ABSTRACT
Potential negative outcomes of machine learning and algorithmic bias have gained deserved attention. However, there are still relatively few standard processes to assess and address algorithmic biases in industry practice. Practical tools that integrate into engineers' workflows are needed. As a case study, we present two tooling efforts to create tools for teams in practice to address algorithmic bias. Both intend to increase understanding of data, models, and outcome measurement decisions. We describe the development of 1) a prototype checklist based on existing literature frameworks; and 2) dashboarding for quantitatively assessing outcomes at scale. We share both technical and organizational lessons learned on checklist perceptions, data challenges and interpretation pitfalls.
- Ricardo Baeza-Yates. 2016. Data and algorithmic bias in the web. In Proceedings of ACM Web Science. ACM, 1--1. Google ScholarDigital Library
- Joy Buolamwini and Timnit Gebru. 2018. Gender shades: Intersectional accuracy disparities in commercial gender classification. In Conference on Fairness, Accountability and Transparency. 77--91.Google Scholar
- Kate Crawford. 2017. The trouble with bias. In Conference on Neural Information Processing Systems, invited speaker.Google Scholar
- Michael D Ekstrand, Mucun Tian, Mohammed R Imran Kazi, Hoda Mehrpouyan, and Daniel Kluver. 2018. Exploring author gender in book rating and recommendation. In Proceedings of RecSys'18. ACM, 242--250. Google ScholarDigital Library
- Avriel C Epps and Travis L Dixon. 2017. A Comparative Content Analysis of Anti-and Prosocial Rap Lyrical Themes Found on Traditional and New Media Outlets. Journal of Broadcasting & Electronic Media 61, 2 (2017), 467--498.Google ScholarCross Ref
- Batya Friedman and Helen Nissenbaum. 1996. Bias in computer systems. ACM TOIS 14, 3 (1996), 330--347. Google ScholarDigital Library
- Alexandra Olteanu, Carlos Castillo, Fernando Diaz, and Emre Kiciman. 2016. Social data: Biases, methodological pitfalls, and ethical boundaries. (2016).Google Scholar
- D Sculley, Todd Phillips, Dietmar Ebner, Vinay Chaudhary, and Michael Young. 2014. Machine learning: The high-interest credit card of technical debt. (2014).Google Scholar
- Aaron Springer and Henriette Cramer. 2018. Play PRBLMS: Identifying and Correcting Less Accessible Content in Voice Interfaces. In Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems. ACM, 296. Google ScholarDigital Library
- Elizabeth H Stokoe. 2004. Gender and discourse, gender and categorization: Current developments in language and gender research. Qualitative Research in Psychology 1, 2 (2004), 107--129.Google ScholarCross Ref
- Allison Woodruff, Sarah E Fox, Steven Rousso-Schindler, and Jeffrey Warshaw. 2018. A Qualitative Exploration of Perceptions of Algorithmic Fairness. In Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems. ACM, 656. Google ScholarDigital Library
Index Terms
- Translation, Tracks & Data: an Algorithmic Bias Effort in Practice
Recommendations
Mitigating bias in algorithmic hiring: evaluating claims and practices
FAT* '20: Proceedings of the 2020 Conference on Fairness, Accountability, and TransparencyThere has been rapidly growing interest in the use of algorithms in hiring, especially as a means to address or mitigate bias. Yet, to date, little is known about how these methods are used in practice. How are algorithmic assessments built, validated, ...
Problematic Machine Behavior: A Systematic Literature Review of Algorithm Audits
CSCWWhile algorithm audits are growing rapidly in commonality and public importance, relatively little scholarly work has gone toward synthesizing prior work and strategizing future research in the area. This systematic literature review aims to do just that,...
How to Train a (Bad) Algorithmic Caseworker: A Quantitative Deconstruction of Risk Assessments in Child Welfare
CHI EA '22: Extended Abstracts of the 2022 CHI Conference on Human Factors in Computing SystemsChild welfare (CW) agencies use risk assessment tools as a means to achieve evidence-based, consistent, and unbiased decision-making. These risk assessments act as data collection mechanisms and have been further developed into algorithmic systems in ...
Comments