impactmania has been awarded a follow-on grant for our work with the FINDHR organization on EU algorithmic hiring tools. This project, funded by the EU, involves top European universities and institutions. Learn more here: impactmania and EU CORDIS.
We arrived at the conclusion that bias in AI systems is not merely a technology problem, since machine learning algorithms are trained on historical data which are inadvertently shaped by societal structures, reflecting and perpetuating existing power dynamics, and inequalities. Even if the previously mentioned technical recommendations are institutionalized, job applicants optimize their resumes, and the system recommends a fairly diverse workforce — ultimately a company’s hiring manager may still select the candidate based on similarity of past hires.
Technology can definitely become better advisors to humans making better decisions once we correct existing datasets, conduct continuous bias audits, and add human oversight by a professional diverse team of people.
However, the overriding key recommendation we would have is that the AI field should diversify itself. A diverse and inclusive AI talent pool, especially from under-represented communities would anticipate and build a more equitable system. This will require investment in education, mentorship, and access to opportunities. Only then will we have an environment where sophisticated systems are created by people that represent our union.
For the full report, see publication on FINDHR.