← CHI 2026 map

CHI '26 · Honorable mention · full-paper review · confidence medium-high

Conversational Inoculation to Enhance Resistance to Misinformation

Dániel Szabó , Chi-Lan Yang , Aku Visuri , Jonas Oppenlaender , Bharathi Sekar , Koji Yatani , Simo Hosio

DOI PDF Program page Page JSON

This is a timely CHI paper with a clear method contribution: it turns inoculation against misinformation into a conversational chatbot interaction and backs that claim with a controlled online comparison. The novelty is credible and the evaluation is solid, though the scope is still bounded by topic selection, short-delay testing, and a passive control.

Axes Lens

Rare contribution shape, typical evidence profile. The point here is not a score. It is to show what kind of claim the paper makes, and whether the evidence pattern is unusual or baseline in this 268 -review set.

Review Summary

This paper’s main value is that it reframes misinformation inoculation as an interactive conversational practice rather than a static informational intervention. That is a meaningful CHI contribution because it connects a well-known psychological idea to a concrete HCI system design: MindFort, an LLM-powered chatbot that guides users through refutational rehearsal in real time. The evidence packet supports both the novelty claim and the validation claim: the authors explicitly introduce Conversational Inoculation as a paradigm, describe the system, and evaluate it in an online within-subject experiment with 65 participants against reading and writing conditions. That makes the paper more than a speculative design concept; it is an empirically grounded method paper with a system artifact. At the same time, the paper’s own limitations matter for interpretation. The study is bounded to a small set of topics, all in health, wellbeing, and nature, and the authors note difficulty creating a long delay between inoculation and attack. They also acknowledge that the passive control condition prevents isolating the specific effect of refutation content, and that the linguistic analysis did not yield significant overall markers. So the strongest reading is not that the paper solves misinformation resistance in general, but that it demonstrates a promising and scalable interaction pattern for a specific class of topics and a specific experimental setup. In CHI terms, that is a strong honorable-mention-level contribution: novel, well-motivated, and empirically supported, but still early-stage in generality and deployment maturity.

What Changed

Canon before

Prior CHI work on inoculation and misinformation resistance largely emphasized static messages, reading-based interventions, or writing/refutation exercises rather than a structured conversational chatbot that actively guides refutation practice.

Departure from common sense

The paper’s core move is to treat misinformation resistance as something that can be trained through an interactive, real-time chatbot conversation. That departs from the more familiar intuition that inoculation is mainly delivered through passive exposure, reading, or one-off refutation tasks, and instead frames the bot as an active debate partner that rehearses resistance.

Actual novelty

The paper introduces Conversational Inoculation as a novel paradigm and implements it in MindFort, an LLM-powered web system. The novelty is not only the chatbot artifact but the interaction design: structured, interactive conversation for inoculation, with the bot proactively guiding users through threat recognition and refutational rehearsal.

Evidence

The paper defines Conversational Inoculation as a structured, interactive chatbot-based inoculation paradigm and evaluates MindFort in an online within-subject experiment with 65 participants. It compares chatbot-based inoculation against reading and writing treatments across four topics, collecting certainty scores, IMI responses, open feedback, and chat/writing logs. The discussion also states concrete limitations around delay, topic scope, control condition, and linguistic analysis.

Limits

Method limits

The study is constrained by an online within-subject design, a limited set of topics, and a passive control condition that does not isolate the specific contribution of refutation content. The paper also notes that linguistic analysis did not identify significant overall markers, suggesting possible Type II error.

Deployment limits

The approach depends on a chatbot-mediated interaction and therefore inherits practical constraints around model behavior, user trust, and friction in conversation. The paper’s evidence is strongest for the tested web-based setting and the selected misinformation topics, not for broad deployment across all misinformation domains.

Boundary conditions

The paper itself notes difficulty introducing a long interval between inoculation and attack, limited coverage to health, wellbeing, and nature topics, and inability to separate refutation-content effects from the passive baseline. These conditions bound the claims to short-delay, topic-specific, online experimental use.

Manual check: false

Position in field

This work sits at the intersection of cognitive inoculation, misinformation interventions, and conversational AI. Its contribution is best read as a method and system proposal that extends inoculation from static or writing-based formats into an interactive chatbot setting, with empirical evidence that the approach is viable and promising.

Abstract

Proliferation of misinformation is a globally acknowledged problem. Cognitive Inoculation helps build resistance to different forms of persuasion, such as misinformation. We investigate Conversational Inoculation, a method to help people build resistance to misinformation through dynamic conversations with a chatbot. We built a Web-based system to implement the method, and conducted a within-subject user experiment to compare it with two traditional inoculation methods. Our results validate Conversational Inoculation as a viable novel method, and show how it was able to enhance participants' resistance to misinformation.A qualitative analysis of the conversations between participants and the chatbot highlighted adaptability, independence, trust and friction as the main factors affecting Conversational Inoculation.We discuss the opportunities and challenges of using Conversational Inoculation to combat misinformation. Our work contributes a timely investigation and a promising research direction in scalable ways to combat misinformation.