When AI explains its decision, humans may stop thinking independently
Executive Take
Adding an explanation to an AI recommendation doesn't build trust, it kills scrutiny. If your teams use AI to screen deals, hires, or projects, strip out the rationale and force people to verify the call themselves.
Executive Summary
A study by researchers from Harvard Business School, MIT, and the University of Washington tested 228 evaluators reviewing MIT innovation proposals with AI recommendations. Evaluators agreed with AI decisions 75% of the time but human experts only 54%. AI recommendations with written explanations increased false negatives, while unexplained pass/fail recommendations improved decision accuracy.
Why It Matters
AI and technology leaders rolling out AI-assisted decision tools should care because this study shows explanations make people trust bad AI calls more, not less. That's a direct risk in hiring, funding, and screening decisions.