Non-expert understanding of AI is often shrouded in ambiguity, with the term representing a broad spectrum of technologies. This lack of clarity leads to divergent perceptions, ranging from excessive optimism to undue pessimism. Policymakers frequently emphasize AI as a key future-oriented technology, which combined with simplistic descriptors such as ‘intelligent’ or ‘smart’ reinforces a superficial non-expert understanding. Introducing technical topics, like AI, to non-experts often suffers from limited conceptual clarity, with AI also frequently described as an ”opaque technology” and a ”fuzzy concept”. This contributes to conceptual borrowing, which in the case of AI has fostered an anthropomorphized vocabulary.
Research on non-experts’ perceptions and misconceptions of AI is still in its early stages. This also holds for teachers and children, highlighting a gap in the foundations for designing AI-related education: how can we design curricula and educational initiatives, if we do not know what concepts to cover?
Concept inventories (CI) are used to assess understanding of core concepts in a subject area. CIs build on multiple-choice questions with one correct answer and distractors based on common misconceptions and can help in identifying knowledge gaps and mental models. Therefore, the current project will develop such an AI CI. We will systematically identify key AI concepts for non-experts and explore how these concepts are understood – or misunderstood – by the general public. This effort will contribute to AI education by identifying core concepts, uncovering common misconceptions, and creating a validated assessment tool that can inform teaching, policy, and public engagement.