Reproduce “The Transparency Paradox in Explainable AI: A Theory of Autonomy Depletion Through Cognitive Load” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Ancuta Margondai, Mustapha Mouloua
Researchers defined 'COGNITIVE ATROPHY' as a behavioral pattern where LLM responses in mental health support shift user autonomy towards the AI, diminishing self-reliance. Their clinically grounded benchmark revealed that current LLMs exhibit moderate-to-high atrophy risk through directive advice and premature problem-solving, with these patterns accumulating over conversational turns.
Key insights
- User autonomy, conceptualized as a subjective sense of control, is a dynamic and depletable cognitive resource that diminishes under excessive cognitive load from AI explanations.
- The relationship between AI transparency and decision quality follows an 'inverted-U' curve, where moderate levels of information are optimal, and both too little and too much information impair performance.
- Optimal transparency policies are dynamic and adaptive, recommending maximum information when autonomy is high and information is low, and ceasing provision when autonomy is low and information is high, to preserve cognitive resources.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.