US scientists are using cellphones to decipher the human behaviour

MOBILE phone technology that is enabling millions of people connect with each other every day is giving scientists some new insights into how humans communicate with friends, connect with strangers and get influenced by others.

The research could result in new tools to let people improve their communication skills. Soon you may know if you are persuasive enough while making a sales pitch or are handling job and salary negotiation talk the right way or are dominating a group discussion.

On a larger scale, new tools may let scientists predict behaviour of people in a given setting and even spread of contagious diseases or political opinion prior to elections. Scientists at the Media Laboratory of the Massachusetts Institute of Technology (MIT) are excited over the initial results they have got using specially designed gizmo called ‘sociometer’, a highly complex software to collect data and mathematical models to map this data.

So far, researchers have depended on the traditional method of a questionnaire to collect data about human behaviour. But such data is biased because we don’t tend to report everything we do correctly.

Dr Alex Pentland, who is leading this research at the Media Lab, has pioneered the concept of ‘reality mining’ — processing of our conversational and movement data collected through mobile phones and other wearable devices developed at the laboratory. Pentland and his research team have already used some of these techniques in actual settings such as call centres, hospitals and banks to decipher how people communicate in organisations. In the call centre experiment, researchers used an algorithm to predict whether a call would result in a sale within a few seconds of data. Successful operators spoke little and listened more.

When they spoke, their voices fluctuated strongly in amplitude and pitch. A call centre in UK is now using the technique while recruiting new workers.

In a larger experiment currently on in MIT campus, Anmol Madan — a doctoral student of Pentland — is studying how ideas diffuse and people get influenced in a close- knit community.

The study includes 65 undergraduate residents living in one university dormitory.

Each participant is using a Windows Mobile smartphone for a year, which can scan for Bluetooth wireless devices in proximity as well as wifi access points. This data helps identify nodes and edges in the social network and tell if a cluster of users tends to visit similar locations frequently. The music player allows participants to play, share, rate and search through the music library. In addition, participants input data about smoking behaviour, exercise, fitness and diet. All phone call logs and SMS logs are captured. Only tone, length of speaking duration and time when a call is made are recorded and not voice.

“ We can predict relationships, we can predict who are close friends based on what time a person calls another person, what time you meet him or her, etc. And we have got it right with over 90 per cent accuracy,” pointed out Madan. “ We have noted that people who eat salads and veggies tend to flock together though they may not know it”. Madan said the information being collected through ‘ socially aware mobile phones’ had higher accuracy than what people tell when asked to recall. “ It is more accurate than you are. It exposes the biases that we have when we think about someone.

Cell phones can predict better than you can,” Madan said.

Sociometric badge or sociometre is similar to badges which employees wear in large organisations to identify themselves to gain access to certain locations.

When used in a meeting, sociometric badges can help users better understand the flow of the meeting and possibly improve their participation.

Often people do not realise that they are dominating a conversation while other people may need reminders to participate more in the conversation, explained Taemie J. Kim, another PhD student and a member of Pentland team.

The sociometric badge provides real- time voice analysis on whether a user is speaking or not and aggregated data allow researchers to know who are dominant speakers and who are in the periphery. The badge can also give real- time feedback to the user based on their speaking behaviour. “ In our experimental scenario, each person in the meeting wears a badge and carries a mobile phone. Every five seconds, the badge broadcasts its unique identifier number and signal that tells whether the wearer is speaking or not. The movement data is transferred in a similar manner”, said Kim.

“ The biggest challenge of our approach is privacy concerns which we are working to tackle,” added Kim.

dineshc.sharma@mailtoday.in