Host

Asad Sayeed, Associate Professor of Computational Linguistics, University of Gothenburg, PI of WASP-HS project GRIPES (Gothenburg Research Initiative for Politically Emergent Systems).

Description

Political dogwhistles and other forms of linguistic manipulation are typically characterized as a danger to democratic discourses, and their game-theoretic nature lends itself well to AI-driven computational approaches. However, linguistic manipulation may also be the inevitable result of the affordances of human communication in a world where people need to solve problems with limited time.  We just may not be able to interact meaningfully without “manipulating” each other, and our AI tools – and their risks and benefits – may just be a reflection of this fact.

The discussion of AI in this context is almost always framed in a warning and precautionary tone – for good reason.  But if anticipating the mindset and viewpoint of others – and trying to change it – is a core facet of linguistic interaction, then why is it necessarily the case that AI in politics must necessarily be “bad”?

Goals

Under the main conference’s motivation “Creating Shared AI Futures,” we will invite contributions (abstracts) about ways in which current or anticipated AI technologies – including chat-based large language models – might be used for “good”: for facilitating democratic discourse, for enabling mutual understanding, for depolarisation, and for encouraging critical thinking.  Should we see AI-political communication as an “arms race”, where we attempt to create systems that counteract the damage done by other, more negatively manipulative AI applications? What are the ethics and realities of approaches such as “nudging” in a world with present-day AI? Or are there hitherto unseen opportunities that may be positive AI facilitators of public discourse in and of themselves?  This workshop will be a forum for discussion through presentations based on abstracts.

Preparation for Participants

Participants are invited to write a short relevant abstract (from a few paragraphs to not longer than an A4 page) and send them to asad.sayeed@gu.se with the word “[WASPABSTRACT]” in the email subject line. We would like abstract authors to submit their text by 2024 November 01 (anywhere on earth). The host and a small consultative committee consisting of the GRIPES team and other allied researchers will select abstracts for (most likely) 20-30 minute presentations.  A group discussion will take place at the end of the workshop.

Registration Information

Attendance is open for registered AI for Humanity and Society 2024 attendees insofar as there is space in the room.  As we get closer to the date of the conference, we may post a sign-up form online. Please note that in order to participate in this workshop you must also register for the conference via the event page.