Host

Nikhil Mahant, MSCA Postdoctoral Fellow, Uppsala University, Uppsala, Sweden, nikhil.mahant@filosofi.uu.se

Description

The visual and literary arts are rife with narratives of AI systems ‘manipulating’ humans or ‘lying’ to them to further their own aims. Some recent academic studies also seem to indicate that AI systems can induce non-ideal mental states (e.g., false beliefs) in their users in pursuing certain objectives (e.g., winning a game (Diplomacy), or performing tasks that are out of their reach (solving a captcha)).

This workshop will focus on the topic of deception by artificial systems. The possibility of deception by AI systems lies at the heart of many technological, philosophical, and governance related issues posed by AI systems including the question of alignment, the risks posed by ‘black box’ AI systems, trustworthiness of AI systems when implemented in specific areas of human activity (like recommender systems, autonomous weapons, etc.) Further, the issue of AI deception cuts across several domains of application of AI systems, and requires a combined expertise in philosophy, ethics, and policymaking to be appropriately addressed.

Goals

The workshop will bring together experts from various areas to pursue two main objectives. The first objective is to engage the participants with the latest research on the main theoretical positions concerning AI deception, which range from treating AI systems as capable of real deception (i.e., AI systems can deceive humans just as humans can deceive each other) to treating AI systems as potentially dangerous tools but principally incapable of deception (owing to lack of deceptive intentions or other relevant mental states).

The second objective is to foster a discussion and reflection on the possibilities and risks of AI deception in various areas of application and research. One way that the workshop aims to achieve this is by examining cases of purported deception by AI systems. Students, developers, and AI researchers will be invited to present cases/arguments through short talks or poster presentations.

Preparation for participants

Abstracts of talks and cases will be circulated, which the participants are expected to read before the workshop.

Registration information

Please first register for the main conference, then send an email to aideception2024@gmail.com expressing your interest. Register before 1 October, 2024.