On 16-17 October 2024, Lithuanian and foreign researchers presented papers on human-centric artificial intelligence at the international conference "Human-Centric AI: ethics, regulation and safety", which took place at Vilnius University (VU). The event, organised by the Faculty of Law of Vilnius University, was an interdisciplinary one, and the speakers considered the ethical, safety and regulatory challenges of artificial intelligence in the context of various potential uses of the technology. On the first day of the conference, Monika Šukytė, a junior researcher and a PhD student at the Law Institute of Lithuanian Centre of Social Sciences and Vilnius University Faculty of Law, presented her research.

The PhD student started her presentation "Bias in Facial Recognition Technologies: How it Happens and What is at Stake" by presenting the consequences of biased results in facial recognition technologies. She discussed a case that was reported in the international media, in which US police officers arrested a person unrelated to the crime after relying on a false identification result from facial recognition technologies. The two men were linked only by their dark skin colour.

In her assessment of biased AI systems, Šukytė drew attention to a research-based conclusion: algorithm errors are not just an isolated technical problem, but also a combination of mistakes made by the actors who develop, use and regulate facial recognition technologies.

In the context of regulation, she said, it is the EU legislators who have an important role to play in addressing the challenges posed by biased facial recognition technologies and the discriminatory practices they entail. "However, the new EU AI Act does not fully fill the gaps in the application of these technologies in law enforcement, as it introduces a confusing system of exceptions," concluded M. Šukytė.

The video of the presentation by PhD student M. Šukytė (from 37:49 to 49:30) can be viewed here.

Presentation slides (in English).