èƵ

App guesses your emotions to target you with adverts

Data from smartphone sensors can be used to predict our mood, leading to better movie recommendations – and more effective ads
Man with phone
How are you feeling?
EschCollection/Getty

In the mood for love? If so, you probably won’t want your phone to suggest a slasher movie or a thrash metal album. An app that works out how you are feeling could allow recommendation systems to only make suggestions that chime with that mood.

Called MoodExplorer, the app has been designed by Wenzhong Li at Nanjing University, China, and his colleagues. They say that assessing someone’s mood from moment to moment has been a key factor missing from personalised recommendation engines, often leading to inappropriate suggestions that people simply dismiss.

“Someone who likes romantic movies might prefer to watch an action film if they are feeling angry, perhaps, to relieve that anger,” says Li. “So a smart recommendation system really needs to take our moods into account.”

Smartphones are ideal devices for such a system because they are filled with sensors that detect light, sound, motion and location, all of which might help deduce a user’s emotional state. The way you use your phone’s apps can also say a lot about your mood. Previous work has shown that phone use can pick up on depression, for example. Now Li and his colleagues are aiming for a fuller emotional palette.

Emotional audit

To create the app, the team trained a machine learning system on data from 30 students who were asked three times a day for a month if they felt happy, sad, angry, surprised, afraid or disgusted. In tests, the AI could then guess the mood of participants correctly from their smartphone data 76 per cent of the time.

Li thinks that MoodExplorer could help humans interact with robots as well. A team in Sweden is for robots to let them pick up on a person’s mood through touch, for example. But MoodExplorer would give a robot extra information to work with.

Knowing a person’s emotional state could have wider benefits, says at the Massachusetts Institute of Technology’s Media Lab. “There are much more important reasons to sense mood than to modulate which ads we get,” she says. “If a person’s mood data shows that they’ll get the most benefit from advice like: ‘take a walk with a friend’, ‘call your buddy, Joe’, ‘skip Facebook today’ or ‘go to bed 2 hours earlier tonight’, then that should be prioritised.”

Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies

Topics: Artificial intelligence / Machine learning / Psychology / Robots / Smartphone