Human-Centered Explainable AI for Cybersecurity: A Design Science Research Approach

AI systems are becoming increasingly common in cybersecurity threat detection; however, their unexplained decision-making processes lead to diminished user trust and slow adoption rates. The research addresses this obstacle by developing a Human-Centered Explainable AI (HC-XAI) dashboard that enhances transparency and user trust in AI threat detection classifications. Grounded in Design Science Research (DSR), the artifact integrates three explanation modalities: explanations in the dashboard are derived from rule-based logic, combined with natural language explanations generated by large language models, and visual heatmaps that display token-level attention. Justificatory knowledge from cognitive psychology and human-AI interaction, along with explanation theory, provides the basis for the design process. The artifact was evaluated through a two-phase strategy. The artifact evaluation consisted of an expert walkthrough using HC-XAI heuristics and a mixed-methods user study with 23 cybersecurity students. The findings demonstrate excellent usability levels and perceived trustworthiness, while also showing varied preferences depending on the type of explanation. The study presents a multimodal XAI artifact, along with validated principles for HC-XAI applications in critical systems, and investigates user preferences regarding explanation clarity and cognitive load. The study yields new insights into the implementation of human-aligned XAI within practical cybersecurity environments.