New Scale Reveals How AI Dependence May Quietly Erode Human Autonomy

Every time you ask a chatbot what to cook, which job to apply for, or how to reply to a difficult email, a small transaction takes place: you hand over a piece of your decision-making. For most people these trades feel trivial, even liberating. But a new study published in Current Psychology suggests that the cumulative effect of outsourcing choices to artificial intelligence may be measurable — and that the people who feel their autonomy slipping away also report lower psychological well-being, weaker authenticity, and diminished self-esteem. The research, conducted by Mustafa Selim Altınışık of Ondokuz Mayıs University in Türkiye, introduces one of the first validated instruments designed specifically to capture what the author calls perceived AI-related autonomy erosion, a psychological construct that may become increasingly relevant as generative AI embeds itself into everyday cognition.

The study, published in September 2026 in the Springer journal Current Psychology, set out to answer a deceptively simple question: can the subjective sense that AI is eroding one’s autonomy be measured reliably? To do so, Altınışık employed a rigorous two-stage quantitative design. The first stage was devoted to scale development, following the classical psychometric playbook. A pilot study with 65 participants allowed the initial item pool to be refined, after which an exploratory factor analysis was conducted on a sample of 256 adults. This statistical technique examines the patterns of correlation among questionnaire items to discover whether they cluster around underlying dimensions. A subsequent confirmatory factor analysis on a separate sample of 312 participants then tested whether the emerging structure held up when confronted with fresh data — a crucial guardrail against findings that merely reflect the quirks of one particular dataset.

The result is a compact 12-item instrument with a three-factor structure, and the three factors are strikingly intuitive. The first, Reduced Decision Ownership, captures the feeling that one’s choices no longer genuinely belong to oneself — that recommendations, rankings, and algorithmic nudges have quietly taken over the steering wheel. The second, Weakened Autonomous Judgment, reflects a perceived decline in the confidence and capacity to evaluate options independently, a kind of subjective atrophy of one’s own critical faculties. The third, Decision Delegation, measures the behavioral side of the phenomenon: the extent to which a person habitually hands decisions over to AI systems rather than deliberating personally. Together, these three dimensions sketch a portrait of autonomy under pressure that is both psychological and behavioral, capturing not just what people do with AI but how they feel about what they have become.

The second stage of the research moved from measurement to modeling. Using an independent sample of 473 adults, Altınışık tested a structural model linking dependence on artificial intelligence — a construct already measured by existing instruments — to three well-established psychological outcomes: psychological well-being, authenticity, and state self-esteem. Psychological well-being, in the tradition of Carol Ryff’s foundational work, encompasses dimensions such as purpose in life, personal growth, and self-acceptance. Authenticity, as conceptualized by researchers like Michael Kernis and Brian Goldman, refers to the congruence between one’s true self and one’s actual behavior and awareness. State self-esteem, measured with the scale developed by Todd Heatherton and Janet Polivy, captures moment-to-moment fluctuations in how people evaluate themselves, as opposed to their stable baseline self-regard.

The correlational findings were consistent and, in one sense, sobering. Perceived AI-related autonomy erosion was negatively associated with all three outcomes: adults who reported greater erosion of autonomy also reported lower psychological well-being, lower authenticity, and lower state self-esteem. In other words, the subjective sense that AI has colonized one’s decision-making travels together with a cluster of indicators that psychologists have long linked to flourishing. The study is careful to frame these as cross-sectional associations — snapshots taken at a single point in time — which means the direction of causality remains an open question. It is plausible that people who already feel less autonomous gravitate more heavily toward AI assistance; it is equally plausible that heavy reliance on AI gradually hollows out the felt experience of self-direction. Most likely, the relationship runs in both directions, forming a feedback loop that future longitudinal research will need to untangle.

The mediation analyses added a further layer of theoretical interest. In statistical mediation, a researcher tests whether an observed relationship between two variables flows through a third variable — whether the third variable acts as a conduit or mechanism. Here, the model proposed that dependence on AI relates to lower well-being, authenticity, and self-esteem partly because it fosters the perception that one’s autonomy has eroded. The results supported this: significant indirect associations emerged between AI dependence and each of the three psychological outcomes when perceived autonomy erosion was included as a statistical mediator. Read plainly, the findings suggest that the psychological cost of AI dependence, if there is one, may be routed through the experience of lost self-direction. It is not merely that people use AI a great deal; it is that heavy use appears to coincide with the feeling that their choices are no longer their own, and that feeling is what tracks with diminished psychological functioning.

These results land at the intersection of several rapidly growing literatures. Research on problematic technology use has documented links between smartphone and internet dependence and anxiety, depression, and reduced well-being, and recent work has begun to extend this framework to AI specifically, with emerging scales for AI addiction and problematic generative AI use. Parallel research in human-computer interaction has long studied automation bias — the well-documented tendency of humans to over-trust automated systems and defer to their outputs, sometimes against their own better judgment. The new study contributes a distinct angle: rather than focusing on addiction or on trust in machines, it centers the user’s own sense of agency, drawing on the philosophical and psychological tradition of autonomy that runs from self-determination theory through decades of work on choice, will, and the inner self. Self-determination theory holds that autonomy is one of three basic psychological needs, alongside competence and relatedness, whose satisfaction is essential for well-being. A tool that systematically displaces autonomous action therefore strikes at something motivationally fundamental.

The practical implications extend well beyond the laboratory. Educators wrestling with students who outsource essays and problem sets to chatbots may find in this research a vocabulary for what is at stake: not just academic integrity but the developmental work of building autonomous judgment. Clinicians and counselors may eventually use the scale to identify clients whose AI reliance has become entangled with low self-esteem or a diminished sense of authenticity. Designers of AI systems, meanwhile, face a pointed challenge: products engineered for maximum convenience may, as a side effect, encourage decision delegation that users themselves come to experience as a loss of ownership. The study’s author notes that the scale provides initial psychometric evidence only, and that further validation is needed before it can be deployed broadly — an appropriately cautious caveat for a first-of-its-kind instrument. The data underlying the study are not publicly available due to privacy and ethical restrictions but can be requested from the corresponding author, and the research protocol received approval from the Ondokuz Mayıs University Social and Human Sciences Ethics Committee in accordance with the Declaration of Helsinki.

What makes this study resonate beyond its statistics is the mirror it holds up to a cultural moment. Generative AI has been adopted faster than almost any technology in history, and the conversation has largely revolved around what these systems can do for us. This research reframes the question around what they may be doing to us — or more precisely, to how we experience ourselves as choosers. The three factors of the new scale read like a checklist for modern life: Do my decisions still feel like mine? Can I still trust my own judgment? Am I delegating more than I intend? None of these questions has a definitive answer yet, and the cross-sectional nature of the evidence forbids dramatic causal claims. But the study offers something the AI debate has lacked: a way to quantify the subjective texture of human-AI co-dependence, and early evidence that this texture matters for well-being, authenticity, and self-esteem. As AI assistants grow more capable and more pervasive, the erosion of autonomy may prove to be one of the defining psychological questions of the decade — and researchers now have a ruler to measure it.

Subject of Research: Development of a scale measuring perceived AI-related autonomy erosion and its associations with well-being, authenticity, and self-esteem in adults

Article Title: Perceived AI-related autonomy erosion in adults: scale development and associations with well-being, authenticity, and self-esteem

Article References: ALTINIŞIK, M. S. (2026). Perceived AI-related autonomy erosion in adults: scale development and associations with well-being, authenticity, and self-esteem. Current Psychology, 45(17), Article 1492. https://doi.org/10.1007/s12144-026-09999-2

Image Credits: AI Generated

DOI: 10.1007/s12144-026-09999-2

Keywords: artificial intelligence, autonomy, AI dependence, psychological well-being, authenticity, self-esteem, scale development, psychometrics, decision-making, self-determination theory, mental health, generative AI