In recent years, the debate on the threats that the online universe can pose, led by algorithms that make use of artificial intelligence, when it comes to manipulating users, has intensified.
According to Ujué Agudo and Helena Matute from the University of Deusto. “It is important to educate people so that they do not trust and blindly follow the advice of algorithms. A discussion about who owns and can use the data. That makes these algorithms work efficiently is also necessary.” This is the main conclusion of a study carried out with the aim of examining whether algorithms can persuade people, overtly or covertly, about who to vote for and who to date.
Through Four Experiments
The researchers found that persuasion was possible and that different styles of persuasion (explicit, covert) were more effective depending on the decision context (political, dating). Each of the experiments started, according to The Next Portugal B2B list Web, with a fake personality test. When the participants complete it, they given a personality profile. They were told that they would inform the algorithm so that it determines the best results for them. But the reality is that all the profiles were the same.
In the first experiment, explicit manipulation was use to make the participants vote for a candidate. Told which candidates matched their personality in a high percentage and then asked who they would vote for. The result? People were much more likely to vote for the candidate the AI had told them.
In the second experiment it was based on indirect manipulation. The algorithm chose four politicians -in a covert way. And exposed their images to the participants so that they could become familiar with them. “Our prediction was that if we could make some candidates look more familiar. Then others through mere pre-exposure. Pre-exposed candidates would covertly attract more votes than control candidates,” the researchers explain. However, contrary to their prediction, the results suggest that the covert algorithm cannot influence the voting preferences of the participants.