New research from the University of Georgia attempts to test whether individuals trust algorithms or a consensus of other people.
The study, “Humans rely more on algorithms than social influence as a task becomes more difficult,” was co-authored by Eric Bogert, a Ph.D. student in Terry College, as well as Professor Rick Watson and Assistant Professor Aaron Schecter.
Algorithms work in a couple of different ways to try and help users when they interact with technology. YouTube uses algorithms to recommend videos that are similar to the ones you have already watched, and a directions app will recommend alternate routes if they think it will help a driver avoid traffic. The algorithms that the three authors studied worked to count the number of objects in a picture.
“The original motivation for this problem came from the original wisdom of the crowd idea. If I show you a jar of jelly beans and I ask you to guess how many jellybeans are in the jar. One person is probably going to get it wrong. But if a thousand people guess and we take the average of their guesses we'd probably get something close to the right answer.” Schecter, the study’s co-author, explained, “We used a very similar task where we showed people a picture with a lot of different objects in it and we asked them to guess how many were in there.”
The researchers then suggested a different number, either calculated by an algorithm or by a consensus of other people. Following that, each participant was asked whether they wanted to change their previous answer.
“Whatever the algorithm said, whether it was good advice or bad advice, they were changing their answers. In most cases, they were taking the average of their original answer and whatever the algorithm said. Whereas if it was other humans providing a guess, they did not actually change that much.” said Schecter.
The applications for this research are far-reaching. The study was originally funded through a $300,000 grant from the U.S. Army Research Office, but Schecter believes the conclusions can be built on to help shape how algorithms are used for setting bail, and whether or not someone is approved for a loan.
Schector added, “I think what we’re trying to show is that people need to be aware of the fact that even if they are giving you junk, you still may have a tendency to believe it. Because it comes from an algorithm.”
This conclusion is something that many people already believe. Janet Geddis is the owner of Avid Bookshop in Athens, Georgia. She stresses the importance of one on one connections when making a book recommendation.
Geddis said, “More often than not people come in able to describe the sort of reading experience they want, and I along with my colleagues at Avid Bookshop are all experts at books. So we are able to narrow down some options for customers based on what they’ve said they want to read.”
Geddis, like many business owners, is hesitant to fully embrace using algorithms.
“We do use data to help inform our decisions about what to stock in the store and what to recommend.” Geddis added, “At Avid we do have records for customers going back a decade now of what people have bought from us. That plus our conversations in the store, plus just knowing our customers on a personal level in a way that an algorithm never could, allows us to cater our recommendations, as well as what we stock in the store.”
Even though most people interact with algorithms quite frequently, many are cautious to include them more in their daily lives. Joe Silva hosts Athens 441, a weekly music show. The show tries to capture modern music through an eclectic lens. Silva is skeptical of streaming services that rely heavily on algorithms because he feels they do not offer a diverse enough spectrum of music.
“If you use a platform like Spotify or something it’s always going to drive you to the most common denominator.” Silva added, “So it is like, oh you like 80s music, and here is an 80s playlist. You might get one thing in a playlist of 100 tracks that you don’t know.”