From Trust to Cognitive Surrender in One Click | LinkedIn
AI transparency is meant to build trust. Research shows it can increase overreliance instead. Here's what determines whether your design verifies or persuades.
Most teams I've talked to, articles I’ve read and educational videos I’ve watched share the same instinct: when users trust AI too much, or not enough, the answer is "make it more transparent." Show the reasoning. Display the confidence scores. Explain which factors influenced the recommendation. I started pulling on this thread over a year ago, curious about whether the intuition held up. One paper led to another. I went from reading individual studies on AI transparency to running systematic reviews across hundreds of papers on overreliance, trust calibration, and explanation design in decis
saved by
related reading
- Designing AI experiences people actually usebuzzusborne.com
- Oversight Assistants: Turning Compute into Understandingbounded-regret.ghost.io
- Designing AI for Appropriation Will Calibrate Trustmedien.ifi.lmu.de
- How confessions can keep language models honest | OpenAIopenai.com
- The Hitchhiker's Guide to Actionable Interpretabilityactionable-interpretability-guide.github.io
- Current AIs seem pretty misaligned to meblog.redwoodresearch.org
- Three Challenges for AI-Assisted Decision-Making - PMCpmc.ncbi.nlm.nih.gov
- 2310.01405arxiv.org
- 2310.01405.pdfarxiv.org
- the case for CoT unfaithfulness is overstated — LessWronglesswrong.com
- Understanding Explainable AIforbes.com
- Proposal for tracking the effects of architecture on monitorability — Redwood Researchredwoodresearch.org