ISEDJ

Abstract: Simulation-based and gamified platforms have become central to cybersecurity education; however, numerous training environments remain procedurally static and offer limited real-time instructional scaffolding. This constraint is particularly significant in self-paced learning contexts, where novice learners require adaptive guidance to navigate decision-rich tasks such as phishing analysis, log interpretation, and incident response triage. This study proposes a feasibility framework for integrating AI-driven Non-Playable Characters (NPCs) as conversational guides within cybersecurity training simulations. The framework combines a curriculum-informed knowledge base aligned with established workforce and threat frameworks such as NIST NICE and MITRE ATT&CK, a large language model (LLM)-enabled conversational layer for real-time dialogue management, and a simulation interface deployable across browser-based and game-engine environments. To evaluate feasibility and instructional alignment, the study conducts a prompt-based comparative assessment of two NPC designs: a curriculum-informed agent grounded in structured course materials and a general-purpose LLM agent without domain-specific tuning. The results indicate that the curriculum-informed NPC generates more instructionally accurate, standards-aligned, and pedagogically structured responses across representative cybersecurity scenarios. This work offers a scalable design model and evaluation methodology that support future empirical research and the development of immersive, AI-guided cybersecurity training systems.

Download this article: ISEDJ - V24 N4 Page 36.pdf

Recommended Citation: Ihekweazu, C., Aghado, N., Ihekweazu, I., Adelowo, E., (2026). Toward Curriculum-Aligned Conversational NPCs for Cybersecurity Training Simulations: A Feasibility Framework and Prompt-Based Comparative Evaluation. Information Systems Education Journal 24(4) pp 36-53. https://doi.org/10.62273/KRUV1548