Evaluated by a robot – an experimental-ethnographic study of the automation of the recruitment process in a Swedish municipality

AI (artificial intelligence) is changing the way modern society functions in several domains.

The automation of the assessment of individuals, such as in recruitment, has been identified as risky from an ethical perspective, as it is concerns assessments that affect people´s life chances.

At the same time, development in this area is driven by large cost reductions, as human assessors are more costly than those made by AI.Proponents of AI-based recruitment also point out that the incidence of discrimination may decrease with AI as it can be programmed to ignore factors such as applicants´ gender, ethnicity, age, etc.

At the same time, sceptics argue that AI risks reproducing the prejudices that already exist in the society in which they are used, and that AI may thus reproduce inequality to a greater extent than human assessors.The purpose of the project is to investigate how assessments of job seekers made by AI-based recruitment systems differ from assessments carried out by recruiters.

When in the process do differences arise, what type of differences are there, and why do they arise?

For which social categories - gender, ethnicity, or age, do differences in assessments have the greatest impact?The project is based on a unique opportunity to study a large Swedish municipality that for a year will use an advanced AI-based interview robot in its recruitment processes.

The project combines an experimental approach, where we compare the interview robot´s and the municipality´s recruiters´ evaluations of the same jobseekers, with an ethnographic study where we follow the municipality´s work during the year when the robot is tested.

This will provide a deeper understanding of the results of the experiment and of how recruitment processes change with automation.Research on the real effects of AI-based recruitment in organizations is so far very limited, even though the few studies that have been conducted show that AI risks reinforcing already established patterns of inequality.

The project thus contributes with new much-needed knowledge, speaking to organizations that are interested in automating their recruitment processes, as well as to policy makers on both AI and the labor market.

The project also contributes to the general discussion about the societal effects of AI implementation.We apply for funding for three years for two part-time researchers: for preparation, the one-year data collection, and for subsequent analysis and presentation of results.