Can AI’s Recommendations Be Less Insidious? (IEEE Spectrum 2 Oct 2022)

Netflix. YouTube. TikTok…
Are recommendation engines helping us a bit too much?


Read the full article here

Recommender systems often use machine learning to find patterns in data. Reinforcement learning (RL) is a sort of machine learning that lets computers predict behaviour several steps ahead. It is what gave DeepMind’s AI the power to beat humans at Chess and Go.

With RL, “you have an incentive to change a chessboard in order to win,” says Micah Carroll, a computer scientist at the University of California, Berkeley, doing work on the subject. “There will be an incentive for the system to change the human’s mind to win the recommendation game.”

But new research suggests a way to measure—and reduce—such manipulation. Find the paper here.

1 Like