Balancing Accuracy and Acceptance of Algorithmic Hiring Decisions: Put the Human Touch into the Equation
研究提出两种混合决策方法（机械合成和临床合成），在保留决策者自主权的同时提高预测有效性，为解决算法使用中的自主性-有效性困境提供方案。
Abstract Although more valid predictions are made when information is combined algorithmically (mechanical prediction) rather than in the decision-maker’s mind (holistic prediction), decision makers rarely use algorithms in practice. One main reason is that decision-makers’ autonomy is restricted when algorithms are used to combine information. Unfortunately, affording decision makers greater autonomy in information combination decreases predictive validity compared to consistent algorithm use, creating an “autonomy-validity dilemma”. We hypothesized that two hybrid approaches to decision making—clinical and mechanical synthesis—should retain decision-makers’ autonomy while increasing predictive validity compared to pure holistic prediction. In Study 1 ( N = 261), mechanical and clinical synthesis resulted in higher predictive validity than holistic prediction, but user perceptions regarding these procedures were mixed. In Study 2 ( N = 610), mechanical and clinical synthesis again resulted in much higher predictive validity than holistic prediction, and these procedures were perceived much more positively than the strict use of a prescribed algorithm. Mechanical synthesis forms a promising solution to balance decision-maker autonomy and predictive validity in decision making.