The Persuasion Paradox: When LLM Explanations Fail to Improve Human-AI Team Performance

📰 ArXiv cs.AI

LLM explanations can increase user confidence but not necessarily improve human-AI team performance

advanced Published 7 Apr 2026
Action Steps
  1. Identify the types of tasks where LLM explanations may not improve performance
  2. Analyze the relationship between explanation fluency and user confidence
  3. Evaluate the impact of LLM explanations on task accuracy and user reliance on AI
Who Needs to Know This

Data scientists and AI engineers can benefit from understanding the limitations of LLM explanations to improve human-AI collaboration, while product managers should consider the potential impact on user trust and decision-making

Key Insight

💡 Fluent LLM explanations can increase user confidence without improving task accuracy

Share This
🤖 LLM explanations can boost confidence but not always accuracy #AI #LLMs
Read full paper → ← Back to Reads