Falling Asleep at the Wheel by Fabrizio DellAcqua

The "Falling Asleep at the Wheel" by Fabrizio DellAcqua discusses the challenge of incorporating artificial intelligence (AI) into organizational settings while still ensuring that human workers remain engaged and attentive. While firms use AI algorithms for recruitment decisions, they are often hesitant to rely solely on AI for hiring, preferring humans to have the final say. However, as AI performance improves, there is a greater temptation for humans to delegate decision-making to the AI, which may result in mindlessly following its recommendations without critical thinking. To address this tension, the author developed a formal model that examines the trade-off between AI accuracy and human efforts. As AI accuracy increases, humans may become less incentivized to put in effort. This study conducted a field experiment with a recruitment firm where professional recruiters evaluated job candidates based on algorithmic recommendations of varying quality. The experiment revealed that higher-quality AI assistance resulted in less accurate evaluations by experienced recruiters, suggesting that their ability to think independently and improve on AI recommendations was hampered. The findings of the experiment highlight the need to consider the objectives of the organization and the specific task when designing human-AI collaborations. Maximizing human-AI performance does not necessarily mean maximizing AI performance alone. In some cases, lower-quality AI may be preferable to encourage human effort and critical thinking. The paper provides a theoretical overview, a formal model, and details of the experimental design and analyses. The results emphasize the complex interaction between skills and technology in human-machine collaboration and the importance of considering task features when designing effective structures for such collaborations.