"I Have a Narrow Thought Process": Constraints on Explanations Connecting Inferences
summary
Study measures how people link platform-derived traits to themselves and finds explanations hinge on memory and plausibility. Participants who actually saw their inferred labels are rare, with 95 respondents in the survey and 21 interviews for followup. Codified reactions fall into plausible, implausible, and no connection categories; all 21 assume the inferences target advertising, 17 tolerate the business model while 20 report discomfort. Recommendation is to limit automated inferences to those users can plausibly relate to.
AI-written summary · about
chapters (6)
related talks
09mAnchoring Bias Affects Mental Models and User Reliance in Explainable AI Systems
46mAI for High-Stakes Decision Making
05mConceptualizing attribution and why it matters
24mYou Are Who You Know and How You Behave: Attribute Inference Attacks...
12mSafety Perceptions of Generative AI Conversational Agents: Uncovering Perceptual...
32mConsequences of an Insightful Algorithm
37mWhen better retrieval makes agents worse
01h 08mUnwanted Learning: What It Is, Why It Matters, and What To Do About It