DOI:10.24900/ijss/01034660.2017.1201 Advancing Safety by ...€¦ · Somayeh Asadi1, Ebrahim...
Transcript of DOI:10.24900/ijss/01034660.2017.1201 Advancing Safety by ...€¦ · Somayeh Asadi1, Ebrahim...
International Journal of Safety Science
Vol. 01, No. 03, (2017), pp. 46-60
DOI:10.24900/ijss/01034660.2017.1201
46
Advancing Safety by In-Depth Assessment of Workers Attention
and Perception
Somayeh Asadi1, Ebrahim Karan
2, Atefeh Mohammadpour
3
1 Department of Architectural Engineering, Pennsylvania State University, 104
Engineering Unit A, University Park, PA 16802, USA
2Department of Applied Engineering, Safety, and Technology, Millersville
University, Osburn Hall, PO Box 1002, 40 East Frederick Street, Millersville, PA
17551, USA
3Department of Manufacturing & Construction Engineering Technology, Indiana
University – Purdue University Fort Wayne, 2101 East Coliseum Boulevard, Fort
Wayne, IN 46805, USA
Abstract
Considering the complexity and sometimes unpredictability of human behavior, and its
role on workplace injuries, it is surprising that psychological research has contributed
relatively little in studying workplace safety compared to technology development. The
focus of this paper is thus shifted to psychological investigation of human factors to
understand how a hazard is perceived, how a seen hazard is recognized, and how a
decision must be made to avoid the hazard. The study is conducted through a three-step
experiment including a questionnaire, Balloon-Analogue-Risk-Task (BART) test and
safety photos and video clips with an eye tracking system to measure and analyze the
relationship between the worker’s personality and perceptions of risks and hazards. A
very close relationship was found between the length of time a participant spent looking
at a hazard and the number of hazards identified correctly. Participants with a higher
safety score were able to recognize more safety hazards than those with lower safety
score. Moreover, it was found that risk-takers were underestimating a hazard, whereas
more cautious and less risky participants were overestimating a hazard.
Keywords: Construction safety, attention, perception, risk-taking tendency, eye-
tracking
1. Introduction
In the past decade (2005-2014) more than 50,000 U.S. workers have died at work, with
more than 9,600 fatalities in construction it continues to be the most dangerous industry
[1]. That equates to approximately four construction worker deaths every working day.
Despite this high fatality rate, considerable progress has been made in improving
workplace safety and thus injury fatality rates have decreased approximately 25% from
2005 to 2014. The reasons are varied, but some of the more common ones include
technological advances in personal protective equipment (PPE), safety incentive and
disincentive policies and practices, effective safety training programs, involvement of
workers in safety by including them in decision-making, helping them become safety
leaders and trainers, and encouraging their ideas for improvement [2].
Safety research has provided valuable strategies to develop construction worker
competency in hazard recognition and communication [3]. Studies have shown that safety
training can improve hazard identification skills for managers and workers and help
preventing workplace injuries and accidents [4]. There is a significant negative
relationship between the level of safety-focused worker cognitive engagement with
accident rates, therefore safety management systems and worker engagement levels can
International Journal of Safety Science
Vol. 01, No. 03, (2017), pp. 46-60
DOI:10.24900/ijss/01024660.2017.1201
47
be used individually to predict safety performance outcomes such as accident rates [5]. To
find the reason of unsafe behaviors, individual and environmental conditions that
significantly affect safety awareness and safety behavior must be identified [6]. This
would lead to develop foremost indicators that can be used by workers and contractors to
improve identification and awareness of the most common risk factors in workplace [7].
In recent years, several studies have focused on the application of PPE and engineering
control technologies to improve safety in construction projects. However, the statistics
show that these advances are not fully preventing accidents from happening in
construction projects, despite the major declines in work injuries of other industry sectors
[8]. This can be largely attributed to the uniqueness of construction projects that expose
workers to “non-design” emergencies (i.e. emergencies that could not have been foreseen
by the designers).
A rich volume of existing publications have solely focused on technology development
and neglected the human side of the equation. Examples of technology-oriented solutions
include efforts such as sharing safety knowledge using a mobile device [9], real-time data
collection and visualization technology for construction safety [10] and [11], safety
assessment of construction site using geographic information systems (GIS) [12], vision-
based recognition of construction equipment and hazard detection [13], game technology
combined with visualizing safety assessment [14], autonomous jobsite safety proximity
[15], location-aware sensing technology to determine a worker’s safety situation
regarding proximity to dangers [16], use of digital models to enhance construction safety
[17], enhancing hazard recognition with augmented virtually [18], and the applications of
building information modeling (BIM) to investigate potential fall hazards [19]. Among
these studies few of them have investigated the application of eye-tracking technology in
construction safety [20]. Dzeng et al. [21] used an eye-tracking system to compare how
the experienced and novice workers pay attention to the obvious and unobvious hazards.
Their results showed that experienced workers are able to detect hazards faster than
novice workers. It was also found that there is a significant difference between general
work experience and safety-specific experience, and general work experience may not
necessarily improve workers’ accuracy in identifying hazards. In another study conducted
by Hasanzadeh et al. [22], different scenarios within a real-world construction site were
developed to quantify workers’ situation awareness using a mobile eye-tracker. They
found that some factors such as workload, the state of the area of interest, and the level of
experience of workers have a significant impact on the allocation of situation awareness
and visual attention.
The complexity and sometimes unpredictability of human behavior, and its role on
workplace injuries, direct us to shift our emphasis from technology development
approaches towards psychological investigation of human factors. However, few studies
have applied psychology in the construction domain to understand the role of human
factors in safety of workers. The findings point out that risk perception are often linked to
socio-demographic factors such as personality facets [23], gender, education, income, and
social context [24], training, ethnicity and socio-economic status [25]. Tixier et al. [26],
hypothesized that there are significant differences in construction safety risk perception
among people having different emotional states and found that workers in positive and
neutral emotional states may perform risk perceptions that are notably lower than people
with negative emotions. Hallowell [27] quantified the current level of safety risk and the
risk tolerance of workers and managers and compared the risk perceptions and tolerance
of workers with managers. He found a statistically significant difference in the risk
tolerance between workers and managers showed that the level of current perceived risk is
approximately five times higher than the tolerable risk value.
International Journal of Safety Science
Vol. 01, No. 03, (2017), pp. 46-60
DOI:10.24900/ijss/01034660.2017.1201
48
The theoretical basis for these methods rests on the assumption that human behaviors
depend largely on environmental factors rather than on personal characteristics. As
opposed to current practices that largely focus on technology development, this study
aims to prevent workplace injuries and fatalities by exploring the human side of the
equation and providing an in-depth understanding of human behavior and actions in terms
of safety. After a brief overview of the human factor aspects of accidents, the possibility
of investigating contributing human factors for failures in the perception of a hazard and
the decision to avoid are discussed. This paper will suggest an innovative use of
psychological measurements, an experimental procedure and quantitative analysis of the
results to understand how people perceive a hazard and then evaluate a perceived hazard.
2. Safety and Human Factor
Ramsay's accident sequence model illustrated in Figure 1 demonstrates how a hazard is
perceived (through sensory inputs), how a seen hazard is recognized (or understood to be
a hazard), and how a decision must be made to avoid the hazard based on risk tendency
and personality [28]. Safety decisions or hazard identification by occupants are made in
very short periods of time and it makes the role of attention very important in the present
study. The perceptual process explains how a person takes sensory information
(information collected from person’s senses such as sight, hearing, and smell), or stimuli,
from the environment. Attention explains how a person selects information for more
extensive processing [29] and depends on two important theoretical concepts, including
mental load and situational awareness. The mental load is defined as the number of things
that are objectively relevant to the task while the concept of attention is applied in a more
subjective sense for situational awareness in order to predict which aspects of the current
situation a person will become aware of [30]. Therefore, attentional process plays a much
greater role when the decision must be made in a short period of time (e.g. in risky and
hazardous situations), usually less than 200 ms.
Figure 1. Accident sequence model representing various stages in the
occurrence or avoidance of accident (adopted from Ramsey et al, 1986)
In a recent study, Chi et al. [31] analyzed 9,358 accidents that occurred in the U.S.
construction industry and found a significant correlation between workers’ behavior and
International Journal of Safety Science
Vol. 01, No. 03, (2017), pp. 46-60
DOI:10.24900/ijss/01024660.2017.1201
49
their impact on injury severity. Approximately 80 to 90 percent of these accidents were
strongly associated with workers' unsafe behavior and acts. Behavior models and human
factors approach consider human errors as being the main cause of accidents [32]. Quinn
[33] studied the demographic information of field and office personnel who most
frequently are involved in workplace accidents and found that 56.3 percent of accidents
were caused by a lack of attention or awareness. Hutchings et al. [34] visited twenty
jobsites and interviewed eighteen general contractors and subcontractors to determine
both causes of and solutions to accidents among on-site construction workers. The
interviewees agreed that lack of training, lack of attention to safety, and negligence are
the most important factors contributing to job-site accidents among workers. In another
study, Gillen et al. [35] evaluated injured construction workers and found a positive
significant correlation between injury severity sustained by the workers and their
perceptions of workplace safety. This finding was confirmed by the result of another
study [36], in which it was found that shared perceptions of safety conduct at work were
negatively and significantly associated with injury rates.
Humans use information from perceptual systems such as vision, audition, touch, taste,
and smell to make decisions. Although inputs from all the perceptual systems are
important, visual perception strongly influences the thinking and acting of human beings.
Eye tracking is the corresponding computer-aided research method to record and analyze
this behavior. Eye-tracking is the process of measuring the point of gaze (the spot a
participant is looking at), and an eye tracker is the portable hardware device that performs
this measurement by measuring eye movement [37]. It is a method to assess the eye
movement (as a dependent variable) in the anticipation of stimuli (as an independent
variable) [38]. Even though the application of eye-tracking technology has been
recognized in a wide variety of domains and its use has become prevalent in a number of
industries, considerably less attention has been paid to its potential to improve
construction safety. Workers act based on their perceptions of risk at the jobsite, so we
can conclude that safety improvement in the construction industry is the most appreciated
beneficial effect of enhancing worker’s risk perception. Therefore, systematic
understanding of unsafe behavior has great potential to contribute to the reduction and
prevention of injuries and fatalities in construction projects.
3. Research Objective
The objective of this study is to find a proper answer for following questions regarding
the source of human error that have intrigued researchers for decades: Do the construction
workers pay enough attention to what is going on around them (error in attention)? Can
construction workers understand or identify the risk if they see a hazard (error in
perception)? What are the causes of systematic deviations from the sequence of
operations restricted or required in safety standards (error in memory)? The significance
of this study is to apply psychology to answer these questions; accurately identify risk-
takers and individual-level interventions; and ultimately advance safety by reducing
worker exposures to workplace hazards.
To achieve this objective, the present study analyzes and understands the relationship
between participant’s attention, visual perception, and the way she or he interprets the
registered information. In addition, authors examine whether individual differences in on-
the-job risk taking tendencies by participants can be predicted by psychological variables.
International Journal of Safety Science
Vol. 01, No. 03, (2017), pp. 46-60
DOI:10.24900/ijss/01034660.2017.1201
50
The risk taking behavior of participants is measured to identify workers who are more
likely to engage in risky decision making on the construction site. To achieve this
objective, eye-tracking technology is used to analyze the relationship between the
worker’s personality (i.e. attitude towards risks/safety) and perceptions of risks and
hazards. The need for clarity on these issues thus forms a gap in the current body of
knowledge on the area of human behavior in construction safety to examine whether
individual differences in on-the-job risk taking by construction workers can be predicted
by psychological variables. This study can also be used to identify workers who are more
likely to engage in risky decision making on the construction site. If one can identify
those individuals who are more likely to engage in risky behaviors, then they can be
targeted for specialized training.
4. Research Methodology
At least two challenges remain in studying the relationship between workers’ behavior
and assessment of safety performance at construction sites. The first challenge concerns
the amount of time and effort needed to assess risk or safety perceptions among
construction workers. The second challenge is to understand how personality and
perception can influence risk taking. To address these challenges, a three-step experiment
was designed to measure and analyze the relationship between the worker’s personality
and perceptions of risks and hazards and assess the safety performance at construction
sites. Figure 2 shows the overall architecture of the study. The focus of the present study
is to understand the sources of human error (i.e., attention, perception, and/or memory)
through different experiments including a questionnaire, Balloon-Analogue-Risk-Task
(BART) test as well as safety photos and video clips with the eye tracking system.
Figure 2. Overall architecture of the study
Each step of the present study is designed to simulate construction worker’s general
awareness on recognizing hazards on a construction site. At the first step, a questionnaire
International Journal of Safety Science
Vol. 01, No. 03, (2017), pp. 46-60
DOI:10.24900/ijss/01024660.2017.1201
51
was designed and conducted with the objective of assessing the subject's working
experience and knowledge on safety information which included age, gender, work
experience, and safety training (e.g. 10-hor or 30-hour Occupational Safety and Health
Administration (OSHA) training). Twenty multiple choice questions, obtained from
OSHA training program, were used to evaluate the participants’ prior safety knowledge.
The score is represented on a scale from zero percent to a maximum of 100 percent and
used as the participant’s safety score. The purpose of designing a questionnaire was to
understand whether participants with higher safety score perform better in identifying
safety hazards than participants with lower score or inexperienced participants.
In the second step, the OSHA safety training photos and video clips were displayed on
the monitor and participants were asked to take a look at the photos for a certain period of
time and identify the actual or potential hazards they see, while their interaction with the
photos and video clips is recorded using eye tracking. These photos and video clips
ranged from near misses to unsafe behaviors/conditions with high potential for accidents.
Video clips of potential hazards were recorded from one of the on-going construction
projects at the Pennsylvania State University campus. The eye movement data and
participants’ responses were used to quantify visual attention and measure participants’
perception of hazard and risk, respectively. Therefore, it was possible to understand
“where participants look” and “what they look for”.
Along with identifying potential hazards, the participants were asked to rate the level of
danger (or level of hazard) presents in the safety photos and video clips to determine the
severity score. The presented study used Safety Assessment Code (SAC) developed by
the National Center for Patient Safety (NCPS) for doing comparative analysis. Three
faculty members at Millersville University - Occupational Safety & Environmental Health
(OSEH) program were interviewed to verify the severity level of hazards and those were
used as the baseline for hypothesis testing. Four different levels of severity including
Catastrophic, Major, Moderate, and Minor were defined. The hypothesis is that the
participants who are taking greater risks (based on the results of the BART experiment)
would underestimate the level of danger while those who are less risky would
overestimate the level of danger present in the safety photos. The following equation is
used to calculate the severity score.
Severity Score =
where, Catastrophic = +4, Major = +3, Moderate = +2, and Minor = +1.
For example, if the correct answer is Major and the subject picked Minor, his/her score
will be +3 – 1 = +2, which means that the hazard is underestimated and if the subject
picked Catastrophic, his/her score will be +3 – 4 = -1, which means that the hazard is
overestimated.
In the third step, the BART test was carried out to identify risk-takers [39]. BART is a
computerized measure of risk taking behavior that models real-world risk behavior
through the conceptual frame of balancing the potential for reward versus loss. In the test,
the participant is presented with a balloon and offered the chance to earn money by
pumping the balloon up by clicking a button. Each click causes the balloon to
incrementally inflate and money to be added to a counter up until some threshold, at
which point the balloon is over inflated and explodes. Thus, each pump confers greater
risk but also greater potential reward.
International Journal of Safety Science
Vol. 01, No. 03, (2017), pp. 46-60
DOI:10.24900/ijss/01034660.2017.1201
52
5. Experimental Design
In this study a Tobii X2-30 compact eye-tracker was used to record eye movement data.
The technical specification of the eye-tracker includes: gaze accuracy of 0.4 to 0.5 degree
range, precision of 0.32 to 0.45 degrees, data rate 30 Hz, freedom of head movement 50
(width) x 36 (height) x 90 (depth) cm, calibration 9 points (~ 15 seconds to complete),
monocular and binocular tracking capability. To perform a BART test, participants were
asked to pump up a computerized balloon. The maximum number of pumps that can be
made to a balloon is 128, and for each pump, the participant is awarded a small amount of
money that is placed in a temporary bank (¢5 per pump). The participant can pump the
balloon 1 to 128 times, and must decide when to stop pumping. If they stop pumping
before the balloon bursts, the money in the temporary bank is placed in a permanent bank.
However, if the balloon bursts before they stop pumping, the money in the temporary
bank disappears. Thus, people who are willing to pump up the balloon more are
essentially taking greater risks, whereas people who quit early can be viewed as being
more cautious and less risky. Lejuez et al. (2002) has shown that one’s propensity to
pump up the balloon is positively correlated with real-world risk taking behaviors.
In this study a within-subjects experiment is implemented in which every single
participant was subjected to every single safety photo and video clip. The results were
analyzed using 122 data samples collected from 32 participants (24 male and 8 female).
Since the focus of this study is on construction safety, all participants for this experiment
were drawn from construction workers, people who had worked in construction jobsites
or students who were studying construction management or construction engineering.
Table 1 summarizes the descriptive statistics of the subject pool and the safety
questionnaire results. Eleven participants (34%) reported that they had formal safety
training and eight other participants (25%) reported no previous work experience. There is
no doubt that prior safety knowledge plays a vital role in the level of hazard identification
(safety photo score), however, this study examines whether construction workers with
similar prior safety knowledge but higher attentional control are able to recognize and
perceive more safety hazards than those with lower attentional control. The results of the
statistical analysis in the next section provide a quantitative way to answer this question.
Table 1. Summary of the Experimental Designs
N Mean/Percentage Std. Deviation
Age - 29.1 11.0
Gender - male 24 75% N/A
Gender - female 32 25% N/A
Work experience - 5.6 yr. 8.1 yr.
Safety score - 54% 3%
In total, ten safety photos and four safety video clips were created and showed to the
participants. All photos and video clips were taken into consideration for the correlation
analysis. Each participant spent 10 and 17 seconds seeing the photos and video clips
respectively, before the evaluation. To further analyze the eye movement (e.g. fixation)
data, static and dynamic areas of interest (AOI) were defined for each photo and video
clip, respectively. AOIs are relevant areas on the stimulus that are usually defined before
starting the experiment and are closely connected to the research question and the
corresponding research design. To ensure consistency and proper balance between
International Journal of Safety Science
Vol. 01, No. 03, (2017), pp. 46-60
DOI:10.24900/ijss/01024660.2017.1201
53
selectivity and sensitivity in the results, the AOI was defined as 3% of the total area. The
selected size of the AOI was defined based on the largest hazardous situation in each
photo and video. Figure 3 shows the heat map of two of the studied safety photos. Heat
map (or hot spots) is a widely adopted visualization technique for eye tracking studies.
Heat map shows the areas with a large amount of interest and the aggregate eye fixations
of those areas which are shown in red. As it can be seen in these figures, the Time Spent
in AOI, which is defined based on total duration of all respondent's fixations and
measured in second, is significantly higher in the red area in comparison with green areas.
For example, in Figure 3 the Time Spent in AOI is 2s and 1.8s respectively, while in other
areas they range between 0.1-0.8s. The same trend is seen for fixation and time to first
fixation (TTFF). TTFF is the time stamp of the first fixation inside the AOI.
Figure 3. Heat map and the associated AOIs for the photo experiment
6. Results and Discussions
One assumption is if the participants don’t see the stimuli, they can’t identify the
hazard (error in attention). To test this hypothesis, an independent-samples t-test is used to
understand whether the hazard identification score (HIS) differed based on the fixation
time (time spent looking at a hazard). Thus, the dependent variable would be the “HIS”
and the independent variable would be fixation time, which has two groups: unaware,
participants who don’t see a hazard (their fixation time is less than a specific duration)
and attentive, participants who see a hazard (fixation time is greater than a specific
duration). The HIS for each participant is defined as:
Where, t is the fixation time (measured in seconds) less than a specific duration (d0) for
unaware group and greater than a specific duration (d0) for attentive group, IH is the
number of hazards that are identified correctly, and UH is the number of unidentified
hazards (hazards that are not identified). For example, for unaware
group shows that 3 out of 5 hazards are identified correctly when the time spent looking at
a hazard are less than 2 seconds. The HIS between unaware and attentive groups are
compared for different time periods.
Table 2 summarizes the results of two-sample t-test associated with a 95% significant
level for the safety photo experiment. The significance (2-tailed) value for fixation time
periods equal or less than 3 seconds is less than 0.05 (p>0.05). There is a statistically
significant difference between the mean number of hazards identified in the safety photos
for the attentive and unaware groups. Since the average for the attentive group was greater
AOI
AOI
International Journal of Safety Science
Vol. 01, No. 03, (2017), pp. 46-60
DOI:10.24900/ijss/01034660.2017.1201
54
than the average for the unaware group, we can conclude that participants who spent more
than 3 seconds at a hazard were able to identify significantly more hazards than
participants who spent less than 3 seconds at a hazard.
Table 2: quantitative analysis of t-test for different fixation time periods (photo experiment)
Fixation time Group Mean score Std. deviation F Sig. (2-
tailed)
0.5 sec Attentive 0.50 0.20
5.22 <0.001 Unaware 0.18 0.29
1 sec Attentive 0.53 0.24
1.412 <0.001 Unaware 0.25 0.29
1.5 sec Attentive 0.53 0.30
0.916 <0.001 Unaware 0.25 0.23
2.0 sec Attentive 0.59 0.27
0.042 <0.001 Unaware 0.26 0.23
2.5 sec Attentive 0.57 0.32
1.379 0.001 Unaware 0.30 0.23
3.0 sec Attentive 0.53 0.33
1.525 0.020 Unaware 0.33 0.23
3.5 sec Attentive 0.40 0.31
7.155 0.495 Unaware 0.35 0.19
A correlation analysis is performed on the data to measure the strength and direction of
relationships between the fixation duration and the HIS. As shown in Table 3, the Pearson
correlation between the fixation durations and the HIS is remarkably high (correlation
coefficient = 0.806) and statistically significant at the 95% confidence level. Therefore
there is a very close relationship between the length of time a participant spent looking at
a hazard and the number of hazard identified correctly.
Table 3. Comparison of fixation time periods with HIS (photo experiment)
Pearson correlation
Fixation time hazard identification score
Fixation time 1 0.927 (Sig. < 0.001)
hazard identification score 0.806 (Sig. < 0.001) 1
Figure 4 shows the relationship between the fixation duration and HIS for the safety
photo experiment. As can be seen, lower scores (when hazards were not identified) are
associated with shorter fixation duration. For example, when fixation duration is less than
0.5 seconds, 100-23=77% of the time the hazards were not identified correctly and when
fixation duration is less than 2 seconds, 100-31=69% of the time the hazards were not
identified correctly.
International Journal of Safety Science
Vol. 01, No. 03, (2017), pp. 46-60
DOI:10.24900/ijss/01024660.2017.1201
55
Figure 4: Relationships between fixation duration and hazard identification
score
Similar analysis was performed for the video experiment, where fifteen participants
were asked to watch four different safety video clips. As shown in Table 4, none of the
differences between the attentive and unaware groups were significant and no conclusion
can be drawn between the HIS and the fixation time for the video experiment. This is
most likely a result of the small sample size in the experiment, because only eighteen data
sets could be used from fifteen participants in the safety video experiment, while a total of
ninety two data sets were used from thirty two participants in the safety photo experiment.
Table 4: quantitative analysis of t-test for different fixation time periods (video experiment)
Fixation time Group Mean score Std. deviation F Sig. (2-
tailed)
0.5 sec Attentive 0.27 0.44
1.427 0.248 Unaware 0.14 0.38
1, 1.5 sec Attentive 0.23 0.41
0.189 0.669 Unaware 0.33 0.43
2.0 sec Attentive 0.15 0.34
2.666 0.120 Unaware 0.38 0.46
2.5 sec Attentive 0.33 0.52
0.350 0.563 Unaware 0.35 0.45
3.0 sec Attentive 0.40 0.55 2.304 0.150
After examining the relationship between visual attention and hazard identification, the
eye-movement data classified as “seen” were further analyzed to understand whether the
participants with knowledge and experience in the field of safety perform better in
identifying safety hazards than participants with lower safety score and experience. The
prior knowledge was assessed through a series of multiple-choice questions and is
reported as the participant’s safety score. In order to perform the statistical analysis, the
participants were grouped based on their safety score. The average safety score for the
participants was 58%. The participants with high or above average safety score are listed
as knowledgeable and participants who scored less than 58% in the safety questionnaire
are listed as less knowledgeable.
International Journal of Safety Science
Vol. 01, No. 03, (2017), pp. 46-60
DOI:10.24900/ijss/01034660.2017.1201
56
Table 5 shows the results of two-sample t-test associated with a 95% significant level
for the participants who spent more than 3 seconds at a hazard (note: fixation time of 3
seconds is classified as “seen”). Although authors fail to reject the null hypothesis (p-
value = 0.086 > 0.05) that when a hazard is seen, participants with higher safety score (or
knowledgeable participants) have significantly higher safety performance in identifying
safety hazards than non-experienced participants, but it was noticed that they are able to
recognize more safety hazards than those with a lower safety score (or less knowledgeable
participants). When the average time in seconds spent looking a hazard is more than 3
seconds, participants with higher safety score could identify 64% of hazards; however,
less knowledgeable participants could only identify 33% of seen hazards.
Table 5: quantitative analysis of t-test for “seen hazards” (photo experiment)
Fixation
time
Group Mean
safety
score
Mean hazard
identification
score
F Sig. (2-
tailed)
3 sec
Knowledgeable 46% 0.60
1.217 0.317 Less knowledgeable 66% 0.41
Even if a safety hazard is seen and recognized, construction workers with less
experience or high risk-taking tendencies may underestimate or neglect to avoid a hazard.
To find the causes of deviations from the sequence of operations required in safety
standards, participants’ experience and BART scores are compared with the severity
scores related to the identified hazards. A correlation analysis is performed to measure the
extent of interdependence between the risk-taking tendencies (based on the results of the
BART experiment) or experience and the severity scores. As shown in Table 6 and Figure
5, there was a strong, negative correlation between “years of experience” and “severity”,
which was statistically significant (r= -0.545, p = 0.002). Thus, the participants with
greater work experience would overestimate the level of danger while those who had no
or little work experience would underestimate the level of danger present in the safety
photos.
Table 6. Comparison of participants’ work experience and severity scores
Pearson correlation
Fixation time hazard identification score
Severity score 1 -0.545 (p = 0.002)
Work experience -0.545 (p = 0.002) 1
International Journal of Safety Science
Vol. 01, No. 03, (2017), pp. 46-60
DOI:10.24900/ijss/01024660.2017.1201
57
Figure 5: Relationships between participants’ work experience and severity
score
A correlation analysis between the BART and severity scores is performed to test the
hypothesis that participants with higher risk-taking tendencies would underestimate the
level of danger while those who are risk-avoider would overestimate the level of danger
present in the safety photos (see Table 7 and Figure 6). On average, participants pumped
up the computerized balloon 50 times and earned $37.6. Although there is inconclusive
evidence about the significance of the association between the risk taking tendencies and
severity scores (p-value = 0.059 > 0.05), we notice a moderate linear relationship between
the variables. Risk takers (participants who pumped up the balloon more) were
underestimating a hazard (higher severity score), whereas participants who quit early
(more cautious and less risky) were overestimating a hazard (lower severity score).
Table 7. Comparison of participants’ BART score and severity scores
Pearson correlation
Fixation time hazard identification score
Severity score 1 0.400 (p = 0.059)
BART score (average pumps) 0.400 (p = 0.059) 1
International Journal of Safety Science
Vol. 01, No. 03, (2017), pp. 46-60
DOI:10.24900/ijss/01034660.2017.1201
58
Figure 6: Relationships between the average number of pumps and severity
score
7. Conclusions
Many construction workers are injured because of safety hazards on the job site. This
study shows a 95% statistically significant difference between the mean number of
hazards identified in the safety photos for the attentive and unaware groups. Participants
recognized hazards when they spent more than 3 seconds on a hazard in comparison with
participants who spent less than 3 seconds. As a result, there was a very close relationship
between the length of time a participant spent looking at a hazard and the number of
hazard identified correctly.
This study failed to prove that knowledgeable participants have higher safety
performance in identifying safety hazards than non-experienced participants. But it is
important to note that when participants spent more than 3 seconds looking at hazards,
participants with safety knowledge and experience were able to recognize 66% of hazards
while less experienced participants could identify 46% of the hazards. In correlation
analysis between the BART and severity scores there was a moderate linear relationship
between risk-taker and risk-avoider behaviors. There was a strong, negative correlation
between “years of experience” and “severity”, which was statistically significant. It was
found that the participants with greater work experience would overestimate the level of
danger while those who had no or little work experience would overestimate the level of danger present in the safety photos.
The results of this study can be used to develop an assessment tool to gauge how well
construction workers understand safety hazards in a workplace and determine the sources
of human error. This information will be combined with the workplace hazards identified
through a job hazard analysis to predict how (e.g. who might be involved and what is the
source(s) of his or her error), when, and where an accident occurs. A possible subject to
be left for future research is to couple the eye –tracking technology with an immersive
virtual reality environment in order to expose subjects to more realistic construction
environments in order to actively interact with it.
References
1. BLS. Fatal occupational injuries. 2015; Available from:
http://www.bls.gov/iif/oshwc/cfoi/all_worker.pdf.
2. OSHA, Construction Industry Digest. 2015, U.S. Department of Labor: Occupational Safety and
Health Administration. p. 76.
3. Albert, A., M.R. Hallowell, and B.M. Kleiner, Experimental field testing of a real-time construction
hazard identification and transmission technique. Construction Management and Economics, 2014.
32(10): p. 1000-1016.
4. Bahn, S., Workplace hazard identification and management: The case of an underground mining
operation. Safety science, 2013. 57: p. 129-137.
5. Wachter, J.K. and P.L. Yorio, A system of safety management practices and worker engagement for
reducing and preventing accidents: An empirical and theoretical investigation. Accident Analysis
& Prevention, 2014. 68: p. 117-130.
6. Jiang, Z., D. Fang, and M. Zhang, Understanding the causation of construction workers’ unsafe
behaviors based on system dynamics modeling. Journal of Management in Engineering, 2014. 31(6):
p. 04014099.
International Journal of Safety Science
Vol. 01, No. 03, (2017), pp. 46-60
DOI:10.24900/ijss/01024660.2017.1201
59
7. Shea, T., et al., Leading indicators of occupational health and safety: An employee and workplace
level validation study. Safety Science, 2016. 85: p. 293-304.
8. Statistics, B.o.L., National Census Of Fatal Occupational Injuries In 2014, USDL-15-1789, Editor.
2015, , U.S. Department of Labor
9. Li, R.Y.M., Construction Safety Mobile Knowledge Sharing Among Generation Y. Romanian
Review of Social Sciences, 2012(3): p. 3-12.
10. Cheng, T. and J. Teizer, Real-time resource location data collection and visualization technology
for construction safety and activity monitoring applications. Automation in Construction, 2013. 34:
p. 3-15.
11. Navon, R. and O. Kolton, Algorithms for automated monitoring and control of fall hazards. Journal
of computing in civil engineering, 2007. 21(1): p. 21-28.
12. Karan, E. and A. Ardeshir, Safety assessment of construction site layout using geographic
information system, in Architectural Engineering National Conference: Building Integration
Solutions. 2008, American Society of Civil Engineers: Denver, CO. p. 634-647.
13. Golparvar-Fard, M., A. Heydarian, and J.C. Niebles, Vision-based action recognition of
earthmoving equipment using spatio-temporal features and support vector machine classifiers.
Advanced Engineering Informatics, 2013. 27(4): p. 652-663.
14. Li, H., G. Chan, and M. Skitmore, Visualizing safety assessment by integrating the use of game
technology. Automation in construction, 2012. 33: p. 498-505.
15. Luo, X., Requirements, specifications and deployment models for autonomous jobsite safety
proximity monitoring. 2013.
16. Luo, X., F. Leite, and W.J. O’Brien, Location-Aware Sensor Data Error Impact on Autonomous
Crane Safety Monitoring. Journal of Computing in Civil Engineering, 2014.
17. Zhou, W., J. Whyte, and R. Sacks, Construction safety and digital design: A review. Automation in
Construction, 2012. 22: p. 102-111.
18. Albert, A., et al., Enhancing construction hazard recognition with high-fidelity augmented virtuality.
Journal of Construction Engineering and Management, 2014. 140(7): p. 04014024.
19. Zhang, S., et al., BIM-based fall hazard identification and prevention in construction safety
planning. Safety Science, 2015. 72: p. 31-45.
20. Pinheiro, R., et al. Eye-Tracking Technology for Construction Safety: A Feasibility Study. in 33rd
International Symposium on Automation and Robotics in Construction (ISARC 2016) 2016. Auburn,
Alabama.
21. Dzeng, R.-J., C.-T. Lin, and Y.-C. Fang, Using eye-tracker to compare search patterns between
experienced and novice workers for site hazard identification. Safety science, 2016. 82: p. 56-67.
22. Hasanzadeh, S., B. Esmaeili, and M.D. Dodd. Measuring Construction Workers’ Real-Time
Situation Awareness Using Mobile Eye-Tracking. in Construction Research Congress 2016. 2016.
San Juan, Puerto Rico: ASCE.
23. Chauvin, B., D. Hermand, and E. Mullet, Risk perception and personality facets. Risk analysis,
2007. 27(1): p. 171-185.
24. Sjöberg, L., Factors in risk perception. Risk analysis, 2000. 20(1): p. 1-12.
25. Hampson, S.E., et al., Risk perception, personality factors and alcohol use among adolescents.
Personality and Individual Differences, 2001. 30(1): p. 167-181.
26. Tixier, A.J.-P., et al., Psychological antecedents of risk-taking behavior in construction. Journal of
Construction Engineering and Management, 2014. 140(11): p. 04014052.
International Journal of Safety Science
Vol. 01, No. 03, (2017), pp. 46-60
DOI:10.24900/ijss/01034660.2017.1201
60
27. Hallowell, M., Safety risk perception in construction companies in the Pacific Northwest of the USA.
Construction management and economics, 2010. 28(4): p. 403-413.
28. Ramsey, J.D., C.L. Burford, and M.Y. Beshir, Systematic classification of unsafe worker behavior.
International Journal of Industrial Ergonomics, 1986. 1(1): p. 21-28.
29. Vickers, J.N., Perception, cognition, and decision training: The quiet eye in action. 2007: Human
Kinetics. 267.
30. Stanton, N., P.M. Salmon, and L.A. Rafferty, Human Factors Methods: A Practical Guide for
Engineering and Design. 2013: CRC Press. 656
31. Chi, S., S. Han, and D.Y. Kim, Relationship between Unsafe Working Conditions and Workers’
Behavior and Impact of Working Conditions on Injury Severity in US Construction Industry. Journal
of Construction Engineering and Management, 2012. 139(7): p. 826-838.
32. Abdelhamid, T.S. and J.G. Everett, Identifying root causes of construction accidents. Journal of
Construction Engineering and Management, 2000. 126(1): p. 52-60.
33. Quinn, K., An in depth study of how the psychological behaviors affect construction safety, in
College of Technology Directed Projects. 2010.
34. Hutchings, D.M., et al., Safety Issues among Hispanic Construction Workers along the Wasatch
Front in Utah, in 48th ASC Annual International Conference Proceedings. 2012, Associated
Schools of Construction (ASC): Birmingham, UK.
35. Gillen, M., et al., Perceived safety climate, job demands, and coworker support among union and
nonunion injured construction workers. Journal of safety research, 2002. 33(1): p. 33-51.
36. Smith, G.S., et al., The relationship between safety climate and injury rates across industries: The
need to adjust for injury hazards. Accident Analysis & Prevention, 2006. 38(3): p. 556-562.
37. Murray, I.C., et al., Feasibility of saccadic vector optokinetic perimetry: a method of automated
static perimetry for children using eye tracking. Ophthalmology, 2009. 116(10): p. 2017-2026.
38. Yousefi, M., et al., Implementing Eye Tracking Technology in the Construction Process, in
Associated Schools of Construction (ASC) 51st Annual Internation Conference. 2015: College
Station, TX.
39. Lejuez, C., et al., Evaluation of a behavioral measure of risk taking: the Balloon Analogue Risk
Task (BART). Journal of Experimental Psychology: Applied, 2002. 8(2): p. 75.