EyeWarn focus group: Science and ergonomics

At the recent EyeWarn stakeholder event, participants were split into three focus groups to discuss a scenario set in a near-future workplace, where worker wellbeing and productivity are supported through advanced biometric technologies.

This focus group consisted of those working in roles focusing on science and ergonomics.

Scenario

A mid-sized logistics company has implemented an eye-tracking fatigue detection system across its workforce, including office staff, warehouse operatives, and long-haul drivers. The company reports improvements in safety metrics and reduced accident rates. However, employee reactions are mixed. Some workers appreciate the proactive support for wellbeing, while others express concerns about surveillance, data use, and potential disciplinary consequences. Trade unions and regulators are beginning to take interest, and the company is considering scaling the system across multiple regions. The company needs expert perspectives to evaluate the broader implications of this technology.

Discussions

This focus group explored the use of eye tracking and social gaze as indicators of mental state in workplace settings, particularly for fatigue detection. Participants were introduced to a prototype system using camera-based gaze estimation and a data collection application capturing visual and sensor data.

Due to the absence of audio recording, participants contributed via sticky notes arranged on a shared whiteboard. This produced a non-linear, collaborative mapping of ideas, which broadly divided into “Risks” and “Benefits.” However, discussions frequently crossed this boundary, suggesting that acceptability depends less on sensing itself and more on how the system is governed, interpreted, and applied.

A central theme was that fatigue detection technologies cannot be evaluated in isolation. Their impact is shaped by workplace context, scientific validity, and power relations, raising questions about whether they support wellbeing or reinforce monitoring.

Validity and reliability of eye-tracking measures

Participants questioned whether eye-tracking can reliably measure fatigue across individuals and tasks. While eye behaviour may contain meaningful signals—such as blink rate or gaze instability—fatigue was not seen as a single, stable state with universal indicators.

Fatigue varies with task type, time of day, stress, and individual physiology. One concern was the lack of a “learning loop” that adapts to individual baselines, increasing the risk of misclassification. This highlighted a key scientific limitation: systems based only on population averages may overlook meaningful individual differences.

Participants emphasised that validity depends on recognising fatigue as context-dependent and dynamic. Without this, outputs may appear objective while remaining only partially grounded in real human variability.

Limitations and risks (false positives/negatives)

Participants identified risks of both false positives (detecting fatigue where none exists) and false negatives (missing genuine fatigue). These errors are significant because they may influence workplace decisions and safety interventions.

A key issue was how thresholds for fatigue are defined. If set too low, normal variation may be misread as risk; if too high, dangerous fatigue may go undetected. Participants also questioned whether observed improvements might reflect behavioural changes caused by monitoring (a Hawthorne effect), rather than accurate detection.

There were concerns about “gaming” or “subversion.” Workers may adapt behaviour, appearance, or routines to influence the system, producing misleading signals. This suggests that error is not purely technical, but also shaped by social responses to monitoring.

Environmental and contextual factors

Environmental conditions were seen as a major challenge to reliability. Factors such as lighting, screen use, posture, and movement can significantly affect eye-tracking data.

Participants stressed that workplaces are not controlled environments. Changes in visual behaviour may reflect environmental conditions rather than fatigue. As a result, interpreting eye data without context risks misclassification.

This reinforced an ergonomic insight: the system must be evaluated within real-world conditions. Without accounting for environmental variability, contextual noise may be mistaken for fatigue, undermining both accuracy and trust.

Design for human performance

Participants agreed that design choices would determine whether the technology supports or disrupts workers. While there was recognition that fatigue detection could improve safety, this depended on how it is implemented.

Potential benefits included preventing accidents, improving working conditions, and informing better shift design or interfaces. Importantly, participants suggested the system may be most valuable when used to identify patterns and redesign work environments, rather than continuously evaluating individuals.

A key concern was that fatigue could be framed as an individual problem rather than a structural issue. If so, responsibility shifts from organisations to workers. Participants therefore emphasised that systems should be context-aware, avoid punitive use, support better work design, and retain human judgment.

Benefits, risks, and acceptability

Participants identified both promise and risk. Benefits included improved safety in high-risk settings and insights that could support better scheduling, workflows, and wellbeing.

However, risks were substantial. These included surveillance, misclassification, data misuse, and the normalisation of monitoring. A major concern was that fatigue could be reframed as personal failure rather than evidence of poor working conditions.

Acceptability was seen as conditional. Participants supported limited use in safety-critical contexts, but only with strong evidence of validity, clear purpose limitation, and robust privacy protections. Use would be unacceptable if it functioned as covert surveillance, lacked meaningful consent, or enabled disciplinary action.

Data access, transparency, and trust

Data ownership and access were key concerns. Participants questioned whether fatigue data belongs to the worker, employer, or system provider. There was strong resistance to unrestricted managerial access, due to risks of secondary use.

Participants emphasised that access must be limited, purpose-specific, and clearly communicated. Data should not be repurposed beyond safety or wellbeing objectives.

Trust was identified as essential. Transparency must include clear explanations of what is measured, how it is interpreted, and what consequences follow. Trust also depends on fairness: if only workers are monitored while management remains unobserved, systems may be perceived as unequal and illegitimate.

Participants also stressed the need for contestability. Workers must be able to challenge or question fatigue classifications; otherwise, transparency becomes superficial.

Unintended consequences

Several unintended consequences were identified. One key risk is redefining fatigue as an individual trait rather than an organisational issue, potentially shifting responsibility away from employers.

Behavioural adaptation was also expected. Workers may perform compliance or adjust behaviour to satisfy the system, reducing both validity and authenticity.

Participants raised concerns about fairness and discrimination. Factors such as age, health, disability, or neurodivergence may influence how fatigue is detected, meaning the system could reproduce inequalities under a scientific appearance.

Finally, the system may deepen divides between managerial and worker perspectives—appearing as a safety tool for organisations but as surveillance for employees—potentially undermining trust and workplace relationships.

Conclusion

The focus group did not reject eye-tracking fatigue detection outright, but expressed cautious, conditional acceptance. The key issue was not only whether fatigue can be measured, but how such measurement is governed and used.

Participants emphasised that the technology is most legitimate when it supports safer environments and improved work design, rather than individual monitoring. Ultimately, its success depends not only on technical accuracy, but on whether it is scientifically valid, socially accountable, and ethically trustworthy.