Investigating algorithmic systems - Algorithm auditing.
summary
Panel lays out concrete audit methods and real-world tests: code review, scraping and sock-puppets, user surveys and browser-plugin collaborative audits, with tradeoffs (Terms of Service, ethics, gaming). Empirical study separates Twitter’s input data from its ranking bias; FOIA requests force agencies to reveal scoring like CBP’s 0–1 risk metric; examples (VW, Uber) show systems can be deliberately gamed. Legal levers (GDPR, FOIA, sector rules) plus funded civil‑society capacity are necessary to inspect, reproduce and remediate.
AI-written summary · about
chapters (10)
- 0:00Algorithm auditing introduction
- 3:15EPIC's transparency work
- 11:52Limitations of algorithm focus
- 17:51Methods and challenges in auditing
- 27:07Legal frameworks for transparency
- 32:45Auditing from inside versus outside organizations
- 41:08Regulatory and democratic challenges
- 56:22Designing audit frameworks
- 1:01:16Balancing checklists and safeguards
- 1:11:45Q&A
related talks
19mThe Case for Auditing Algorithms: Disloyalty and Bias M
01h 16mBEYOND ETHICS WASHING:IMPACT ASSESSMENTS, AUDITS, AND OVERSIGHT FOR AI
32mPlatform-supported Auditing Of Social Media Algorithms For Public Interest
01h 08mBeyond ink and paper: software tools for monitoring ...
01h 06mPanel on Interpretability in the Law
01h 12mObscured by algorithms: how to prove "yes" if computer says "no"
01h 18mACADEMIC SESSION 3
53mIt’s Not Fair! Detecting Algorithmic Bias with Open Source Tools