What our smartphones reveal about our social habits

alex pentland MIT

As you might have seen this week, we can learn a lot about people and their habits from their smartphones. Dr. Alex “Sandy” Pentland is an MIT professor and thought leader, and has been working as a data scientist for decades. Pentland’s work has made him one of the most-cited scientists in the world. Below, we’ve broken down some of the findings from his research papers on mobile data, social networks (both online and offline) and artificial intelligence.

We can detect and fight spam

twitter normal account vs spam account
Adapted from “If it looks like a spammer and acts like a spammer, it must be a spammer: analysis and detection of microblogging spam accounts” A. Almaatouq et al. (Figure 2) (2016)

Spam isn’t just the stuff of Monty Python musicals, Hawaiian delicacies and junk email folders. As evidenced by the 2016 U.S. presidential election, there are social media accounts designed to clog networks with automated messages, some of which might be indistinguishable from real tweets from real users.

Based on research conducted a few years ago, there was a 355 percent growth in social media spam in the first half of 2013 alone, a rate which has likely increased exponentially as both the number of social media users has increased and as software for creating spamming accounts and messages has become more intelligent. Twitter itself identified spamming practices as one of its primary concerns in its initial public offering filing when the company went public.

But how can a spam account be tracked and verified?

In a paper published in 2016, Pentland at MIT and his team collected data on more than 100 million messages sent during a single month in 2013, sent by 30 million accounts. They found an estimated 7 percent of accounts were suspended by Twitter for violating its algorithms, posting too frequently and tripping an alarm that it might be operating fraudulently. Spam accounts can be identified by “aggressive” following, spreading misleading or false information or malicious links to websites containing viruses or other activity that indicates a real person is not operating the account.

In the graph above, we see what a “normal” Twitter account looks like (top row) versus a spam account (bottom row), based on the number of tweets sent and the number of accounts followed.

Ask any American voter whether inauthentic social media accounts can cause problems, or talk to a reporter about the uphill battle with “fake news” in the past few years.

It takes very little to create a whole army of spam and bot accounts that can be used to create the impression of a viewpoint or a “majority” of people demanding action when it’s really just a shadowy campaign creating the illusion of support or controversy. A large number of fake accounts retweeting information could be enough to create a trending topic, leading to people clicking on inaccurate or wholly fabricated “news stories” filled with lies, half-truths and misleading information. Being able to detect, root out and eliminate these accounts helps keep good data in the hands — and feeds — of people who need it.

We can create better event and trip experiences

smartphone tourism map
Analysis of Tourism Dynamics and Special Events through Mobile Phone Metadata” Y. Leng et al. (Figure 4) (2016)

We use our smartphones to show off our travels, snapping photos and posting to social media to make our friends at home jealous. However, data collected from anonymous call detail records can also help tourism and event planning companies devise better occasions and festivities to further attract tourists.

Instead of relying only on surveys and interviews, it’s possible to track the country of origin for people attending an event, follow where they go while in another country or location and determine repeat visitors. Big data applications provide better and deeper information than surveys alone.

Using data originally collected for billing purposes, from calls made to and from phones visiting the tiny country of Andorra, Pentland and his team determined 90 percent of tourists visiting the country for a Cirque du Soleil event were from Spain or France. The remainder of tourists for the event were from Belgium, the Netherlands, Portugal and Russia. Also using call records, Pentland and his team were able to distinguish iPhone users from those with other, non-Apple devices and determine the percentage of each and where they came from (Spain, Russia, etc.).  

There’s an obvious benefit to this kind of information if you’re in the tourism business: It’s possible to design better events and opportunities with the shared interests of your ticket buyers in mind and can help them spend more money in your town or during the course of your event without having to go too far afield.

But as a traveler, it also helps you get a better overall experience on a single trip: If event planners have access to this information, or can learn from previous years’ data, you’re more likely to have better services (read: more bathrooms, shorter food lines, better on-scene transportation) on a  trip to a major international destination, with everything from translation services to currency exchanges to specialty maps categorized by points of interest.

Shared experiences matter more than Facebook posts when it comes to making friends

MIT social network dorm
Adapted from “Modeling the Co-evolution of Behaviors and Social Relationships Using Mobile Phone Data” W. Dong, B. Lepri, A. Pentland (2011)

Remember the first few weeks of college, when everyone was new and terrified but wanted to come across as cool and fun and confident, and therefore willing to overlook some bad habits in order to make new friends?

Studying a college dorm for one academic year, and relying on provided smartphones to monitor activity, social media use, interaction with other users of the same phone and call records, Dr. Pentland and his team could observe that friendships were formed most commonly based on shared experiences. The research indicated a student was five times more likely to consider someone a friend if they lived on the same floor of a small residence dorm or in the same building than a classmate that lived in a different dorm.

However, as the year went on, students branched out: During the fall semester, students socialized with half of the people living in their dorm, but by the end of the year, that number fell to one-third, with students picking up friends who lived in different buildings.

All age groups reported adding new people to their friend group through the duration of the year, but graduate students adding on average nine new friends, compared with just five new friends for freshmen.

When looking at the way in which people interacted – discussing politics, interacting with Facebook posts or tweets, etc. – online activities (Facebook and Twitter networks) are the least structured and friendships created through shared living situations are the strongest and most important to those students.

That’s not to say everyone in this particular dorm considered everyone else in the building a best friend, but there were more common experiences and points of reference than with people who they met from class or another social group.

We’re more likely to do what Yelp suggests

social restaurant order distribution
Adapted from “The Impacts of Just-In-Time Social Networks on People’s Choices in the Real World” K. H. Lee, A. Lippman, A. Pentland (Figure 5) (2011)

“What do you want for dinner tonight?”

When visiting in a new city or just trying a new restaurant, how do people decide what to order?

People are so used to researching their purchases online — comparing prices, shipping fees, the availability of various options — that not being able to make an informed choice based on considering all variables makes some uncomfortable. But when ordering dinner in an unusual setting, the only data available other than a list of special dishes comes from reviews left by previous patrons.

Pentland and his team created a Social Menu for an expensive seafood restaurant. Survey participants were given iPhones instead of traditional menus; the Social Menu listed food items by category (appetizers, sides, fish, etc) but also included a Friends’ Choice section which would populate recommendations based on what other survey participants ordered. One group was able to see their friends’ names along with what they had ordered or what they had indicated they preferred before going out to dinner. Another group saw how many other participants ordered a dish but without any names affiliated with each selection; the fourth group saw the number of their friends that ordered a given item or expressed a preference for it.

The group that knew precisely what their friends ordered made their meal selections faster than those who did not have that specified data on which to make their choices. They were deemed confident in their on-the-fly choices, more so than a group that had more anonymous data or more than 12 minutes to order dinner.

In other words: If you know your friend liked the seared tuna from a fancy seafood restaurant, you’re more likely to order it, and quickly, than if there’s no frame of reference upon which to make a choice.

Mom was right: Not getting enough sleep makes people cranky

Adapted from “Using Social Sensing to Understand the Links Between Sleep, Mood, and Sociability” S. Motoru et al. (Figure 3) (2011)

We all feel better, happier and more outgoing after getting a good night’s rest, but how much sleep do we need to feel refreshed?

Dr. Pentland and his team, instead of relying solely on self-reporting, studied a group of people living in the same area over the course of a few months. Between March and September 2010, they collected information from 54 people out of a community of 400, all of whom provided information via a social and behavioral software monitoring platform. Face-to-face interactions were tracked through Bluetooth proximity sensing and other social interaction patterns. Those involved in the study also provided monthly, daily and weekly surveys with information about relationships, mood, diet, exercise and sleep.

The results? People are far more likely to feel sociable after longer periods of sleep (7-8 hours) than fewer intervals of rest. The most social people were the ones who had received the most sleep the previous night, with people who slept less reporting less social activity. Most people (57 percent) reported getting about 7-8 hours of sleep, with 32 percent reporting 6 hours or less and the remainder clocking 9 hours of sleep or more per night. The study also found that people who were less socially inclined were more likely to be in a “poor mood” than those who were more sociable and had more sleep.

The study also focused on couples and found that one partner’s sleep patterns and mood might be affected by their partner’s amount of sleep and mood as well. But that’s probably not surprising either.  

About Futurithmic

It is our mission to explore the implications of emerging technologies, seeking answers to next-level questions about how they will affect society, business, politics and the environment of tomorrow.

We aim to inform and inspire through thoughtful research, responsible reporting, and clear, unbiased writing, and to create a platform for a diverse group of innovators to bring multiple perspectives.

Futurithmic is building the media that connects the conversation.

You might also enjoy
an automized forklift truck in a warehouse
Distributed cloud’s role in the future of work