| Summary: | <p>The Internet has rapidly revolutionised recruitment practices. Over the last two decades or so, there has been one particularly prominent change: the rise of AI recruitment technology (AI-rec-tech). This includes technologies which source, screen, assess, and select candidates. Despite their widespread use, there has been a significant lack of research concerning what lies behind the design.</p>
<p>This research offers one of the first in-depth, empirical, qualitative explorations into how vendors understand ‘gender bias’. Moreover, it studies how this understanding is mobilised in system design. This investigation takes place within eight AI-rec-tech vendors which produce assessment technology. It uses data from online qualitative interviews, walkthroughs and autoethnography of the assessments, alongside material from conferences, company websites, and reports.</p>
<p>An adapted version of Goffman’s theory on ‘frames’ (1974) is used to illustrate vendors’ understanding (their conceptualisation) of gender bias, and to unpack how this is transposed to system design (their operationalisation). This thesis argues that we can make sense of how vendors frame gender bias in hiring through three discrete, but not mutually exclusive, frames: (1) the intrapersonal frame, (2) the structural frame, and (3) the statistical frame. When these frames are held together, they display a socio-technical interpretation of gender bias in hiring.</p>
<p>This research is the first to apply theory on frames to study AI design. I argue this theory is valuable due to its translatable analytical provocations. Throughout, I employ elements of Goffman’s theory to scrutinise the ‘vulnerabilities’ of these frames. The vulnerabilities don’t offer certainty on the impacts of the systems but highlight complexities within or surrounding the frame itself. This enables recommendations concerning vendor claims, system design, regulation, auditing, and an agenda for future research on this topic. Overall, this thesis supplies an important contribution, offering rare insight into these companies, their rationale, and practices.</p>
|
|---|