How addicted are people to social media? We found a way to measure it

According to our new study, a randomized experiment involving roughly 2,000 Americans, the truth is somewhere in between: We value social media, but we use more than we think is good for us. Our study finds that about 31 percent of social media use among people in our sample is caused by self-control problems. In other words, if people in our study could choose their preferred screen time in advance instead of scrolling uninhibited in the moment, they’d spend nearly one-third less time on social media.

We designed our study to measure two concepts that are central to most definitions of addiction: habit formation and self-control problems. Habit formation means that using now makes people want to use more in the future. For example, smoking cigarettes eventually makes people crave nicotine. Self-control problems mean that people want to use less in the future, but when the future arrives it’s hard to follow through. For example, most smokers would like to quit smoking, but few actually succeed. (There are differences between addictions to substances such as nicotine and behaviors such as gambling or Internet use, but clinicians increasingly recognize commonalities.)

Many academic fields study addiction, including medicine, psychology, neuroscience and economics, and each has its own well-established way of measuring habit formation and self-control problems. Our study used standard approaches that social scientists have used to study behaviors such as smoking, alcohol use and exercise.

For our study, we used Facebook ads to recruit about 2,000 American adults who were willing to install an Android app that would confidentially and securely measure their smartphone screen time. We randomly assigned them to treatment and control conditions and asked them to answer regular surveys over a three-month period. Our analysis focused on five social media apps — Facebook, Instagram, Twitter, Snapchat and YouTube — plus web browsers.

To test for habit formation, we randomly assigned some participants to be paid to reduce their social media screen time for three weeks. (During the three-week incentive period, we paid this group $2.50 for each hour that they reduced their social media consumption below their baseline levels.) We then tested whether the reductions persisted after the three-week incentive period was over. Such persistence would be evidence of habit formation.

Unsurprisingly, people responded to the incentive: It caused people to reduce their screen time by 39 percent (about 56 minutes per day) over the three weeks it was in effect. More interestingly, the effects of the incentive persisted well after we had stopped paying people. Six weeks after the incentive period ended, the group that had received the incentive was still using social media about eight percent less often than the control group. This provides strong evidence of habit formation. As a related example, when you pay people to stop smoking for a period of time, some of them stay off cigarettes even after you’ve stopped paying them.

To test for self-control problems, we randomly assigned some participants to have access to screen time limits in our study’s Android app. This functionality allowed people to set personalized daily time limits for each app on their phone, effective the next day. Once they reached the time limit they had set for themselves on an app, our software would force-quit that app. (We randomly assigned participants to settings that varied how long they had to wait before they could override the limits. Shorter waits provided more flexibility, while longer waits did more to prevent in-the-moment temptation.) These screen time limits combat self-control problems by allowing people to force their future selves to follow through with their current desires to use less.

Remarkably, 89 percent of people in this group used the screen time limit function. On average, they used the limits to reduce social media use by 17 percent over a 12-week period. Once people had experienced the screen-limitation feature, they were willing to pay to keep using it — an average of $4.26 for three weeks of access. These facts are evidence of self-control problems. Continuing our cigarette analogy, many smokers want analogous “commitment devices” to control their future smoking — for example, they’ll agree to give up money if they smoke again.

Habit formation and self-control problems interact. Using the data from our experiment, we created an economic model that predicts the long-run effects of self-control problems in the presence of habit formation. This model is what allows us to estimate that self-control problems account for 31 percent of our participants’ social media use. (There was a lot of variation across people: Self-control problems increase social media use by less than 10 minutes per day for 26 percent of participants, and by more than 100 minutes per day for 13 percent.)

These results from the experiments are consistent with separate qualitative evidence from our surveys. For example, 57 percent of people said they use their smartphone too much, while almost nobody said they use their smartphone too little. Both of our treatments — the incentive to reduce use and the screen time limits — reduced this survey measure of addiction (and others).

There is also some evidence that our treatments may have improved traditional measures of subjective well-being such as life satisfaction, anxiety and depression, but the sample isn’t large enough for these results to be statistically significant. (In a previous study, we found that quitting Facebook for the four weeks before the November 2018 elections significantly improved subjective well-being.)

Like any other study, our work has limitations. While we tried to avoid recruiting a sample that was too different from the average American, it was not nationally representative. Our participants knew they were in an experiment, and that knowledge could have affected their behavior and survey responses. And our economic model involves many assumptions.