Alex Pentland‘s research focuses on how interactions between people can reveal information about the individual and group. His innovations have included novel methods to measure and analyze the interactions of groups empirically and correlate the interactions with group performance. Such innovations have led to many spin-off entrepreneurial ventures.
I love this sort of work because it focuses on design for group interactions and builds connections between academic research and social business opportunities.
Two Minds
Part of Pentland’s inspiration comes from the Two Minds model of Kahneman and Simon. We have the Habitual mind, which is fast, parallel, and automatic. It is far older than the Attentive mind, which is slow, serial and rule based. Research has found that for simpler problems, with fewer factors, the Attentive mind is immensely successful. Whereas for complex problems, the Habitual mind is able to make better decisions. We are not aware of all of the decisions and processes of the Habitual mind, and Pentland’s work seeks to reveal its influence. What are factors that may be taken into account by the Habitual mind, which significantly affect individual and group behavior?
Primitive Economies
To understand group dynamics, the researchers go ‘back to the beginning’. Initial research in primitive economies focuses on simplified models of myopic individuals (considering only one interaction at a time) who interact only in local exchanges- approximating primitive society. How do these local exchanges shape the larger-scale structure of society?
What Pentland found is that the global structure is stable and efficient, with a system that truly includes as much opportunity for the rich and the poor (very different from our current system) and the myopic dynamics converge to a stable local exchange network. Maximization of the group and indiviaul in this scenerio are the same, so the invisible hand actually works. Such a stability allows for trust- people will see each other often- which provides the infrastructure for an information economy and language. Individuals can now trade tips and valuable information.
Backchannel Comments
In a paper on collective intelligence and information exchange, researchers find that important factors in predicting the success of a group include commenting, contributions and also backchatter comments. Backchatter are phrases usually less than a second long, but can help indentify how engaged individuals are. Backchatter is identified in this research through analysis of audio recordings, so it is an automated procedure. The same process can be used for an individuals cell phone to determine health factors, such as signaling depression.
While individuals can learn skills to work better in groups, Pentland finds that such skills are difficult to use for manipulation. For example, if people learn how to show more empathy, they will actually become more empathetic.
Informal Communication
In analyzing a call center, the density of informal communication predicts productivity. Informal communications include face to face conversations in passing, as well as emails or phone calls between groups. Such information can be retrieved from badges that can sense where individuals are, and from such data one can determine who is communicating with who.
This was a Harvard Business Review idea of the year. Informal communication patterns account for 40%-60% of group performance. For the example of the call center, in the informal communication conduit, people trade tips on what works for customers. For this call center case, $15 million in savings results from changing the coffee break structure.
In constructing a successful business environment, if there are strong lateral and informal communication channels, information will clear better. (Consider the connection between this informal communication, and the sidewalk ballet of Jane Jacobs. They are the same in their function, in creating healthy productive groups.)
Who You Spend Time With
Another research project outfits a dorm of students with cell phones with sensors that reveal proximity and communication patterns between students. The research takes place during the 2008 presidential election, which allows researchers to see what factors contribute to shifts in political opinion. Surprisingly, the findings suggest that it is not your friends who influence you, but the people you spend time with. A very small proportion of people you spend time with are your friends.
Previous techniques in network research ask individuals to name their top three friends, and a network is constructed. The problem with this sort of work is that it fails to take into account other social influences- it assumes the outcome within the method. Remember the Habitual mind in the Two Minds theory- that factors people are not consciously attentive of can have a large influence in behavior.
Social Networks from Behavior Patterns
Other techniques to construct social networks focus instead on patterns of behavior and locations. In San Francisco for instance, the movement of cabs create a network of the city between locations. The network reveals groups of nodes with few connections to other nodes, the locations of two groups could be right next to each other, but people didn’t mix between them. Based on one’s behavior, the group of nodes they are most likely to move within can be determined. People within groups also have similar characteristics, which can be useful in understanding purchasing patterns, as well as disease spread.
Influencials can be identified within social networks by having high centrality. These individuals have particular behaviors. They are information harvestors, and become the conduit to connect other people in their social network with useful information.
Successful group structures consist of individauls who are able to go out of the group, harvest information, and bring it back to clear with the group. Such a theory is represented in Nathan Eagle’s work on network properties and socioeconomic status, which correlated each individuals ability to clear information within communities and between communities with a socioeconomic indicator of their neighborhood.
Further research shows that to predict the success of a business plan, for instance, one can measure the number of information harvesters in the group.
On Privacy
Monitoring workplace interactions, cell phone calls, and interpersonal interactions delves into serious privacy issues, especially given that such information can be used to influence individual’s behavior. Currently, individuals data are collected and centralized- in many cases the data can be harvested and sold to other companies for purposes including advertising. Pentland’s insight into the issue is that information flows are most stable and efficient in systems where individuals own their own data.
I am excited about further research exploring the benefits of these personal-owned-data systems, where people are in charge of their own data. Such research will encourage applications built on the decentralized infrastructure, and these applications will take advantage of improved information flows. Once a few successful models come out with their documentation, the programming architecture will spread. The norm will follow that individuals control the dissemination of their data.
[...] few weeks ago I wrote about Sandy Pentland’s research in constructing social networks using mobile phone data. The core ideas are the same. Your [...]
[...] No secret that I dig the work of Alex Pentland at the Media Lab, but this next project is awesome. funf an Android app that will help you collect [...]