-
Title
-
Developing AI-Security Self-Efficacy Through Prompt Injection Research in a High School Classroom
-
Author
-
Thomas Heverin
-
Year Published
-
2026
-
Description
-
In high school cybersecurity classrooms, girls often experience a confidence gap when confronting the unpredictable and ill-defined vulnerabilities of modern artificial intelligence systems. This action research project examined how shifting students from passive users of AI tools to adversarial investigators through prompt injection testing influenced the AI-security self-efficacy of 12 girls enrolled in a high school Cybersecurity and Ethical Hacking class at The Baldwin School. Pre- and post-action surveys, student reflections, interviews, and work artifacts provided a comprehensive dataset capturing students’ transition from technical uncertainty to investigative authority. Findings indicate that self-efficacy increased substantially when girls engaged in mastery-based experiences that positioned them as active security researchers. Through iterative experimentation and creative prompt design, students successfully bypassed AI safeguards and demonstrated significant gains in confidence in their ability to analyze, question, and test AI systems. The findings also suggest that hands-on exploration of AI vulnerabilities promotes systems-level thinking, critical inquiry, and ethical awareness. While discovering the fragility of AI guardrails initially produced skepticism about the reliability of these technologies, this realization ultimately strengthened students’ sense of responsibility and agency in evaluating emerging AI systems. Future research should examine how adversarial exploration of AI technologies influences girls’ long-term persistence in cybersecurity pathways and how investigative learning models can support AI literacy and confidence among girls in secondary education. Developing AI-Security Self-Efficacy Through Prompt Injection Research in a High School Classroom To prepare students to navigate and lead in a digital landscape defined by rapid technological disruption, they must be empowered to see themselves as sophisticated agents of change rather than passive consumers of technology. However, girls in advanced technical domains often face a significant confidence gap where their belief in their own capabilities, rather than their actual abilities, serves as the primary barrier to participation (Francis et al., 2024). This is particularly visible in high-stakes fields like cybersecurity, where technical uncertainty can lead to a hesitation or worry about learning. The development of self-efficacy, as a core component of agency, is therefore essential to enable girls to boldly thrive and assert their authority in technical domains that have traditionally felt exclusionary, including cybersecurity and artificial intelligence (AI). Furthermore, according to Chiu et al. (2025), understanding how AI works represents a key step to living a safe and healthy life in a society dominated by AI.
-
Tags
-
Technology, Teaching and Learning