐416. Carlile, P. R. 2004 ʺTransferring, translating, and transforming: an integrative framework for managing knowledge across boundaries.ʺ Organization Science, 15: 555‐568. Chang, L. T. 2008 Factory Girls: From Village to City in a Changing China, 1st ed. New York: Spiegel & Grau. Chen, C. 2008 Some Assembly Required: Work, Community, and Politics in Chinaʹs Rural Enterprises. Cambridge, Mass.: Harvard University Asia Center, Distributed by Harvard University Press. Chwe, M. S.‐Y. 2001 Rational Ritual: Culture, Coordination, and Common Knowledge. Princeton: Princeton University Press. Cottrell, N. B. 1972 ʺSocial facilitation.ʺ In C. G. McClintock (ed.), Experimental Social Psychology. New York: Holt, Rinehart & Winston. Cross, R. L., and A. Parker 2004 The Hidden Power of Social Networks: Understanding How Work Really Gets Done in Organizations. Boston: Harvard Business School Press. Csikszentmihalyi, M., and K. Sawyer 1995 ʺCreative insight: the social dimension of a solitary moment.ʺ In R. J. Sternberg, and J. E. Davidson (eds.), The Nature of Insight: 329‐363. Cambridge, MA: MIT Press. Dalton, M. 1959 Men Who Manage: Fusions of Feeling and Theory in Administration. New York: Wiley.


P a g e | 64

Deming, W. E. 1986 Out of the Crisis. Cambridge, MA: Massachusetts Institute of Technology, Center for Advanced Engineering Study. Detert, J. R., and A. C. Edmondson 2011 ʺImplicit voice theories: taken‐for‐granted rules of self‐censorship at work.ʺ Academy of Management Journal, Forthcoming. Doz, Y. L. 1996 ʺThe evolution of cooperation in strategic alliances: initial conditions or learning processes?ʺ Strategic Management Journal, 17: 55‐83. Dunbar, R. I. M. 1992 ʺNeocortex size as a constraint on group size in primates.ʺ Journal of Human Evolution, 22: 469‐493. Edmondson, A. C. 1999 ʺPsychological safety and learning behavior in work teams.ʺ Administrative Science Quarterly, 44: 350‐383. Edmondson, A. C. 2002 ʺThe local and variegated nature of learning in organizations: a group‐level perspective.ʺ Organization Science, 13: 128‐146. Ellis, C., and A. Bochner 2000 ʺAutoethnography, personal narrative, reflexivity: researcher as subject.ʺ In N. K. Denzin, and Y. S. Lincoln (eds.), Handbook of Qualitative Research: 733‐768. Thousand Oaks, CA: Sage. Ellis, C., and M. Flaherty 1992 Investigating Subjectivity: Research on Lived Experience. Newbury Park, CA: Sage. Ely, R. J., and D. E. Meyerson 2000 ʺTheories of gender in organizations: a new approach to organizational analysis and change.ʺ Research in Organizational Behavior, 22: 103‐151. Emerson, R. M., R. I. Fretz, and L. L. Shaw 1995 Writing Ethnographic Fieldnotes. Chicago: University of Chicago Press. Eysenck, H. J. 1995 Genius: The Natural History of Creativity. Cambridge: Cambridge University Press. Foucault, M. 1977 Discipline and Punish: The Birth of the Prison, 1st American ed. New York: Pantheon Books. Gabbard, G. O. 2005 Psychodynamic Psychiatry in Clinical Practice, 4th ed. Washington, D.C.: American Psychiatric Pub. Gambetta, D. 1996 The Sicilian Mafia: the Business of Private Protection: Harvard University Press. Gibson, C., and F. Vermeulen 2003 ʺA healthy divide: subgroups as a stimulus for team learning behavior.ʺ Administrative Science Quarterly, 48: 202‐239. Gilovich, T., K. Savitsky, and V. H. Medvec 1998 ʺThe illusion of transparency: biased assessments of othersʹ ability to read oneʹs emotional states.ʺ Journal of Personality and Social Psychology, 75: 332‐346. Glaser, B. G., and A. L. Strauss 1967 The Discovery of Grounded Theory. New York: Aldine de Gruyter. Goffman, E. 1959 The Presentation of Self in Everyday Life. New York: Doubleday. Grief, M. 1991 The Visual Factory. Portland, OR: Productivity Press.


P a g e | 65

Hackman, J. R. 1976 ʺGroup influences on individuals.ʺ In M. Dunnette (ed.), Handbook of Industrial and Organizational Psychology: 1455‐1525. Chicago: Rand McNally. Hackman, J. R., and G. R. Oldham 1975 ʺDevelopment of the job diagnostic survey.ʺ Journal of Applied Psychology, 60: 159‐170. Hackman, J. R., and R. Wageman 1995 ʺTotal quality management: empirical, conceptual, and practical issues.ʺ Administrative Science Quarterly, 40: 309‐342. Hamper, B. 1986 Rivethead: Tales From the Assembly Line. New York, NY: Warner Books. Hansen, M. T. 1999 ʺThe search‐transfer problem: the role of weak ties in sharing knowledge across organization subunits.ʺ Administrative Science Quarterly, 44: 82‐111. Hargadon, A. 2003 How Breakthroughs Happen: The Surprising Truth About How Companies Innovate. Boston: Harvard Business School Press. Ingham, R. 1978 ʺPrivacy and psychology.ʺ In J. B. Young (ed.), Privacy: 35‐58. Chichester, UK: Wiley. Janis, I. L. 1983 ʺGroupthink.ʺ In J. R. Hackman, et al. (eds.), Perspectives on Behavior in Organizations. New York: McGraw‐Hill. Jett, Q. R., and J. M. George 2003 ʺWork interrupted: a closer look at the role of interruptions in organizational life.ʺ The Academy of Management Review, 28: 494‐507. Jones, S. R. G. 1992 ʺWas there a Hawthorne effect?ʺ The American Journal of Sociology, 98: 451‐468. Kanter, R. M., and R. Khurana 2009 ʺTypes and positions: the significance of Georg Simmelʹs structural theories for organizational behavior.ʺ In P. S. Adler (ed.), The Oxford Handbook of Sociology and Organization Studies: Classical Foundations. Oxford: Oxford University Press. Karabel, J. 2006 The Chosen: The Hidden History of Admission and Exclusion at Harvard, Yale, and Princeton: Mariner Books. Lambert, A. 2003 Russian Prison Tattoos: Codes of Authority, Domination, and Struggle. Atglen, PA: Schiffer Pub. Lamont, M., and V. Molnar 2002 ʺThe study of boundaries in the social sciences.ʺ Annual Review of Sociology, 28: 167‐195. Lazer, D., and A. Friedman 2007 ʺThe network structure of exploration and exploitation.ʺ Administrative Science Quarterly, 52: 667‐694. Levinson, A. A. 2009 ʺIndustrial justice: privacy protection for the employed.ʺ Cornell Journal of Law & Public Policy, 18: 609‐688. Lewis, C. A. 1979 ʺComment: healing in the urban environment.ʺ Journal of the Institute of American Planners, 45: 330‐338. March, J. G. 1991 ʺExploration and exploitation in organizational learning.ʺ Organization Science, 2: 71‐87. Marchetti, V., and J. Marks 1983 The CIA and the Cult of Intelligence: Laurel Leaf.


P a g e | 66

Marcus, G., and M. M. J. Fisher 1986 Anthropology as Cultural Critique: An Experimental Moment in the Human Sciences. Chicago: University of Chicago Press. Mas, A., and E. Moretti 2009 ʺPeers at work.ʺ American Economic Review, 99: 112‐145. Mayo, E. 1933 The Human Problems of an Industrial Organization. New York: McMillan. McGrath, J. E. 1976 ʺStress and behavior in organizations.ʺ In M. D. Dunnette (ed.), Handbook of Industrial and Organizational Psychology. Chicago: Rand McNally. McGrath, J. E. 1984 Interaction and Performance. New Jersey: Prentice Hall. Mill, J. S. 1859 On Liberty. London: Longman, Roberts, & Green Co. Morrison, E. W., and F. J. Milliken 2000 ʺOrganizational silence: A barrier to change and development in a pluralistic world.ʺ Academy of Management Review, 25: 706‐725. Murray, F., and S. OʹMahony 2007 ʺExploring the foundations of cumulative innovation: implications for organization science.ʺ Organization Science, 18: 1006‐1021. OʹMahony, S., and B. A. Bechky 2008 ʺBoundary organizations: enabling collaboration among unexpected allies.ʺ Administrative Science Quarterly, 53: 422‐459. Parent, W. A. 1983 ʺPrivacy, morality, and the law.ʺ Philosophy and Public Affairs, 12: 269‐288. Paulus, P. B., T. S. Larey, and A. H. Ortega 1995 ʺPerformance and perceptions of brainstormers in an organizational setting.ʺ Basic and Applied Social Psychology, 17: 249 ‐ 265. Perlow, L. A. 1999 ʺThe time famine: toward a sociology of work time.ʺ Administrative Science Quarterly, 44: 57‐81. Pierce, J. L., T. Kostova, and K. T. Dirks 2003 ʺThe state of psychological ownership: integrating and extending a century of research.ʺ Review of General Psychology, 7: 84‐107. Pisano, G. P. 1994 ʺKnowledge, integration, and the locus of learning: an empirical analysis of process development.ʺ Strategic Management Journal, 15: 85‐100. Roethlisberger, F. J., and W. J. Dickson 1939 Management and the Worker. Cambridge, Mass.: Harvard University Press. Rosenblatt, P. C., and L. G. Budd 1975 ʺTerritoriality and privacy in married and unmarried cohabitating couples.ʺ Journal of Social Psychology, 97: 67‐76. Roy, D. F. 1952 ʺQuota restriction and goldbricking in a machine shop.ʺ American Journal of Sociology, 57: 427. Roy, D. F. 1960 ʺBanana time: job satisfaction and informal interaction.ʺ Human Organization, 18: 156‐168. Sanday, P. R.


P a g e | 67

1979 ʺThe ethnographic paradigm(s).ʺ Administrative Science Quarterly, 24: 527‐538. Scheler, M. 1957 Uber Scham und Schamgeful (On Ethics and Theory of Knowledge). Bern: Francke. Schoeman, F. D. 1984 ʺPrivacy and intimate information.ʺ In F. D. Schoeman (ed.), Philosophical Dimensions of Privacy: An Anthology: 403‐417. Cambridge: Cambridge University Press. Schofield, C. B. P., and A. N. Joinson 2008 ʺPrivacy, trust, and disclosure online.ʺ In A. Barak (ed.), Psychological Aspects of Cyberspace: Theory, Research, Applications: 13‐31. Cambridge, UK: Cambridge University Press. Schwartz, B. 1968 ʺThe social psychology of privacy.ʺ The American Journal of Sociology, 73: 741‐752. Sewell, G. 1998 ʺThe discipline of teams: the control of team‐based industrial work through electronic and peer surveillance.ʺ Administrative Science Quarterly, 43: 397‐428. Simmel, G. 1950 The Sociology of Georg Simmel. Glencoe, Ill.: Free Press. Simmel, G. 1964 ʺThe secret and the secret society.ʺ In K. Wolff (ed.), The Sociology of Georg Simmel: 334. New York: Free Press. Simmel, G., M. Landmann, and M. Susman 1957 Brücke und Tür; Essays des Philosophen zur Geschichte, Religion, Kunst und Gesellschaft. Stuttgart: K. F. Koehler. Simonton, D. K. 2003 ʺScientific creativity as constrained stochastic behavior: the integration of product, person, and process perspectives.ʺ Psychological Bulletin, 129: 475‐494. Sitkin, S. B., K. M. Sutcliffe, and R. G. Schroeder 1994 ʺDistinguishing control from learning in Total Quality Management: a contingency perspective.ʺ The Academy of Management Review, 19: 537‐564. Snook, S. A. 2000 Friendly Fire: The Accidental Shootdown of U.S. Black Hawks over Northern Iraq. Princeton, N.J.: Princeton University Press. Solove, D. J. 2008 Understanding Privacy. Cambridge, Mass.: Harvard University Press. Sorenson, O., J. W. Rivkin, and L. Fleming 2006 ʺComplexity, networks, and knowledge flow.ʺ Research Policy, 35: 994‐1017. Spear, S., and H. K. Bowen 1999 ʺDecoding the DNA of the Toyota Production System.ʺ Harvard Business Review, 77: 96‐106. Spear, S. J. 2009 Chasing the Rabbit: How Market Leaders Outdistance the Competition and How Great Companies Can Catch Up and Win. New York: McGraw‐Hill. Staats, B. R., and F. Gino 2010 ʺThe task and temporal microstructure of productivity: evidence from Japanese financial services.ʺ Boston: HBS Working Paper. Star, S. L., and J. R. Griesemer 1989 ʺInstitutional ecology, ʹtranslationsʹ and boundary objects: amateurs and professionals in Berkeleyʹs Museum of Vertebrate Zoology, 1907‐39.ʺ Social Studies of Science, 19: 387‐420.


P a g e | 68

Sutton, R. I., and T. A. Kelley 1997 ʺCreativity doesnʹt require isolation: why product designers bring visitors ʺbackstageʺ.ʺ California Management Review, 40: 75‐91. Tapscott, D., and D. Ticoll 2003 The Naked Corporation: How the Age of Transparency Will Revolutionize Business. New York: Free Press. Taylor, F. W. 1911 The Principles of Scientific Management. University of Wisconsin Madison: Harper & Brothers. Thomke, S., E. von Hippel, and R. Franke 1998 ʺModes of experimentation: an innovation process‐‐and competitive‐‐variable.ʺ Research Policy, 27: 315‐332. Thomke, S. H. 1998 ʺManaging experimentation in the design of new products.ʺ Management Science, 44: 743‐762. Tilly, C. 2003 ʺPolitical identities in changing polities.ʺ Social Research, 70: 605‐620. Tsui, A. S. 2007 ʺTaking stock and looking ahead: MOR and Chinese management research.ʺ Management and Organization Review, 3: 327‐334. Tucker, A. L., A. C. Edmondson, and S. J. Spear 2002 ʺWhen problem solving prevents organizational learning.ʺ Journal of Organizational Change Management, 15: 122‐137. Van Maanen, J. 1988 Tales of the Field: On Writing Ethnography. Chicago: University of Chicago Press. Waterworth, J., and E. Waterworth 2001 ʺAffective creative spaces: the interactive tent and the illusion of being.ʺ In K. Helander, and T. Helander (eds.), Proceedings of the International Conference on Affective Human Factors Design. London: Asean Academic Press. Westin, A. F. 1967 Privacy and Freedom. New York: Atheneum. Zajonc, R. B. 1965 ʺSocial facilitation.ʺ Science, 149: 269‐274. Zalesny, M. D., and R. V. Farace 1987 ʺTraditional versus open offices: a comparison of sociotechnical, social relations, and symbolic meaning perspectives.ʺ The Academy of Management Journal, 30: 240‐259.

The Transparency Paradox: A Role for Privacy in ... · PDF fileWORKING PAPER E. Bernstein, The...

Documents

Transcript of The Transparency Paradox: A Role for Privacy in ... · PDF fileWORKING PAPER E. Bernstein, The...