What if AI had prejudices

Human prejudices can be transferred to Artificial Intelligence (AI), leading to decisions influenced by biases similar to our own. The topic was explored during Week against racism through an activity organised by Università della Svizzera italiana (USI) and SUPSI, in collaboration with Zonaprotetta and the Centre for the Prevention of Discrimination. Monica Landoni, Full Professor at USI Faculty of Informatics, spoke about it on "Il Quotidiano" (RSI).

The risk is that the same prejudices influencing human decisions may also appear in AI choices, unjustly disadvantaging minorities or groups historically facing more challenges, such as women and foreigners.

But how can we establish the right principles to follow? "Each of us has our own ethics, values and principles," explained Professor Monica Landoni. "A society typically identifies itself based on shared values, which means that its members establish what is considered ethical and right. This community determines what actions are acceptable and what is not."

However, not everything is arbitrary: "Some values, such as the well-being of individuals, are assumed to be universal. As informatics experts, we adhere to a code of ethics that mandates our work be used for the benefit of both users and the community."