Reframing machine learning and fairness in organizations and recruitment

Machine learning, a technology with surprisingly useful applications in many fields, has had inconsistent success in its organizational applications, notably in recruitment, where problems of fairness to applicants and opacity in reasoning within the machine’s “black box” have arisen. It is argued here that opening this black box is not so much to look inside the machine itself, as it is to examine the context, such as recruitment, in which the machine is put to work. The machine learning and fairness problem can be usefully reframed as a problem in organizational learning first of all, where the organization must learn not only in interactions with the worlds of practice in which it operates, but in interactions too with its own machines, two spheres of learning which must somehow be brought together through their associated routines. To address this problem in recruitment, it is suggested that regulatory routines, lotteries, experiments, and policies and audits can play an important role in furthering learning and fairness.