<p dir="ltr">Human interaction with different levels of automation, ranging from simpler decision aids to fully autonomous systems, is becoming increasingly common in society. This is particularly true in safety-critical applications. If humans exhibit inappropriate reliance, meaning depending either too much or too little, the consequences can range from failure in completing a task to injuring a human worker. For this reason, researchers have studied human reliance behavior on different levels of automation. Seminal results have shown that human reliance decisions are strongly influenced by their self-confidence and trust in the automation. However, even if a user trusts the automation, the environmental conditions may be sufficiently risky such that the user should not use it. Indeed, researchers have shown that risk perception can influence decisions, but despite this, relatively little research has studied human risk perception as it relates to decisions about reliance on automation. In this thesis, a new model for how human risk perception (RP) and human self-confidence (SC) affect their reliance on a visual aid is proposed and validated. A novelty of the model is its emphasis on the relative difference between RP and SC and demonstrating that this difference can accurately predict human reliance behavior. A secondary contribution of this thesis is the utilization of the model to design an algorithm that automatically decides when to provide the aid to the user. Main findings include the ability to calibrate a user’s cognitive state when using the algorithm that automatically provides the aid to the user. Another finding is the improvement in cognitive states during trials without the aid after trials in which an algorithm decided when to provide the aid. Other findings include analyses of the statistical differences in cognitive states based on an individual participant’s subpopulation membership. Extensions of these contributions to human-machine interaction contexts is discussed in future work.</p>