Are public decisions made by artificial intelligence democratically okay?

Are public decisions made by artificial intelligence democratically okay?

The reason this is an important question is that already AI is being used in public decision-making. For example, in social services, healthcare, and in the legal system. While this may increase efficiency, will people accept these decisions? And do they live up to democratic standards?

We ask Eva Erman, professor of political science, who currently leads a research project on AI and democratic legitimacy at IFFS.

What is needed for decisions made by AI to be legitimate in a democracy?

"It will revolve around questions of transparency and accountability, and here AI raises some specific problems. It is usually said that for every demand for legitimacy, there must be a counter-demand for accountability. So what if AI has made a decision that is wrong? Who should be held accountable? Thus, there must be a human somewhere in the chain who can be held accountable. Another aspect is transparency, that decisions should be explainable. AI is very much a "black box". You don't know what happens in an AI."

Will it matter for which decisions AI is used?

"Yes. If you think of legislation as the most robust case of decisions, I find it very difficult to see how AI can legislate in a democratically legitimate way. Then there is a sliding scale towards policy-making down to more administrative decisions where AI is already used today. A working hypothesis is that different forms of decision-making may involve different normative criteria for legitimacy."

What potentially positive effects can arise from AI in public decision-making?

"The potential to streamline public services is significant. One danger often raised is that AI will be biased because the data it is fed with is biased. But humans are also extremely biased. Humans are influenced by upbringing and background, which AI is not. AI can also use much more data in a decision, which can make the decision appear more neutral. AI can also play a role in reducing corruption in a society by being more impartial and not giving advantages to certain groups based on friendship for example.