Why not knowing how the AI decides makes things worse
A monitoring system that only counts keystrokes or tracks location is already stressful, but an AI system that also judges performance adds a second layer: employees often cannot tell what the system is measuring or why it reached its conclusion. One review of AI-enabled surveillance describes this opaque nature—employees being unable to grasp how AI decisions are made—as a core feature of the technology, alongside surveillance intensity, performance anxiety, and stress [1].
The anxiety that results has been given a name in research on gig workers: "algorithmic anxiety," described as arising from continuous digital surveillance combined with opaque decision-making that shapes both working conditions and worker well-being [2]. The important point is that the uncertainty is not a side issue. When the rules are invisible, workers cannot reliably predict how to succeed, so the monitoring never fully switches off in their minds.
How a watched feeling turns into lasting strain
The largest and most direct test of this comes from a national survey of 3,508 Canadian workers, which asked people across all kinds of jobs how much surveillance they perceived at work. Using structural equation modeling—a statistical method for tracing how one factor influences another through intermediate steps—the researchers found that surveillance perceptions were indirectly associated with more psychological distress and lower job satisfaction [5].
The pathway matters more than the headline. Surveillance was linked to three secondary work stressors: greater job pressure, reduced autonomy, and privacy violations. For psychological distress, these three stressors fully accounted for the link—meaning the distress was not simply caused by being watched, but by the pressure, loss of control, and intrusion that watching produced [5]. That is a useful distinction for anyone trying to reduce harm: the damage runs through the working conditions surveillance creates.
Can better oversight protect employees' sense of agency?
If lost autonomy is part of the mechanism, then governance that preserves autonomy should help. Research on AI-enabled human resources systems examines three governance mechanisms—algorithmic transparency, fairness, and human oversight—and asks whether they determine if AI supports or undermines employee agency, trust, and the stress caused by AI [3]. This points to a practical lever: the same AI tool may feel very different depending on whether people can see how it works and whether a human is involved in decisions that affect them.
The AI observer-effect research reaches a compatible conclusion from a different angle, linking intensive, opaque monitoring to burnout and anxiety and arguing that employees should be involved in decisions about how AI monitoring systems are introduced and used [4]. Notably, that recommendation is about process, not just technology—giving workers a voice in the design and rollout is itself a way of restoring the agency that monitoring erodes.
What we still cannot say for certain
The evidence is suggestive rather than definitive on several points. The Canadian survey measured workers' general perception of surveillance rather than AI-specific systems, so it speaks to surveillance broadly, not to opaque algorithms in particular [5]. The AI-focused studies describe links between opaque monitoring and anxiety or burnout, but they are reviews and conceptual analyses rather than controlled tests of cause and effect [1][4]. And the gig-work research on algorithmic anxiety is specific to gig workers, whose conditions may not transfer directly to conventional employees [2].
What the studies do agree on is the direction of the effect and the likely route: opacity and constant watching reduce control, raise job pressure, and intrude on privacy, and those conditions are what appear to translate into anxiety and chronic strain [5][4]. The open question is how much transparency and human oversight can offset that—a question the governance research frames but does not yet answer with outcome data [3].
Sources used in this answer
AI-Driven Threat Intelligence, Real-Time Monitoring, and the Future of Employee Performance
AI-enabled surveillance is characterized by opacity—employees cannot grasp how AI decisions are made—and is discussed alongside surveillance intensity, performance anxiety, and stress.
The Algorithmic Anxiety Trap: Why AI Policy Must Move Beyond Post Work Welfare For India's Gig Workers
Continuous digital surveillance combined with opaque decision-making is described as producing "algorithmic anxiety" among gig workers.
Governing algorithms, empowering people: how ethical AI oversight shapes trust, technostress, and employee autonomy in AI-enabled HRM
Algorithmic transparency, fairness, and human oversight are examined as governance mechanisms that may determine whether AI-enabled HRM supports or undermines employee agency, trust, and AI-related stress.
When the Machine is Watching: Making Sense of the AI Observer Effect at Work
Intensive, opaque AI monitoring is linked to burnout and anxiety, and employees should be involved in decisions about how such systems are introduced and used.
Private Eyes, They See Your Every Move: Workplace Surveillance and Worker Well-Being
In a national sample of 3,508 Canadian workers, perceived workplace surveillance was indirectly associated with greater psychological distress and lower job satisfaction through job pressure, reduced autonomy, and privacy violations.