AI Collaboration Workshop in Professional Practice

Large Language Models (LLMs) present a dual challenge for professional communities: they function simultaneously as a tool that people work with and also as an “active participant” in the generation of novel, shared knowledge. Drawing on Ostrom's design principles for governing common pool resources and research on human-AI collaboration, we present a framework that addresses the gap between ad hoc AI adoption and systematic local governance of AI-assisted professional practice that complements institution-wide policies.

The need for this workshop came from a graduate course being taught at the University of Sydney where students work in collaborative groups on a large project. With the extensive adoption of LLMs it was clear that students needed guidance on how to work with AI as a tool/agent and both the positive and negative influence this can have on a group's collective output. This led to adopting Ostrom's framework for managing common pool resources. Here the common pool is the knowledge generated in teams: “This framework seems well-suited for analysis of resources where new technologies are developing at an extremely rapid pace. New information technologies have redefined knowledge communities, juggled the traditional world of information users and information providers; made obsolete many of the existing norms, rules, and laws; and have led to unpredicted outcomes.” (Hess and Ostrom, 2005) Institutional change is occurring at every level of the knowledge commons, and is informed by recent work on AI `machine behaviour' (Rahwan et al 2019).