Event Details

This seminar is online via Teams only and the joining details are: Meeting ID: 380 801 463 779 - Passcode: 2pV9Fg7X

Abstract: There is considerable discourse about the dangers of “bias” in the production of AI algorithms that results in a discriminatory impact on diversity such that, women, ethnic minorities, the young, the elderly and the disabled can be diversely affected. Although much of the literature concerning ethics and AI concludes that “bias” can be eliminated through “cleaning up” the data, we argue that there are two problems in drawing this conclusion. First, data is accumulated through past events which the algorithms then just replicate and second, any so-called “cleaning up” will be performed by engineers who, from our research data, exhibit and sometimes express the very white, masculine assumptions, norms and values that have generated what are claimed to be the biases that accompany AI. Drawing on our ongoing research on interviewing engineers working in the field of AI, we provide some evidence of this phenomenon and how they responded to questions particularly of gender and ethics.