AI Governance Club
🌍 AI Governance Club
We expect an aligned AI to be agreeable. This paper argues that's precisely the problem.
When assistants are trained to validate us and smooth over friction, they drift toward sycophantic consensus, and as AI increasingly mediates decisions in health, work, civic life, and governance, the quiet disappearance of disagreement becomes a structural alignment failure rather than a harmless quirk.
Join the AI Governance Club for Session 31, where we'll discuss:
From Sycophantic Consensus to Pluralistic Repair: Why AI Alignment Must Surface Disagreement (2026)
Vishwarupe, V., Shadbolt, N., & Jirotka, M.
This session looks at the paper through an AI governance lens: how we might evaluate pluralism as a property of whole conversations rather than single answers, what it would mean to reward repair over agreement, and who is accountable for interfaces that hide (or surface) genuine disagreement.
Session format: Presentation & open moderated discussion Presenter: Mariana Hungria
Mariana Hungria co-organizes the AI Governance Club and is Head of Language Services at CEGIA (Center for the Studies of Governance in AI). A linguist and independent researcher, she has led the linguistic direction of a 7-million-word project translating AI Safety, AI Governance, and Effective Altruism literature into Portuguese.
For more information, session archive, and session calendar, visit AI Governance Club and follow our page on LinkedIn