In a recent experiment conducted by Google DeepMind, researchers examined the effectiveness of Artificial Intelligence in assisting group deliberations. In this study, participants were assigned into six-person groups, with one member designated as a "mediator." The role of the mediator was to draft statements on behalf of the group during discussions.

Throughout the experiment, each group engaged in deliberative exercises that involved reviewing statements. Participants were presented with two different statements per round: one crafted by the human mediator and another generated by an AI system. The task was for the participants to choose which of the two statements they deemed superior.

Interestingly, results from the study indicated a preference for AI-generated statements, with AI statements being selected 56% of the time over those written by human mediators. Participants cited a perception of higher quality in AI-generated content and showed stronger support for these statements. Furthermore, it was noted that the utilisation of AI assistance appeared to bridge some divides within the participant groups, leading to less polarisation in their viewpoints on the issues discussed.

While the ability of AI to synthesise and summarise group opinions presents an intriguing development, the findings highlight potential limitations and ethical considerations. Joongi Shin, a generative AI researcher from Aalto University, emphasised the importance of context and transparency. "Unless the situation or the context is very clearly open, so they can see the information that was inputted into the system and not just the summaries it produces, I think these kinds of systems could cause ethical issues,” he remarked.

Moreover, concerns were raised regarding the AI's limitations in simulating the complex decision-making processes of real-world deliberations. As noted by a researcher involved in the study, Tessler, the current AI model lacks capabilities crucial to effective mediation—such as fact-checking, maintaining relevant discourse, and moderating conversations.

The participants in the experiment were not explicitly informed that AI technology would be used to generate group opinion statements. Although the consent forms mentioned the involvement of algorithms, detailed disclosure about AI involvement was absent.

Future research could be instrumental in determining viable applications for this AI technology. Google DeepMind has stated that further investigations are necessary to explore responsible and safe applications of such models. As it stands, the company does not intend to release the AI model used in the experiment for public use. This cautious approach underscores the need for careful consideration and further development before integrating AI into decision-making contexts broadly.

Source: Noah Wire Services