Wearable devices are pretty good at letting us know how many steps we've taken or how much quality sleep we got last night. Wouldn't it be great if wearables could keep each member of a couple informed of changes in the other's mood — and send alerts when it might make sense to tread lightly to avoid needless conflict?
We're not there yet. But new research suggests that such an app for wearables might be available in the not-too-distant future.
With the help of machine learning methods, researchers at the University of Southern California have created an algorithm that uses subtle changes in an individual's physiology to determine his or her emotional state. The researchers' eventual goal is to create an app that couples can use to reduce the amount of friction in their relationship.
"We're really interested in identifying what are the patterns of interaction of less satisfied couples," says Adela Timmons, a graduate student in psychology at USC and lead author of a paper describing the research. "Certain couples may have ways of responding when they get angry at each other. Some may have explosive arguments whereas other people might withdraw from their partner."
The paper was published in the March 2017 issue of the journal Computer.
The researchers recruited 34 college-age couples and fitted them with wrist and chest monitors that captured a day's worth of physiological data, including body temperature, heartbeat, and perspiration. The couples also carried special smartphones that listened in on their conversations during the day to look for changes in vocal tone and words suggestive of conflict (for example, the word "you" is often used to indicate blame).
Using the phones, the couples also responded to hourly surveys about their emotional states and recorded times of any conflicts that occurred.
The researchers analyzed the collected physiological data and transcripts of the conversations to predict the times when couples had disagreements. Then, they checked these times against responses to the surveys and discovered they could use the physiological data to determine conflict with an accuracy of about 86 percent. For the next phase of their research, the scientists plan to expand the kinds of data they capture to include time spent online, the amount of sun exposure, the frequency of text conversations between partners, and when someone had last eaten.
If the scientists can develop algorithms that predict emotional states in real-time, they think the resulting apps could help solve communication issues in relationships and improve people's emotional and physical health.
"We visualize an application that everyone could download and use in their everyday life," says Theodora Chaspari, a graduate student in engineering at USC and co-author of the paper. "It would track their relationship status each day. It would provide indications of how good or how bad relationships would be on a specific day. And beyond that, it could provide feedback or maybe online intervention."
But before that happens, the biggest obstacle to adoption of these systems is privacy and whether people will be comfortable having their emotional states broadcast to others, says Kathryn Hume, vice president of product and strategy for Integrate.ai in Toronto, which helps companies apply artificial intelligence technology to their brands.
But Timmons believes streamlining the monitors so they're less obtrusive will be the significant hurdle.
"Our hope is to create models that can work across devices as the technology continues to develop," Timmons says. "As the field of wearables becomes more sophisticated we're going to be able to collect more data on smaller, more stylish, single devices."