There is no such thing as bias-free tech. How models shape our world.
summary
Argues that “bias” is not just bad data but assumptions built into models that become invisible constraints in software and hardware, so unbiased tech is impossible. Gives concrete cases: crash test dummy male standard causing lower car safety for women, Amazon’s biased recruiting tool and racist predictive‑policing outputs, Microsoft’s Tay chatbot failure, and differences between Arch Debian Ubuntu installer models that shape user skills and culture. Shows how Android 6 runtime permissions (install‑time vs runtime, internet as a normal permission) fixes some issues but embeds new limits, and calls for reflexive, contestable model design.
AI-written summary · about
chapters (12)
- 0:00Introduction to bias in tech
- 1:58Speaker background and research
- 2:40Examples of biased technology
- 4:08Why data models are broken
- 5:13Cultural subjectivity in models
- 7:43Modeling user behavior in tech
- 9:07Technology scripting behavior
- 11:54Moral inscription in design
- 14:07Models shaping technology
- 15:08Linux distribution examples
- 16:54Android permission models
- 18:38Conclusion and responsibility
related talks
25mIt's not a bug, it's a bias
34m(Un)Ethical Machines? Why AI Bias Is a Problem and What to Do About It
42mUnderstanding the nature of Bias in Natural & AI Systems
38mSocial Implications of Bias in Machine Learning
35mBad computers, bad people
26mAdventures in Ethical Computing
03mInterview with Alexandra Ebert
35mZero Bias AI