Letter in Science (2025): 'AI Mediators May Change Deliberative Outcomes' by Joshua Cohen & Henrik Kugelberg

In a short letter published in Science (May 2025), Joshua Cohen and Henrik D. Kugelberg respond to the “Habermas Machine” study by Tessler et al., raising concerns about the role of perceived objectivity in how participants evaluated AI-mediated deliberation. They argue that trust in the system may stem from a misunderstanding of its design choices—like ranking methods or model tuning and call for further research into how such perceptions shape outcomes.

Read the letter here.