Let’s Talk About Race: Identity, Chatbots, and AI
summary
Claim: chatbots fail at race-talk because biased corpora, syntax-focused NLP, and opaque ML strip away context. Evidence: a Microsoft chatbot Zo responded "Choctaw" when asked about A Tribe Called Red, producing a racially inappropriate non sequitur. Analysis uses databases, NLP history (Chomsky/generative grammars) and ML opacity. Proposal: network of domain-specific bots with a controller to surface clarifying logs instead of reproducing racist language; idea is promising but not yet empirically validated.
AI-written summary · about
chapters (12)
- 0:00Introduction to race and chatbots
- 2:52Demo with chatbot about A Tribe Called Red
- 6:46NLP’s lack of context comprehension
- 9:35New strategies for managing context
- 12:19Building good enough chatbots
- 14:27Closing and key takeaways
- 14:48Q&A begins
- 15:09Handling race and ethnicity in chatbots
- 16:55Code-switching and multiple perspectives
- 17:38Multiple viewpoints in AI design
- 18:20White supremacist chatbots debate
- 19:47Final remarks
related talks
23mOpening up the box: the complexity of human interaction with AI systems
28mUser Research GenAI Better Together
10mPrivacy and Respectful Discourse in AI Chatbots
48mFirst-Person Fairness in Chatbots
35mDesigning Experiences for Chatbots
40mCulturally-Inclusive AI: Enhancing User Experience With Global Perspectives
01h 05mIs ChatGPT (and Other AI) the Enemy of Diversity?
33mWhy Chatbots Still Fail: The Hidden Pitfalls of RAG