Psychological barrier slowing acceptance of independent AI technologies

A new study has found a striking divide in how people perceive artificial intelligence (AI), revealing that users overwhelmingly trust AI systems that assist humans rather than those that act independently. The findings highlight a growing psychological barrier to full AI autonomy, even as such systems become more embedded across healthcare, finance, education, and workplace decision-making.

Published in AI & Society, the study titled "Trust in AI: supportive vs. autonomous roles across four domains" presents a cross-domain analysis of how individuals respond to AI in different roles, offering fresh insight into the human factors shaping AI adoption.

Strong preference for human oversight across sectors

The research identifies a clear and consistent pattern: users feel significantly more comfortable with AI systems that function as supportive tools rather than autonomous decision-makers. This trust gap holds steady across all four domains examined, including healthcare, financial investment, workplace decisions, and education.

Participants reported much higher comfort levels when AI was positioned as an assistant working alongside humans. By contrast, trust dropped sharply when the same systems were described as operating independently without human involvement. The difference was not marginal but substantial, pointing to what researchers describe as a deep-rooted "trust asymmetry."

In healthcare, one of the most sensitive domains, the preference for supportive AI was particularly strong. Users showed a clear inclination toward systems that assist doctors rather than replace them entirely. Similar patterns emerged in finance, where AI-driven investment tools were more trusted when they complemented human judgment instead of making autonomous financial decisions.

The workplace domain also reflected this divide. AI tools that help managers or employees make decisions were viewed more favorably than systems that independently determine outcomes such as hiring or performance evaluations. In education, users preferred AI systems that support teachers and personalize learning rather than fully automated teaching systems.

This consistent pattern suggests that, despite rapid advances in AI capabilities, users remain hesitant to relinquish control in high-stakes or socially sensitive contexts. The presence of human oversight appears to serve as a psychological anchor, reinforcing trust even when AI systems are highly capable.

Experience and openness shape acceptance of autonomous AI

While the overall trust gap remained dominant, the study found that certain individual characteristics influenced how people perceived autonomous AI systems. In particular, prior experience with AI and a personal orientation toward openness to change emerged as the most significant predictors of trust in independent systems.

Users who had more frequent interactions with AI technologies, such as chatbots, voice assistants, or healthcare platforms, were more likely to express comfort with autonomous AI. This finding supports the idea that familiarity plays a crucial role in shaping trust. Repeated exposure to AI systems, even in low-risk settings, appears to reduce uncertainty and increase acceptance over time.

Openness to change, a value orientation linked to adaptability and willingness to embrace new experiences, also showed a meaningful association with trust in autonomous AI. Individuals who scored higher on this dimension were more receptive to AI systems operating without human intervention. This suggests that psychological predispositions, not just technical performance, influence how users respond to emerging technologies.

Age and education, by contrast, showed only minor effects. Older participants demonstrated slightly higher trust in both supportive and autonomous AI, but the impact was limited. Education level had virtually no influence on trust levels, indicating that familiarity and personal values outweigh formal knowledge in shaping attitudes toward AI.

Interestingly, personality traits such as neuroticism showed only weak associations, suggesting that broader psychological frameworks like values and experience are more relevant in understanding AI trust than individual personality differences alone.

Familiarity, not transparency, drives trust in AI systems

The study challenges traditional assumptions about how trust in AI is built. While much of the current discourse focuses on explainability and transparency, the findings suggest that familiarity and consistent performance may be far more important.

The research draws on the psychological concept of the mere-exposure effect, which posits that repeated exposure to a stimulus increases positive feelings toward it. Applied to AI, this means that users are more likely to trust systems they encounter regularly, even if they do not fully understand how those systems work.

This insight has significant implications for AI design. Rather than prioritizing complex technical explanations, developers may need to focus on creating systems that communicate in relatable, human-centered ways. AI systems that provide clear, context-driven explanations and demonstrate consistent behavior are more likely to build user confidence over time.

The study also introduces the concept of "relatable AI," emphasizing the importance of communication that aligns with human expectations. For example, an AI system that explains its recommendations in simple, context-aware language may be more trusted than one that offers detailed but technical justifications.

This approach reflects how trust is built in everyday technologies. Users do not need to understand the inner workings of smartphones or elevators to trust them; instead, trust emerges from repeated, reliable interactions. The same principle appears to apply to AI, where usability and consistency outweigh technical transparency.

Implications for AI deployment and policy

The strong preference for supportive AI suggests that hybrid human-AI systems may be the most effective path forward, particularly in high-risk environments such as healthcare and finance.

By maintaining human oversight, organizations can address user concerns while still leveraging the efficiency and analytical power of AI. This approach not only enhances trust but also aligns with ethical considerations around accountability and decision-making.

The role of familiarity highlights a potential strategy for increasing acceptance of autonomous AI. Introducing AI gradually through low-stakes applications, such as customer service or personal productivity tools, can help build user confidence. Over time, this exposure may reduce resistance to more advanced and autonomous systems.

However, the study also draws focus to the limits of current understanding. The research relied on self-reported comfort as a proxy for trust, focusing on emotional responses rather than actual behavior. This means that while users may express discomfort with autonomous AI, their real-world actions could differ in contexts where convenience or performance becomes more salient.

The study's sample, primarily composed of digitally active adults in Sweden, also raises questions about generalizability. Cultural factors, levels of technological exposure, and institutional trust may influence how users in other regions perceive AI. Future research will need to explore these dynamics across more diverse populations.