The next paper in this session at the 2026 International Communication Association conference in Cape Town is by Inbal Klein-Avraham and Ayelet Baram-Tsabari; their focus is on multiple documents literacy: the ability to work with multiple content sources to evaluate information. This involves content evaluation, source evaluation, and corroboration or triangulation of information from trustworthy sources.
Exposure to conflicting texts creates a cognitive conflict which demands resolution; AI chatbots are increasingly used to help with such processes, but this requires new approaches. The key issue here is to allocate epistemic authority (to persons, institutions, or technologies), to support informed decision-making. Incorrect allocation might lead people to believe in information from problematic and misleading sources.
Generative AI chat systems may appear to have considerable epistemic authority, since they present information confidently and fluently; this might mask issues with the quality of such systems, and a reliance on AI response – for instance for health-related decisions – might be deeply problematic.
The project conducted an online survey of some 800 respondents in Israel which incorporated a vignette-based task with a health question which required an initial decision, engagement with GenAI responses on the question, and a second decision informed by this additional information. This used authentic screenshots from ChatGPT and Perplexity, which were variously aligned or misaligned in their use of reliable or unreliable sources. Several other respondent-level variables were also captured.
The results of this study show that pre-exposure alignment with experts’ views was a key factor; GenAI information did not make a significant impact here. For others, exposure to conflicting AI responses more than doubled the ability to make an informed decision after exposure. This might mean that such exposure engaged participants’ deeper thinking processes: experience of uncertainty was a nudge for them to consider the issue. Asking an chatbot to provide conflicting responses may therefore be a useful strategy in decision-making.
