In this research program proposal, we aim to investigate why experts override AI suggestions and identify design principles for more effective human-AI teams. Specifically, we propose testing whether increasing the perceived locus of control of human decision-makers over AI functions will lead to fewer overrides and improved performance. We present a mixed-factorial, multi-trial experimental design in which participants receive AI recommendations regarding demand forecasting decisions in a business simulation. Prior to each trial, one group specifies how they want the AI to function (experimental), and the other group does not (control). We use electroencephalography and oculometry to capture attention to recommendations and user interface elements. Behavioral data from a preliminary pilot study with four participants align with our hypotheses. We observed that participants in the experimental condition applied smaller adjustments to AI suggestions and had higher decision performance than the control group. The experiment's results will contribute to our understanding of AI aversion and inform the design of human-AI interactions to improve performance.