Wrong at the Root: Racial Bias and the Tension Between Numbers and Words in Non-Internet Data
2019-06-05..2019-06-07
Berkeley, United States
| title | author | length |
|---|---|---|
| Opening Remarks | Shafi Goldwasser, Cynthia Dwork, Patricia Williams | 15m |
Overview of workshop goals and concrete tensions: theoretical tools used to audit census confidentiality tools and to set mathematical limits on fairness… | ||
| Morning Session: Genomics, Medicine, Epidemiology and Data Interpretation | Marzyeh Ghassemi, Joan Fujimura, Nancy Krieger, Sendhil Mullainathan, Dorothy Roberts | 01h 51m |
Randomized controlled trials cover only about 10–20% of treatments and just 6% of current asthmatics would have qualified for those trials; observational… | ||
| Questions and Discussion | 29m | |
Canadian electronic health records often record no race in Canadian EHR, while France forbids racial categories for historical reasons, creating blind spots… | ||
| Afternoon Session #1: Science, Pseudo-science, Canon and Belief | Evelynn Hammonds, Jnoathan Kahn, Harriet Washington, James Zou | 01h 40m |
Google’s 2019 diversity numbers (tech hires: 25.7% women, 4.8% Black, 6.8% Latino, 1.1% Native American) set a context for concrete failures: MEDLINE/MeSH… | ||
| Questions and Discussion | 38m | |
Comparative word embeddings are computed from public corpora like Wikipedia, NYT, WSJ, BBC and tracked over 22 time points to measure covariance and… | ||
| Afternoon Session #2: Policies, Politics, Publics | Marcy Darnovsky, Alondra Nelson, Osagie Obasogie, Aaron Panofsky | 01h 51m |
Panel documents concrete misuses and loose claims: commercial polygenic scores are sold to score embryos and compute national differences; ancestry ads claim… | ||
| Morning Session: Carcerality | Ruha Benjamin, Sendhil Mullainathan | 01h 08m |
Examines concrete tech: the Citizen app pushes 911-based alerts and ankle monitors scale to ~125,000 monitored (2005–2015), while a California gang database… | ||
| Afternoon Session: Governance | Alexandra Chouldechova, Zachary Lipton, Martha Minow, Patricia Williams | 01h 00m |
A participatory study runs five workshops, 83 participants in one US county using a tangible comfort board to test six fictional child‑welfare scenarios while… | ||
| Questions and Discussion | Alexandra Chouldechova, Zachary Lipton, Martha Minow, Patricia Williams | 42m |
Argues that building a risk model to risk stratify families in one county produces targeted service bins but also correlates with higher inadvertent child… | ||
| Morning Session: Causality | Yiling Chen, Moritz Hardt, Jay Kaufman, Issa Kohler-Hausmann | 01h 19m |
Three linked analyses: a labor market model shows Bayesian hiring on noisy signals plus group reputations creates a self-fulfilling disparity but introducing a… | ||
| Afternoon Session: Individuals, Groups, Pattern Recognition and Rich Data | Cynthia Dwork, Jamie Morgenstern | 43m |
Distinguishes group vs individual fairness, arguing individual fairness requires a task-specific metric and showing a small number of queries to a fair arbiter… | ||
| Questions and Discussion | Cynthia Dwork, Duana Fullwiley, Jamie Morgenstern | 42m |
Panelists argue for concrete outputs: a short list of foundational papers, shared reading resources, and new venues that reward domain-specific, reproducible… | ||
| nothing matches | ||