Fair Algorithms (Revisited) Privacy: The Last Stand forPhoto: YouTube screengrab found via Google...
Transcript of Fair Algorithms (Revisited) Privacy: The Last Stand forPhoto: YouTube screengrab found via Google...
Privacy: The Last Stand for Fair Algorithms (Revisited)Data Rights for All
Katharine Jarmul - KIProtectQCon AI 2019
Left Photo: http://peterfleischer.blogspot.com
Peter Fleischer,
Google’s Global Privacy Counsel and longest serving privacy officer
Left Photo: http://peterfleischer.blogspot.com Right Photo: YouTube screengrab found via Google Search
Photo: YouTube screengrab found via Google Search
On the Right to be Forgotten
“... all of my empathy for wanting to let people edit-out some of the bad things of their past doesn't change my conviction that history should be remembered, not forgotten, even if it's painful. Culture is memory.”
Photo: YouTube screengrab found via Google Search Article: https://www.theguardian.com/technology/2015/jul/14/google-accidentally-reveals-right-to-be-forgotten-requests
On the Right to be Forgotten (Factual Edition from The Guardian)
95% of Google privacy requests are from citizens out to protect personal and private information – not criminals, politicians and public figures
🤔
Privacy as Privilege
Loyalty Program
Your Name Here2342 432 4322432
Privacy as Power
Privacy as Security
Privacy as Privilege,PowerAnd Security
Source: https://www.sfchronicle.com/archive/item/A-decade-of-homelessness-Thousands-in-S-F-30431.php
Privacy and Fairness
Fairness Through Awareness
Dwork et al. Fairness Through Awareness, 2011.
Learning Fair RepresentationsMinimum Discrimination Max Delta (Accuracy - Discrimination)
Zemel et al. Learning Fair Representations, 2013
Implement Private and Fair Machine Learning
Source: https://blog.godatadriven.com/fairness-in-ml
Collect Data with Privacy Guarantees: Group Fairness, Individual Privacy
Source: https://www.nytimes.com/2017/09/08/technology/google-salaries-gender-disparity.html
Privacy By Design: Protect User Data in Software Design
Privacy as a Right:
EnablePrivacyFor All Users
Choose Fairness and Privacy Metrics Early and Often...
...But Please Avoid the Fallacy of Metrics
Photo: BoingBoing
Give Users:
- Agency in Defining Their Privacy
- Transparency in Defining Fairness
Photo: danah boyd’s Medium
Privacy and Privilege,
Revisited
Source: Tim Evanson (Flickr)
Thank you!
7scientists GmbHKIProtect
Bismarckstr. 10-1210625 Berlin
Questions? I’d love to hear them!
Or reach out anytime:
[email protected]@KIProtect (Twitter)https://github.com/kiprotect
Katharine [email protected] @kjam (Twitter)
Appendix: Private and Fair Representations
1. the mapping from X0 to Z satisfies statistical parity;
2. the mapping to Z-space retains information in X (except for membership in the protected set);
3. the induced mapping from X to Y (by first mapping each x probabilistically to Z-space, and then mapping Z to Y ) is close to f.
Zemel et al. Learning Fair Representations, 2013
Slide References- Right to be Forgotten Request Data Leak:
https://www.theguardian.com/technology/2015/jul/14/google-accidentally-reveals-right-to-be-forgotten-requests- Facebook IQ: https://www.facebook.com/business/products/ads/ad-targeting/ - Siri: https://www.apple.com/siri/ - Dwork et al., Fairness Through Awareness https://arxiv.org/abs/1104.3913 - Zemel et al., Learning Fair Representations http://www.cs.toronto.edu/~zemel/documents/fair-icml-final.pdf - GoDataDriven - Fairness in ML: https://blog.godatadriven.com/fairness-in-ml - Google Salary Chart: https://www.nytimes.com/2017/09/08/technology/google-salaries-gender-disparity.html - Green Party Privacy Poster: https://www.gruene-cochemzell.de/page/37/ - Fairness Metrics: https://algorithmicfairness.wordpress.com - BoingBoing Project Maven: https://boingboing.net/2018/05/15/drone-be-evil.html- Tim Evanson (Flikr): https://www.flickr.com/photos/timevanson/6974619755 - Microsoft Predictive Policing:
https://enterprise.microsoft.com/en-us/industries/government/predictive-policing-the-future-of-law-enforcement/