Google DeepMind develops AI 'Habermas Machine' that shows a consensus on conflicting social issues
In modern society, where various social issues such as 'immigration' and 'climate change' are being discussed, these issues can trigger conflicts. To address such conflicts, Google DeepMind has reported that it has developed an AI tool called the 'Habermas Machine ' that can generate consensus that reflects the opinions of both the majority and minority.
AI can help humans find common ground in democratic deliberation | Science
AI tool helps people with opposing views find common ground
AI mediation tool may help reduce culture war rifts, say researchers | Artificial intelligence (AI) | The Guardian
https://www.theguardian.com/technology/2024/oct/17/ai-mediation-tool-may-help-reduce-culture-war-rifts-say-researchers
For smooth group life, it is necessary to reach a consensus by combining various opinions. However, when many valid opinions are submitted, it is difficult to reach a consensus that everyone can agree on. So Christopher Summerfield, research director at the UK AI Safety Laboratory, together with Google DeepMind, investigated whether AI can help groups reach consensus in democratic discussions.
The research team created the 'Habermas Machine' based on the large-scale language model ' Chinchilla ' developed by Google DeepMind. The name 'Habermas Machine' is named after the German philosopher Jurgen Habermas .
The Habermas Machine is able to aggregate all the opinions of the individuals in the group to generate a set of group statements that are acceptable to all, and the group members can then evaluate the generated statements to provide feedback and training to the system, or even to generate more refined statements.
The research team gathered 439 British citizens and divided them into 75 groups of six people each, and conducted an experiment in which they discussed three topics related to British public policy, shared their personal opinions on each topic, and then created a comprehensive opinion. In this experiment, one participant from each group was designated as a mediator and summarized the opinions of the group, and the Habermas machine also summarized them at the same time. The participants were then shown both the summary created by the human mediator and the summary created by the Habermas machine, and asked to evaluate which one was better.
As a result of the experiment, 56% of participants supported the summary created by the Habermas machine, while 44% supported the summary created by the human intermediary, and it was found that the majority of participants appreciated the summary created by the AI. In addition, when external judges were asked to evaluate the summaries, it was reported that the summary created by the Habermas machine received high marks in terms of fairness, quality, and clarity.
The research team also reported that when the Habermas Machine mediated discussions between participants, the likelihood of reaching agreement within the group increased by an average of 8% compared to when there was no mediation.
Summerfield argued, 'What the Habermas machine does is to come up with a compromise that broadly respects the majority's opinion while not making the minority feel that their rights are being taken away. The Habermas machine can fulfill some of the functions of modern citizen assemblies and debate-based polls, and can make the conflict resolution process faster and more efficient.' In addition, Ethan Busby of Brigham Young University said, 'The Habermas machine can be used in a variety of fields, such as assisting in deliberations that have traditionally been performed by human moderators. I see the Habermas machine as cutting-edge research with great potential to address pressing social and political issues.'
'We want political leaders to use the Habermas Machine to better understand what their people are thinking,' Summerfield said of the Habermas Machine.
On the other hand, some researchers are skeptical of the Habermas machine. Melanie Gerson of University College London criticized it, saying, 'The Habermas machine does not allow participants an opportunity to explain their feelings, and therefore does not foster empathy with participants who hold different opinions.' 'The role of mediation is not simply to create consensus.'
Related Posts:
in Software, Posted by log1r_ut