Turkish researchers develop scale to measure students' AI confidence

Somewhere between May and November of last year, 284 university students across Turkey sat down at their computers and answered questions about how confident they felt using artificial intelligence. They were part of something quieter than it sounds: the validation of a new tool designed to measure what researchers call AI self-efficacy—not whether students could code or debug, but whether they felt capable and comfortable when they encountered AI systems in their daily academic lives.

The work happened at Koç University's School of Nursing, where Associate Professor Remziye Semerci Şahin and Assistant Professor Seda Güney took an existing English-language scale and adapted it for Turkish speakers. The Artificial Intelligence Self-Efficacy Scale, or AISES, had been developed elsewhere, but Turkey had no validated version. The researchers translated it carefully, checked the translation against the original, gathered expert feedback, ran a pilot study, and then tested it on their sample of students from public and private universities scattered across different regions of the country. The results, published in the International Journal of Human–Computer Interaction, showed something solid: a measurement tool that actually works.

What makes this work matter is the context. Universities are already weaving AI into their classrooms—adaptive learning platforms that adjust to individual students, tutoring systems that respond in real time, automated feedback tools that grade and comment on work. The technology is there. But institutions have been slower to ask a basic question: How do students actually feel about all this? Not whether they think AI is important—most do—but whether they believe they can navigate it, understand it, work alongside it without anxiety or confusion.

The scale that emerged from this study measures AI self-efficacy across four distinct dimensions. The first is assistance—students' sense that they can ask AI systems for help and get useful responses. The second is anthropomorphic interaction, which captures how comfortable students feel treating AI almost like a conversation partner, attributing intention or understanding to it. The third is comfort with AI itself, the emotional ease someone experiences when the technology is present. The fourth is technological skills, the practical know-how to operate these systems. Together, these four factors explained nearly 74 percent of the variation in how students reported their AI confidence. The scale's internal consistency was excellent—a Cronbach's alpha of 0.937, which in research terms means the questions reliably measured the same underlying thing.

The portrait of Turkish university students that emerges from the data is instructive. Most had already used some form of artificial intelligence. But very few had taken a formal class or online course specifically about AI. Yet the majority wanted to learn. About half rated their own AI knowledge as moderate. And a substantial portion saw AI as genuinely useful in their lives—not as a distant technology but as something practical and present.

What the researchers emphasize, though, is that AI self-efficacy is not purely technical. A student might understand how to prompt a language model or navigate an interface, but still feel uneasy, still doubt themselves, still experience anxiety when the system behaves unexpectedly. The emotional and psychological dimensions matter as much as the mechanical ones. This is why the scale includes questions about comfort and anthropomorphic interaction alongside questions about skills.

The implications point toward a gap in how universities are currently preparing students. Many institutions have added AI tools to their learning environments without necessarily building the confidence and literacy that would help students use those tools well. The researchers suggest that structured AI literacy programs—courses or modules that address both the technical and the emotional dimensions of working with these systems—could make a real difference. The Turkish AISES, now validated and reliable, gives educators a way to measure whether such programs are actually working, whether students' confidence is growing, whether interventions are having the intended effect.

What happens next depends on whether universities take the hint. The tool exists. The need is clear. The question is whether institutions will invest in the kind of deliberate, sustained effort required to help students move from passive exposure to genuine confidence.