Ů��С��Ƶ

Excuse me, is now a good time? Smartphones that learn when you’re most likely to respond to notifications could stop apps from interrupting at inconvenient moments.

and his team at Rutgers University in New Jersey are using machine learning to better manage the deluge of smartphone notifications. The group has created an app that uses information about a person’s location, current activity and personality to predict the best time to deliver notifications.

The app surveyed volunteers many times a day – using its own pop-up notifications – and asked them to rate how interruptible they were on a scale from “highly interruptible” to “highly uninterruptible”. It also used the phone’s sensors to collect data on the person’s location and movement. This information was combined with the volunteers’ personality test scores to build a profile of when they would be most happy to see a notification.

Fill the silence

Throughout the four-week trial, the app’s algorithm used this data to learn when best to send the notifications. If it predicted someone would be highly uninterruptible, it didn’t send a survey request. When it did send a notification, the researchers compared its predictions with the volunteers’ self-reported interruptibility scores.

After an average of 16 days, the app could predict someone’s score with an accuracy of 75 per cent. Although every volunteer had their own preferences about when and where they liked to receive notifications, the location data suggests they were generally least happy to be interrupted while out shopping, says Lindqvist. Unsurprisingly, people were happier to be interrupted when they were in a good mood.

People with similar personalities were also likely to have similar preferences. The personality test the volunteers took was based on the “big five” traits – extraversion, agreeableness, conscientiousness, neuroticism and openness. Taking those scores into account increased the algorithm’s accuracy by 10 per cent, says Lindqvist. The team will present the work at a in Denver, Colorado, in May.

Hold the spam

Most smartphones already offer some options for filtering notifications, say at the Technical University of Cologne, Germany, but it’s difficult to create a system that works for all situations. “There is a very broad range of how important notifications are. Not all of them are urgent,” he says. An ideal system could hold back most notifications if a user was at the gym, for example, but still send an alert if they received an important email.

A team at South Korea’s Advanced Institute of Science and Technology has taken a different approach to the same problem. They have developed a system that uses a phone’s microphone and Bluetooth to detect natural “breaks” in social situations as they occur, such as pauses in a meal. It holds notifications until there is a silence, or the Bluetooth detects that a friend’s phone has moved away. In a trial, the system .

But recognising when people are most open to being interrupted won’t necessarily lead to fewer notifications overall, says Lindqvist. Instead, app creators could use that information to target notifications at times they estimate users are more likely to respond.