Requirements and Attitudes towards Explainable AI in Law Enforcement

In high-stakes areas such as law enforcement, where artificial intelligence has the potential to enhance effectiveness and inclusivity, its decisions must be both informed and accountable. Thus, designing explainable artificial intelligence (XAI) for such settings is a key social concern. Yet, explanations in practice are often overly technical or abstract. To address this, our study engaged with police employees in an EU country, who are users of a text classifier. We found that for them, usability and usefulness are paramount in explanation design, whereas interpretability and understandability are less emphasized. Drawing from these insights, we suggest design guidelines centred on clarity and relevance for domain experts. We contribute recommendations which guide XAI system designers to better cater to the specific needs of specialized users and promote the responsible use of AI tools in public service.