Does a Compromise on Fairness Exist in Using AI Models?

Artificial Intelligence (AI) has been increasingly used to assist decision making in different domains. Multiple parties are usually affected by decisions in decision making, e.g. decision-maker and people affected by decisions. While various parties of users may have different responses to decisions regarding ethical concerns such as fairness, it is important to understand whether a compromise on fairness exists in using AI models. This paper takes AI-assisted talent shortlisting as a case study and investigates perception of fairness, trust, and satisfaction with decisions of both recruiters and applicants in AI-informed decision making. The compromises on fairness between decision-maker and people affected by decisions are identified which are then explained by social and psychological theories. The findings can be used to help find compromising points between decision-maker and people affected by decisions so that both parties can reach for a balanced state in decision making.