The next speakers at the SEASON 2026 conference in Hamburg are Joachim Griesbaum, Karen Malin Krüger, and Anna Mierzecka, whose interest is in information-literate human behaviour in AI interactions. Information literacy is defined as an individual’s ability to deal with information effectively and ethically; AI makes such engagement easier, but also partially takes over the process of selecting, evaluating, and presenting information – assessing the quality of that information, however, becomes a great deal more complicated.
By now, several AI literacy frameworks have already been proposed; how these relate to the core elements of information literacy is still poorly understood, however. How might we assess whether human interactions with AI are appropriately information-literate, then?
Such an assessment requires instruments that assess the actual state of information literacy in the interactions between humans and AI, and measure whether information literacy interventions lead to notable improvements in information-literate behaviours. This project proposes one such instrument; it seeks to capture human interactions with AI, and the roles that AI plays in these interactions (including as provider of information but perhaps also as trainer in information literacy).
This builds on past studies that adapt classical search tactics for more effective AI interactions; propose multidimensional query taxonomies that evaluate information-literate behaviour; correlate human-AI interaction capabilities with critical thinking; and distinguish surface information evaluation approaches from critical information competence.
These all transfer conventional information-literate behaviour models to AI interactions, but there is more yet to be done here; in particular, the contributions of the AI systems also need to be studied from an information literacy angle, including indications about the AI’s certainty, invitations to verify its information, and encouragement to engage critically with responses.
The present study therefore focussed on prompt-and-response interaction sequences in engaging with AI chatbots; this examines both the prompting strategies and response features and quality, and the course of interactions over multiple turns. AI here becomes a co-actor in information-literate behaviour, rather than just an information source.
Variables that can be used to assess user approaches here include query design features, output control features, prompting strategies, epistemic trust in AI, user-critical evaluation, and prompt chain logic; while AI responses can be evaluated for their explanation and transparency features, mistakes and hallucinations, and triggers for user follow-ups. These can all be broken down into further subcategories, and AI tools can also be used to help with the analysis of such variables, building on a detailed codebook designed by the project, since the coding is very time-consuming.
This needs to capture user behaviour, AI responses, and the discourse between them, then – that cross-actor and turn-taking analysis is novel about this approach. Human coder agreement in the deployment of this coding approach needs ton be improved still, and the AIs can also be trained to extend this coding approach to larger datasets. This all needs to be done in the context of the specific information seeking interaction, of course; different strategies may be more appropriate for different contexts.
