`It's Reducing a Human Being to a Percentage'; Perceptions of Justice in Algorithmic Decisions
summary
European data-protection rules require meaningful information about the logic behind loan, insurance and hiring decisions; three experiments test four explanation styles (local input influence like LIME, sensitivity analysis, case‑based, demographic) across loans, recruitment, insurance, fraud and overbooking scenarios. Results show explanation style changes justice perceptions only when multiple styles are compared side‑by‑side; repeated exposure to one style yields no effect. Notably case-based explanations decrease perceived fairness, while sensitivity analysis increases fairness by offering actionable alternatives.
AI-written summary · about
chapters (10)
- 0:00Introduction to research
- 2:31Algorithmic decision making and transparency
- 6:23Justice perceptions in decisions
- 8:52Study design and explanation styles
- 10:19Key findings on justice perceptions
- 13:09Conclusion and implications
- 14:19Q&A on transparency and justice
- 15:42Q&A on decision roles
- 17:28Q&A on decision maker expertise
- 18:32Q&A on feedback neutrality
related talks
17mExplaining models: an empirical study of how explanations impact fairness judgment
41mLow rate loans for ladies, stags pay extra: The Role of Ethics in AI/ML
47mDecoding the algorithm - why explainability and transparency matter when buildin...
34mInterpretability and Algorithmic Fairness
44mProcedural Justice in Algorithmic Fairness: Transparency, Control and Participation Perspectives
30mBias, Explainability, and Accountability: The Data Scientist's Burden
40mQuestions and Practices to Make Algorithmic Decision Making More Fair
01h 21mEmpirical methods fairer AI - Auditing a Dutch public sector risk profiling algorithm