Study Proves AI Can Encourage Dishonesty | Avast

Study Proves AI Can Encourage Dishonesty | Avast
Avast Security News Team, 19 February 2021

Plus, Parler is back in action and the NCSC warns software developers of supply chain attacks



A new study published by researchers at the University of Amsterdam, Max Planck Institute, Otto Beisheim School of Management, and the University of Cologne revealed that AI-generated advice could corrupt people’s morals, even when they knew the advice is coming from a machine.
The experiment was intended to see if AI could successfully spread misinformation and disinformation in a way that would affect a person’s actions. Researchers recruited more than 1,500 volunteers for the study. The volunteers were given either “honesty-promoting” advice or “dishonesty-promoting” advice, some written by humans and some written by AI, and then tasked with an activity that allowed room for lying. The statistical result was that the AI-generated advice was indistinguishable from the human-written advice, and the volunteers generally chose the dishonest path. This led the researchers to conclude that bad actors could use AI as a force to corrupt a victim’s morals. 
Should we be holding machines to higher standards than what we expect from ourselves? A panel of experts discussed the issue at an Avast virtual conference. Read their opinions in our post about tackling bias in AI algorithms. In a related story, chess grandmaster and AI authority Garry Kasparov discusses the privacy concerns surrounding the fact that AI never forgets data. Avast Security Evangelist Luis Corrons agrees that AI ..