At the high-profile workshop AI as a Colleague? – Towards a Social Epistemology of AI at the University of Bern, Alina Jacobs will give a talk on epistemic paternalism.

She examines how epistemic paternalism provides a framework for assessing AI’s influence on inquiry. She states that paternalistic structures arise across three domains of AI practice: the operation of AI systems, the design choices of AI developers, and the aims of AI alignment research.

1. Certain AI systems guide users’ inquiries without consultation, aiming to correct biases or improve epistemic outcomes.
2. AI developers form epistemic environments through interface design, data selection, or default configurations, thereby occupying paternalistic roles.
3. AI alignment initiatives presuppose preferred epistemic trajectories and pursue them with limited public input.

She argues that criteria from the debate on epistemic paternalism, such as whether an interference fosters genuine epistemic improvement, whether its benefits outweigh its costs, and whether it is sensitive to the agent’s needs and capabilities, offer tools for evaluating when AI supports, and when it constrains, epistemic autonomy.

Her talk takes place on the first day, Thursday, 26th March 2026 12:00-13:00

About the Event

The past few years have witnessed tremendous progress in the development of AI. DNNs and other AI tools have increasingly demonstrated their ability to perform epistemic tasks, such as recognizing patterns in data, diagnosing diseases, or exploring solution spaces to problems. The epistemic power of some AI tools suggests that they are not just epistemic instruments but rather becoming epistemic agents of their own. If this is true, new perspectives on AI emerge.

The use of AI can be studied through the lens of social epistemology, a thriving branch of epistemology that emphasizes the social nature of acquiring and transmitting knowledge. But:

• Is it really appropriate to analyze AI using concepts and ideas from social epistemology?
• Can AI systems be experts or epistemic authorities?
• Is epistemic trust towards AI models justified?
• Can social epistemology help to deal with difficult situations, e.g., when AI and human judgment disagree?
• Which kinds of collaborations between humans and AI tools are most promising for improving medical diagnoses?
• Might it be sensible to assign group beliefs to coupled systems involving humans and AI tools?
• And what consequences does the social epistemology perspective have for the explainability of AI?

This workshop aims to bring together social epistemology and the philosophy of AI to address such questions. Ultimately, we hope to advance our understanding of AI and its applications by utilizing the tools of social epistemology. A special focus will be laid on the example of medicine and the use of AI tools in this domain.

Program:

Day 1 – Thursday, 26th March 2026

Day 2 – Friday, 27th March 2026

Organizers: Andreas Wolkenstein (LMU Munich) and Claus Beisbart (University of Bern)

Funding by the Swiss National Science Foundation is gratefully acknowledged.