Quantifying AI Fatigue and its Impact on Workplace Performance and Safety

This PhD research positions AI fatigue as a measurable psychosocial risk emerging when artificial intelligence is embedded into everyday workflows. While foundational frameworks of technostress (Tarafdar et al., 2007; Ragu-Nathan et al., 2008) focus on general digital overload, AI fatigue is distinctly driven by the non-deterministic nature of modern algorithms. This interaction necessitates a constant "verification burden", reflecting the "verification-value paradox" (Yuvaraj, 2025), where AI's efficiency gains are heavily offset by the cognitive toll of auditing probabilistic outputs. Driven by information overload, workflow fragmentation, and black-box monitoring pressures, this AI-induced exhaustion degrades attention and encourages superficial compliance, ultimately undermining organisational resilience.

The research proceeds through three interconnected strands:
*Strand 1: Conceptualisation of AI Fatigue. This phase maps the boundary conditions of AI fatigue, distinguishing it from traditional technostress and general burnout. It establishes AI fatigue not simply as a byproduct of "too much technology," but as the specific consequence of the evaluative labour required to manage opaque, adaptive systems. By establishing where AI fatigue departs from traditional technostress (Tarafdar et al., 2007) and general workplace burnout, the project provides a necessary foundation for targeted intervention.

*Strand 2: Scale Development. This strand involves designing and validating a multidimensional measurement toolkit. This modular instrument will be developed to capture the cognitive, emotional, and behavioural components of AI fatigue alongside safety-relevant indicators such as attention lapses, quality drift, and a diminished willingness to raise concerns.

*Strand 3: Empirical Validation. The final strand tests the toolkit within a professional environments to determine if high AI fatigue scores correlate with objective performance drops, such as increased error rates and longer task completion times, or decreased user satisfaction.

The primary contribution of this project lies in transforming the abstract experience of AI fatigue into a tangible asset for evidence-informed AI governance. By providing an empirically validated tool to quantify the hidden costs of AI integration, this research enables organisations to track monitoring intensity and mitigate cognitive overload before it translates into operational failure.

Supervisors

Dr Elaine Chen

Dr Gwen Chen

Dr Hai Dang Nguyen

References

Tarafdar, M., Tu, Q., Ragu-Nathan, B.S. and Ragu-Nathan, T.S., 2007. The impact of technostress on role stress and productivity. Journal of management information systems, 24(1), pp.301-328.

Ragu-Nathan, T.S., Tarafdar, M., Ragu-Nathan, B.S. and Tu, Q., 2008. The consequences of technostress for end users in organizations: Conceptual development and empirical validation. Information systems research, 19(4), pp.417-433.

Yuvaraj, J., 2025. The Verification-Value Paradox: A Normative Critique of Gen AI in Legal Practice. arXiv preprint arXiv:2510.20109.