Growing public awareness around misinformation has led to increased scrutiny of misleading content on social and traditional media. However, the rapid emergence of generative artificial intelligence has further blurred the line between reality and manipulation. Widely accessible AI tools are now capable of producing convincing articles, operating social media bots to influence opinion, and creating realistic images of events that never occurred.

While individuals may learn to recognise false content, misinformation often spreads through trusted social relationships. Misleading beliefs about topics such as mental health, physical activity, or vaccinations can be particularly hard to challenge when they come from friends or family members. According to the researchers, critical knowledge and conversational skills are essential to identify misconceptions and mitigate their impact at both individual and societal levels.

“AI gives us a way to support people at scale without replacing human judgement. Our motivation was to address the gap between knowing that misinformation exists and actually being able to resist it in everyday conversations,” says professor Simo Hosio from the University of Oulu, Finland.

The research applies Cognitive Inoculation, a concept from social psychology originally developed to protect individuals against persuasion attempts. The principle assumes that, much like the immune system, the mind can become more resilient when exposed to weakened forms of misleading arguments in a controlled setting.

Decades of research have shown the effectiveness of this approach across areas such as health education, substance abuse prevention and child safety. However, its real‑world application has been limited due to the lack of professionals available to deliver conversational inoculation at scale.

To address this limitation, the research team, consisting of members from the University of Oulu and the University of Tokyo, led by Hosio, developed an AI‑based chatbot called Forty, designed to deliver structured, science‑based inoculation conversations online. The chatbot focuses on four common health‑related topics and their associated misconceptions: daily tooth brushing, the link between physical activity and mental wellbeing, alcohol use, and environmental protection.

“AI allows us to deliver evidence‑based educational conversations consistently and at population scale. This is something traditional approaches simply cannot do under current resource constraints”, notes researcher Dániel Szabó from University of Oulu.

The effectiveness of the approach was evaluated in a study in which Forty was deployed on the public website meetforty.com, allowing users to engage in cognitive inoculation-based conversations.

To validate the system, 65 participants were invited to compare the chatbot experience with traditional non‑conversational inoculation methods, such as reading educational materials or writing essays. The results showed that interaction with Forty increased participants’ resilience to health misinformation more effectively than either reading or writing. The results were published in the flagship conference of Human-Computer Interaction research field, ACM CHI, in April 2026. The article also received an Honourable Mention Award.

Although the precise underlying mechanisms behind this effect remain unclear, the researchers conclude that AI‑driven conversational approaches can significantly reduce vulnerability to health‑related misconceptions.

“These early results suggest that structured AI dialogue can support critical thinking in ways we did not fully anticipate,” says Academy Award-winning professor Simo Hosio, who has also worked as a visiting researcher at the University of Tokyo, in the Interactive Intelligent Systems Laboratory.

Building on the study, the research team organised a workshop at the Augment Humans conference in Okinawa, Japan, in March 2026. The event brought together researchers from Aalborg University, UNSW Sydney, the University of Melbourne and The University of Tokyo to explore how AI could foster other forms of human resilience.

Participants examined whether similar approaches could help individuals resist negative thought patterns, procrastination, or motivational challenges. The workshop focused on hands‑on discussion and rapid prototyping of potential technological solutions, which are now available on the workshop website beyondoffloading.org.

“The discussion moved quickly from theory to practice. What became clear is that resilience is not limited to misinformation alone”, says Dániel Szabó.

The next phase of research is already underway. Continuing collaboration with The University of Tokyo and expanding to new partners, the team is preparing to extend Forty’s role beyond misinformation. Future iterations will focus on mental health and challenging life situations, aiming to help users develop a resilient mindset before crises occur.

The researchers are exploring how conversational AI could support people in coping with everyday stress as well as unexpected life events, such as loss or sudden change.

“Our long‑term goal is not to replace human support but to provide accessible tools that help people prepare mentally for challenges before they arise”, muses professor Hosio.

Read the original article

by Dániel Szabó, Chi-Lan Yang, Aku Visuri, Jonas Oppenlaender, Bharathi Sekar, Koji Yatani, and Simo Hosio. 2026. Conversational Inoculation to Enhance Resistance to Misinformation. In Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems (CHI '26). Association for Computing Machinery, New York, NY, USA, Article 958, 1–17.

The research was supported by Research Council of Finland (356128, 349637 and 353790), Strategic Research Council (335625, 335729), and JST ASPIRE for Top Scientists (JPMJAP2405)