The World Economic Forum’s (WEF) Global Risks Report ranks social polarization as the fourth-greatest risk to economy and society. The loss of a shared understanding of reality due to algorithmically amplified echo chambers is devastating for the formation of well-informed opinions and dialogue with those who hold different views. The AI chatbot “depolarAIze”, developed at the University of Zurich (UZH), engages in conversation with people, moderates discussions, and thereby promotes a balanced formation of opinions.
Open dialogue between people with different opinions and attitudes is one of the cornerstones of a pluralistic society. However, this principle is under threat, and democratic values are increasingly under pressure. One of the reasons for this is technological advancement.
Thanks to smartphones and computers, more and more people have access to a wealth of information anytime, anywhere. However, the increased use of online services also carries risks: algorithms can steer users into so-called echo chambers, where they are primarily confronted with information and opinions that confirm their existing beliefs. As a result, views can become increasingly entrenched and reinforced, leading to social polarization or, in extreme cases, even radicalization. At the same time, openness to other opinions is rapidly declining due to these echo chambers. Among other things, this can fuel populism, weaken trust in democratic institutions, and foster intolerance and exclusion.
This problem is also coming to a head in Switzerland. Compared to other European countries, our country has one of the highest polarization rates. Countermeasures have been very costly so far. Facilitating face-to-face encounters cannot be implemented on a nationwide scale, as it requires significant human resources and is hardly scalable.
How can this be countered? With the same tools that these “echo chambers” use to polarize users: through the use of digital technologies such as artificial intelligence.
Based on this idea, the AI chatbot “depolarAIze” is set to be developed at UZH. By interacting with users, this little helper aims to help highlight different perspectives on a given topic. Using large language models (LLMs) and access to a broad database, the chatbot provides access to diverse viewpoints, thereby fostering openness toward other perspectives. The chatbot can respond individually to each person and their specific situation.
This is intended to foster a dialogue that addresses not only the communication of facts but also emotional and social aspects. Users learn to better understand other perspectives and to engage with them respectfully. The project is being implemented on an interdisciplinary basis: The Institute of Computer Science at UZH is collaborating with the Institute of Political Science to incorporate diverse academic perspectives. A central component of the project is also the responsible use of artificial intelligence, taking ethical and societal aspects into account.
In the long term, the AI chatbot is to be integrated into a media portal and used there in forums, comment sections, and in response to reader feedback. In this way, it could become part of everyday digital life and promote constructive dialogue among people with different perspectives.
The project relies on third-party funding to implement the chatbot. This funding will enable the development of various key components: The technological foundation is an adaptive LLM chatbot based on a “conversational strategy module” and specifically tailored to the Swiss system. Users will be recruited for studies to conduct a scientific evaluation. Tests will also be conducted to assess the chatbot’s user-friendliness. In addition to funding two postdoctoral positions, the funds will also be used for the necessary infrastructure and server capacity. In the next phase of development, the chatbot is ultimately intended for practical use.
