The Two Roads to AI Dependence in Students

A new peer-reviewed study maps two distinct psychological roads that lead students toward AI dependence — and the findings turn the usual addiction story on its head. The research, led by Feng Sun of Yangzhou University and Renbiao Ma of Changzhou Senior High School of Jiangsu Province, surveyed 1,029 high school and university students in China, testing seven psychological variables with structural equation modeling, according to the peer-reviewed paper published in October 2026. Borrowing the I-PACE framework — a model originally built to explain gambling and internet-use disorders — the team found that dependence on generative AI arrives through two very different doors: being mesmerized by the tool, or coolly treating it as the smartest shortcut. Published in the International Journal of Mental Health and Addiction, the study extends a classic addiction model into unfamiliar territory, as reported by ScienMag.

Rather than asking how often students use generative AI, the researchers mapped why some tip toward AI dependence. The framework traces how a person's traits, moods, and thinking interact with behavior over time — and it has mostly been applied to pursuits built on pleasure, like gaming or social media. Generative AI is different: students value it for performance, not kicks, which forced the researchers to ask whether the old addiction playbook still fits. Their answer: the playbook mostly holds, but the entry points look different. And notably, the students most at risk were not simply the heaviest users.

Absorption: the mesmerizing road to AI dependence

Cognitive absorption — the experience of being so pulled into the flow of a tool that time and surroundings fall away — emerged as the strongest direct predictor of AI dependence. Here the study makes its boldest move: instead of treating absorption as a symptom of dependence, the researchers reframe it as an antecedent, a precursor that comes before dependence sets in. Students entranced by the conversational rhythm of a chatbot are the most likely to drift toward AI dependence, long before their usage looks extreme. That finding lands as a warning for anyone tracking hours logged rather than minds hooked: counting screen time would miss the students quietly slipping. Deliberate pauses and reflective self-monitoring — noticing when a session is pulling you under — could interrupt the drift before it hardens into habit.

The shortcut route: AI dependence by cold calculation

The second road is less cinematic and more deliberate. Students who come to see generative AI as an efficient academic shortcut gradually stop attempting tasks without it — a slow slide driven by perceived usefulness rather than fascination. Where the mesmerized student drifts in sideways, this one walks in with eyes open, choosing the tool because it saves time and mental effort. Positive affect and emotion regulation feed dependence only indirectly, through this very belief in the tool's usefulness, in a link the researchers connect to broaden-and-build theory: pleasant moods widen what we consider doing, and here that widened repertoire gets channeled into ever-warmer appraisals of the chatbot. Even need for cognition, self-efficacy, and avoidance learning motivation matter only indirectly — students who genuinely enjoy hard thinking can still be carried into AI dependence through these nearer mechanisms, as reported by ScienMag.

Why bans miss the mark — and what actually helps

For students and teachers, the takeaway is practical rather than moral. Blanket bans on AI tools miss the mark because the two pathways respond to different medicine: the mesmerized student needs prompts to pause and reflect mid-session, while the shortcut-seeker needs assignments redesigned so the easy way stops looking like the only way. The researchers argue that interventions on absorption — deliberate pauses, reflective self-monitoring — can prevent AI dependence before it forms, and that policymakers need psychological literacy alongside technical literacy when they write the rules. It is also a note of reassurance for learners who love deep thinking: enjoying hard problems neither immunizes you nor condemns you — what matters is noticing which road you are on. Similar questions about young minds and machines came up in our earlier reporting on middle schoolers whose AI friendships left them lonelier.

The deeper significance is what the study does to the addiction map itself. Classic models of technology dependence were built on hedonic gratification — the rush of a win, a like, a view. Generative AI breaks that mold: nobody gets a jackpot of pleasure from a well-formatted essay outline. Dependence here can begin with something as unglamorous as a rational desire to work more efficiently. Extending a framework forged in gambling and gaming into productivity software valued for performance, not pleasure, suggests the next wave of problematic tech use may not look like addiction at all — just smart, tired people choosing the fastest tool, over and over, until the choice disappears.

A few caveats keep the findings in perspective. The data are cross-sectional and drawn from a single country, so they show association rather than proven causation — the arrows could point both ways, and the pattern may differ elsewhere. It is also one study: replication in other cultures and age groups will decide how far the two-roads model travels. The research was funded by the National Natural Science Foundation of China, with ethics approval from the university's medical college; for participants under eighteen, the schools involved approved participation and notified parents.

Still, the message lands: AI dependence is not one story but two — enchantment and efficiency — and each calls for its own response. You can follow this beat on our AI News topic page.