Tag
This position paper argues that AI systems used in high-stakes decision-making should reason similarly to their users and faithfully communicate that reasoning, and outlines a research agenda for achieving such 'cognitively-aligned AI'.
Harvard Business Review article discussing how AI tools can erode critical thinking and offering design principles to strengthen human reasoning instead.
RealMath-Eval is a benchmark of 224 real-world high school math exam responses that reveals a significant 'Evaluation Gap': state-of-the-art LLM judges perform poorly on authentic human reasoning (MSE ~2.96) compared to synthetic LLM-generated solutions (MSE ~1.17), due to higher diversity and surprisal in human error patterns.
This essay argues that the greatest risk of AI is not hallucinations but the gradual erosion of human verification skills, leading to a civilization that cannot question AI outputs.