Dynamic Delegation Thresholds: Real-Time Adjustment of Human–AI Roles via Cognitive Load and Trust Signals

While Lai et al. (2022) introduce conditional delegation via user-defined rules, most current research assumes delegation thresholds are set ex ante and remain static unless manually reconfigured. However, Salimzadeh et al. (2024) and Erlei et al. (2024) show that task complexity, error types, and trust levels dynamically shape human reliance and delegation choices. This idea proposes a system that senses real-time indicators—such as user cognitive load (eye tracking, physiological signals), trust (self-report, behavioral proxies), and AI error likelihood (predictive uncertainty)—to fluidly adjust who is 'in the loop.' The system could, for example, revert control to the human when it detects user under-reliance or cognitive overload, or increase AI autonomy when user trust and performance are high. This approach would move beyond preset rules to a genuinely adaptive delegation framework, potentially reducing both over- and under-reliance, and offering a more nuanced model of human–AI teaming in high-stakes, dynamic environments.

References:

1. Dealing with Uncertainty: Understanding the Impact of Prognostic Versus Diagnostic Tasks on Trust and Reliance in Human-AI Decision Making. Sara Salimzadeh, Gaole He, U. Gadiraju (2024). International Conference on Human Factors in Computing Systems.
2. Understanding Choice Independence and Error Types in Human-AI Collaboration. Alexander Erlei, Abhinav Sharma, U. Gadiraju (2024). International Conference on Human Factors in Computing Systems.
3. Human-AI Collaboration via Conditional Delegation: A Case Study of Content Moderation. Vivian Lai, Samuel Carton, Rajat Bhatnagar, Vera Liao, Yunfeng Zhang, Chenhao Tan, Q. Liao (2022). International Conference on Human Factors in Computing Systems.