Learning Through Interaction: Designing Adaptive AI to Enhance AI and Information Literacy

Abstract

As Large Language Models (LLMs) increasingly mediate how users access, interpret, and generate information, interaction with these systems requires users to reason about both the information produced and the AI system itself. It remains underexplored how direct interaction with LLMs, rather than explicit instruction, shapes users’ AI literacy and information literacy. To address this gap, this project focuses on the design of an interactive system that articulates how adaptive AI interaction can function as a mechanism for experiential learning of AI literacy and information literacy. Through adaptive interaction, such as guided prompting, the system aims to support users with varying levels of prior AI knowledge in enhancing their understanding and critical evaluation of AI systems (AI literacy) and critical evaluation of AI-generated information (information literacy). Using a research-through-design approach, we explore how specific interaction features support the joint development of these literacies during use. We propose a two-stage adaptive mechanism that first triggers information-level evaluation and then provides system-level explanations, while adapting the level of feedback to users’ prior knowledge. Our work proposes that adaptive AI can be positioned not only as a tool for task completion, but also as an environment that facilitates experiential learning embedded within interaction itself.