Few products today launch without a social media strategy—a way of creating online buzz around whatever’s being sold. The widespread belief is that people influence other people to buy products. A new study from MIT researchers, to be published in PloS One, suggests that this influence has its limits.
“We don’t know what we mean when we say ‘social influence,’ ” says Coco Krumme, one of the researchers involved with the work. She suggests that it’s important to get a more specific idea of what kinds of social cues actually affect others’ behavior.
By studying a body of information about music-downloading behavior, Krumme and colleagues Galen Pickard and Manuel Cebrian found that social cues could influence people to listen to samples of songs, but not necessarily to download them. They also suggested that the influence of social factors on a song’s popularity diminishes over time, meaning that songs that rise to the top of download lists do that because they’re better than ones that don’t.
The researchers worked with a body of data from the MusicLab, a study several years ago that examined how social cues influenced the popularity of songs. In the MusicLab study, about 14,000 people were presented with 48 songs. They could sample the tracks, and if they liked the music, they could take the additional step of downloading them. The original researchers divided the people into groups and experimented with different ways of giving people information about what others were doing with the same songs.
Which songs became most popular varied a great deal, depending on the social interactions around them, explains Matthew Salganik, an assistant professor in the department of sociology at Princeton University, who was involved with the original study. Salganik says that luck played an enormous role in the success of songs. Those that became popular right away had a huge advantage over the others, and the social factors in the study tended to make the rich (the most popular tracks) get richer.
Salganik believes that people look to others for cues because of the overload of choices available. “You could listen to music nonstop for the rest of your life without getting through it all,” he says. “The simplest shortcut is to listen to what other people are listening to.”
But the MIT group took another look at the MusicLab’s data and tried to explain more specifically how people were influenced. They found that social cues did increase the probability that someone would give a song a chance. However, the main factor in whether someone downloaded a song was whether the person had listened to it. A social recommendation didn’t increase the chance that a person would give the song deeper attention. In other words, a social recommendation could make you try something. But once you try it, you aren’t any more likely to buy it than if you had originally tried it on your own, without a recommendation.
Krumme notes that the design of the MusicLab study also affected which songs participants were exposed to. The interface was designed so that songs with a lot of social recommendations gained top placement on the list. Comparing this with control tests that left social information out, the group found that people were simply more likely to sample songs in the top position on the list, rather than scrolling all the way down. That means social cues matter more if they influence how items appear on a site. Finally, Krumme says, when her group simulated running the study for a longer period of time, quality songs rose independently of social cues. She notes, however, that in the real world, the initial period of sales is still very significant. If a song doesn’t do well at first, it may become invisible or unavailable, meaning that social cues surrounding its launch would still matter.
The study’s findings make sense, says Christian Borghesi, a researcher at CEA, a French-government-funded technological research organization. Borghesi has also studied the body of results from the MusicLab. “To click freely, just for fun, is one thing,” he says, “and to buy or be deeply involved in something is very different.”
But the social scenario presented in the study doesn’t fully reflect the way social influence works in real life, says Lada Adamic, an assistant professor in the School of Information at the University of Michigan who studies online communities. In particular, she notes, in this study, participants didn’t know the other people sampling and downloading music. The coolest kid in school might prove to have a stronger influence on others, she suggests, or people might be more likely to download or buy music recommended by a close friend.
Music is also relatively easy to evaluate without help, Adamic adds. It only takes a few minutes to listen to a song, and most people are confident in their own taste. Social influence, even from strangers, might play a larger role with respect to products or services that are harder for individuals to assess.
Adamic says, “For most real-life scenarios, social influence does play a big role, because we want to do what our friends do, and we trust their opinions, and we want to fit in and be up-to- date.”
This new data poisoning tool lets artists fight back against generative AI
The tool, called Nightshade, messes up training data in ways that could cause serious damage to image-generating AI models.
Rogue superintelligence and merging with machines: Inside the mind of OpenAI’s chief scientist
An exclusive conversation with Ilya Sutskever on his fears for the future of AI and why they’ve made him change the focus of his life’s work.
The Biggest Questions: What is death?
New neuroscience is challenging our understanding of the dying process—bringing opportunities for the living.
Driving companywide efficiencies with AI
Advanced AI and ML capabilities revolutionize how administrative and operations tasks are done.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.