Using Anti-Personas to Model Children: How to Represent User-System Mismatches

User models, which represent the needs of users an online information access system is meant to serve, can facilitate the design and evaluation of the system. However, these models typically rely on assumptions drawn from adult users, such as advanced literacy levels or the ability to evaluate retrieved content critically. Hence, they often overlook the distinct needs and search habits of children, despite the growing prevalence of information access systems among youngsters. In this work, we introduce a set of child anti-personas, user models representing classes of users for whom the system is not designed. To substantiate these mismatches, we investigate how primary school children's behaviors when interacting with different search tools diverge from the assumptions of typical user models. Thereafter, we identify recurring patterns of misalignment between children's behavior and system expectations and link these patterns to the risks identified by adult stakeholders involved in children's online searching. By representing children as anti-personas of contemporary search tools, this work contributes to the quest for more informed and less stereotypical user modeling when it comes to information access for underrepresented groups, such as children.