Turing Trap / Automation Bias in AI Evaluation
Definition
Cognitive bias of evaluating AI progress by comparing performance to humans on specific benchmarks, then inferring that AI should replace humans. This conflates two different questions: (1) Can AI perform task X? (2) Should AI replace humans on task X? AI systems are imperfect, make hallucinations, and real tasks require judgment. The gap from 'AI beats human on test' to 'AI replaces human in practice' involves quality control, oversight, and judgment—which humans must provide.
Cast within
No topic-region cast yet — this appears once Turing Trap / Automation Bias in AI Evaluation's compiled claims are aligned into a topic region's argument tree.
Claims mentioning Turing Trap / Automation Bias in AI Evaluation (1)
Your notes
Loading notes...