Why Confident Digital Workers Ask AI for Feedback—and When They Hold Back

When organizations flood the workplace with artificial intelligence, something curious happens on the human side of the screen: some employees begin actively asking AI systems for feedback on their performance, while others with access to the very same tools never think to try. A new study published in Current Psychology by Fang Sun of Hubei University of Economics and colleagues Mengyan Liu and Shao-Long Li of Wuhan University offers one of the first systematic explanations for this divergence, and its central finding is deceptively simple. Workers do not treat AI as a passive feedback dispenser that simply appears when digitalization arrives. Instead, they treat it as a resource they must feel competent enough to approach, and the strength of that confidence depends heavily on how cognitively demanding their jobs are.

The research team coined the term feedback-seeking from AI to describe a behavior that earlier studies had largely overlooked. Since Susan Ashford and L. L. Cummings first framed feedback as an individual resource that employees can strategically create for themselves in 1983, the feedback-seeking literature has focused almost exclusively on human sources: supervisors, peers, and mentors. Employees weigh the costs of asking—looking incompetent, bothering a busy colleague—against the value of the information they might gain. But as organizations increasingly deploy AI to evaluate and coach workers, from algorithmic performance dashboards to intelligent assistants that critique sales calls, the classic model of feedback-seeking faces a new kind of feedback source that never gets tired, never judges, and is available at any hour.

Prior work on AI-generated feedback, the authors note, has predominantly portrayed employees as passive recipients of whatever the algorithm delivers. Studies have examined how workers react to AI feedback once it arrives, whether they trust it, and whether disclosing that it came from a machine helps or hurts performance. What has been missing is the proactive half of the equation: the decision to seek out AI feedback in the first place. Sun and colleagues argue that this omission matters because feedback-seeking is fundamentally a self-regulatory behavior. Just as employees decide whether to approach a manager for advice, they now decide whether to query an intelligent assistant about their role expectations, their technical performance, or even how well their behavior aligns with company values.

To explain what drives that decision, the researchers turned to Albert Bandura’s self-efficacy theory, the foundational idea that people’s beliefs about their own capabilities shape whether they attempt, persist at, and succeed in particular behaviors. Bandura’s framework, first laid out in Psychological Review in 1977, has long predicted that confidence is not just a feeling but a causal engine of action. The authors adapted this logic to the digital context, proposing digital self-efficacy—defined as employees’ beliefs in their own knowledge, skills, capabilities, expertise, and experience with digital technologies—as the psychological bridge between a digitalized workplace and the act of asking AI for feedback. Perceived workplace digitalization, in their conceptualization, captures employees’ sense that digital technologies are useful, easy to use, socially endorsed, and supported by management with adequate resources.

The empirical test took the form of a three-wave survey administered to full-time employees at Chinese enterprises, yielding a final sample of 292 participants. Separating the measurements in time is a methodological strength, because it reduces the risk that the relationships simply reflect how people answered questions at a single moment. The study was approved by the Institutional Review Board of Wuhan University, participants gave informed consent, and the datasets were made publicly available on the Open Science Framework. The measurement scales were adapted from established instruments: perceived workplace digitalization drew on Venkatesh and colleagues’ unified technology acceptance model, digital self-efficacy on Shao and colleagues’ 2022 scale, feedback-seeking behavior on VandeWalle and colleagues’ integrated model, and problem-solving demands on the Work Design Questionnaire developed by Frederick Morgeson and Stephen Humphrey in 2006.

The results confirmed the core hypothesis. Perceived workplace digitalization positively influenced employees’ feedback-seeking from AI, and this effect operated through digital self-efficacy: the more employees believed their workplace was digitalized, the more confident they felt in their digital capabilities, and the more actively they sought feedback from intelligent assistants and other AI tools. That mediating pathway is the study’s theoretical heart. It suggests that digitalization does not mechanically produce new behaviors; it works through a psychological transformation in which employees come to see themselves as competent users of the technology. An organization can install the most sophisticated AI feedback systems available, but if workers do not feel capable of engaging with them, the systems will sit largely unused as feedback sources.

The second major finding concerns the boundary condition. Task problem-solving demands—the extent to which a job requires solving problems with no obvious correct answer, dealing with unfamiliar situations, and generating creative or unique solutions—strengthened the relationship between digital self-efficacy and feedback-seeking from AI. In other words, digital confidence translated most powerfully into AI feedback-seeking when the work itself was cognitively demanding. The mediating effect of digital self-efficacy was particularly pronounced for employees in high problem-solving roles. This pattern makes intuitive sense once framed in self-regulatory terms. When tasks are routine, employees may not feel much need for additional feedback, and whatever confidence they have remains dormant. When tasks are ambiguous and novel, the information gap widens, and confident digital users appear to recognize AI as a readily available resource for closing it—asking about technical aspects of their performance, clarifying role expectations, and even probing how their workplace behavior fits organizational values.

The findings carry practical weight for managers navigating what industry surveys from Oracle and Deloitte have described as a rapid infusion of AI into human resources and everyday work. The study suggests that simply deploying AI tools is insufficient; organizations must simultaneously cultivate employees’ digital self-efficacy through training, resources, and visible management support, because self-efficacy is the mechanism that converts digital infrastructure into proactive behavior. Moreover, the moderating role of problem-solving demands implies that digital initiatives should be aligned with task characteristics. Roles heavy on creative problem-solving stand to benefit most from employees who feel digitally capable and actively mine AI for feedback, whereas for routine roles the same investment may yield weaker returns. The authors frame this as the importance of aligning task characteristics with digital initiatives to foster effective human-AI interaction in contemporary workplaces.

The study also contributes to a broader scholarly conversation about how humans and machines negotiate the exchange of evaluative information. Recent research has documented the double-edged nature of AI feedback: a 2021 study in Strategic Management Journal by Tong and colleagues found that the deployment versus disclosure of AI feedback had different effects on employee performance, while work published in Human Resource Management in 2024 explored how AI-delivered negative feedback can be leveraged to enhance job performance. Other studies have examined trust in AI, perceived fairness of algorithmic versus human managers, and even workers’ spiteful reactions to robot supervisors that seem too human. By shifting the analytical lens from reception to initiation, Sun and colleagues extend the feedback-seeking literature’s scope and open a new question: not just how employees respond to machine-generated evaluations, but when and why they go looking for them.

Limitations remain, as with any survey-based design. The sample of 292 full-time employees in Chinese enterprises may not generalize directly to other labor markets, and three-wave correlational data, while stronger than a single snapshot, cannot fully rule out alternative causal explanations. The authors controlled for relevant variables in line with current methodological recommendations for correlational organizational research, and they conducted statistical power analyses to justify their sample size, but experimental and cross-cultural replication would strengthen the causal claims. Still, the core message stands out clearly. As AI becomes an everyday presence in the workplace, the employees who benefit most may not be those with the best tools, but those who feel capable of using them—and whose demanding, creative, problem-saturated jobs give them the strongest reason to ask the machine for advice.

Subject of Research: How perceived workplace digitalization drives employees' proactive feedback-seeking from AI through digital self-efficacy, moderated by task problem-solving demands

Article Title: Perceived workplace digitalization and employees’ feedback-seeking from artificial intelligence: the roles of digital self-efficacy and problem-solving demands

Article References: Sun, F., Liu, M., & Li, S.-L. (2026). Perceived workplace digitalization and employees’ feedback-seeking from artificial intelligence: the roles of digital self-efficacy and problem-solving demands. Current Psychology, 45(17), Article 1429. https://doi.org/10.1007/s12144-026-10001-2

Image Credits: AI Generated

DOI: 10.1007/s12144-026-10001-2

Keywords: feedback-seeking from AI, digital self-efficacy, workplace digitalization, problem-solving demands, self-efficacy theory, artificial intelligence, organizational behavior, employee proactivity, human-AI interaction, Current Psychology, job design, China survey