Artificial intelligence can also develop prejudices

This is shown by thousands of simulations.

You do not have to be a person for having prejudices. A new study published in the journal Scientific Reports shows that machines can develop prejudices as well. They do this simply by taking over this behavior from each other.

Prejudices and machines
It may seem that having prejudices is a human-specific phenomenon. So it would require human knowledge to form an opinion or stereotypical image about someone. But nothing is less true. For example, it appeared earlier that some types of computer algorithms showed racism and sexism. This they learned from public records generated by people. The new study, however, shows that computers are also capable of forming independent prejudices.

Game
The scientists set up a simulated game for this research. In this game, each computer had to choose whether to give something to someone within their own group, or to someone from another group. They made this choice based on the reputation of each computer. Their own strategy also played a role. And the findings show that the computers became increasingly biased towards those from other groups.

Simulations
After thousands of simulations, the computers learned new strategies by copying each other. As a result, the researchers started to get a grip on how prejudices evolve over time and which conditions it promotes or hinders. It turns out that the computers developed strategies that gave them a better outcome in the short term. This means that these decisions do not require high cognitive skills to develop preconceptions.

Autonomous machines
“Our simulations show that having prejudices is a powerful force of nature,” says researcher Roger Whitaker. “In addition, it can easily be stimulated over time in virtual populations.” Something that can sometimes turn out to be disadvantageous. Especially now autonomous machines – for example self-driving vehicles or [Internet] Internet of Things – are increasingly coming to the rise. “It is quite possible that autonomous machines in the future might be susceptible to the adverse phenomena that we encounter in the human population,” warns Whitaker.

Still there is still hope for who feels great disappointment at the moment. For example, under certain circumstances, including the presence of separate subgroups within a population, there are fewer prejudices. Perhaps this eases the pain.

Loading...