pentland-amsci

8/7/2019 pentland-amsci

1/9

A reprint from

American Scientistthe magazine of Sigma Xi, The Scientific Research Society

This reprint is provided for personal and noncommercial use. For any other use, please send a request to Permissions,American Scientist, P.O. Box 13975, Research Triangle Park, NC, 27709, U.S.A., or by electronic mail to [email protected] Xi, The Scientific Research Society and other rightsholders


2/9

204 American Scientist, Volume 98

Featur e Articles

2010 Sigma Xi, The Scientific Research Society. Reproductionwith permission only. Contact [email protected].

Decades of research in social psy-chology have captured the sur-prising ability of people to read oneanother. In contexts as different as

evaluating classroom teachers, select-ing job applicants or foretelling jurydecisions, accurate human judgmentscan be made on the basis of extremelythin slices of observational data. Acrossa wide range of studies, psychologistsfind that research subjects on averageaccurately predict outcomes in suchpursuits 70 percent of the time. Thatsuccess rate holds when predictingend results occurring days, weeks oreven months later.

How is this possible? My theory isthat our ability to read each other startswith what are known in biology as hon-est signals. Evolutionary models predictthat social species are likely to develophonest signals, a reliable communica-tion system that serves to coordinate

behavior between individuals. Typical-ly the signals includegestures, expres-sions and calls. Not only are they gen-erally trustworthy cues, honest signals

are also unusual because they triggerchanges in people receiving the signals,changes that are advantageous to thepeople who send them.

Its likely that our human ancestorsused such signalsto coordinate their ac-tions long before sophisticated humanlanguage evolved. A relative newcomerin hominid evolution, language waslikely layered upon older primate sig-naling mechanisms that used social net-work strategies to find resources, makedecisions and coordinate group action.By better understanding their influencetoday, we can shed light on the struc-ture and function of modern social net-works. For instance, honest signals canincrease the energy level within a hunt-ing team or, for that matter, a creativeteam through contagious excitement.They can create a more cohesive familygroup by increasing empathy and trustthrough mimicry signaling.

When we watch a conversation be-tween two people and carefully mea-sure the timing, energy and variabilityof the interaction, we find several ex-amples of honest signals. My researchgroup concentrates on four compo-nents of this human signaling. Mim-icry is the reflexive copying of one

person by another during a conversa-tion, resulting in an unconscious back-and-forth trading of smiles, interjec-tions and head nodding. Activity in-dicates interest and excitement, famil-iar to us from the connection betweenexcitement and activity level in chil-dren. Influence of one person over an-other can be measured by the extentto which one person causes the otherpersons pattern of speaking to matchtheirs. And consistency, or fluidity, ofspeech and movement is perceived by

others as a marker of expertise.

To measure the impact of theseancient social signals, we developedsome very modern tools in order topractice what we call reality mining.

We collect data mostly with custom-designed electronic badges and some-times with smart phones and otherelectronic devices. The instruments un-cover and quantify the role that socialsignaling mechanisms play in every-day decision-making. By examiningthe back and forth of signaling behav-ior in pairs and small groupspayingno attention to words or the identity ofindividualswe can accurately predictoutcomes of speed-dating encounters,

job interviews, even salary negotiationoutcomes. In a wide variety of situa-tions ranging from business manage-ment to first dates to the effects of po-litical opinion, we find that roughly 40percent of variation in outcomes can

be attributed to signaling-based mod-els of social information processing.That is equivalent to some estimates ofthe influence of genetic makeup on in-dividual behavior and is far too large,we believe, to ignore.

Influential CommunicationHonest signals influence critical activi-

ties including negotiation, group deci-sion making and group management.In fact, they are accurate predictors ofhuman behavior. For example, if onemember of a group is happy and bub-

bly, others will tend to become morepositive and excited, an effect known asmood contagion. Moreover, this signal-ing-induced effect on mood serves tolower perceptions of risk within groupsand to increase bonding. Similarly, peo-ple tend to mimic each other automati-cally and unconsciously. Despite being

unconscious, this mimicking behavior

To Signal Is Human

Real-time data mining unmasks the power of imitation, kith and charismain our face-to-face social networks

Alex (Sandy) Pentland

Alex (Sandy) Pentland is the Toshiba Professorof Media, Arts and Sciences at MassachusettsInstitute of Technology. He is a pioneer in com-

putational social science, organizational engi-neering and mobile information systems. Pent-land directs the Human Dynamics Laboratoryand the Media Lab Entrepreneurship Program,which helps develop international applicationsfor new technologies. Among the most citedcomputer scientists in the world, Pentland hascofounded more than a dozen companies. In1997, Newsweek named him one of the 100Americans likely to shape the 21st century. Hismost recent book, Honest Signals: How TheyShape Our World, was published in 2008 byThe MIT Press. Address: Room E15-387, TheMIT Media Lab, 20 Ames St., Cambridge, MA

02139 Email: [email protected]


3/9

2010 MayJune 205www.americanscientist.org 2010 Sigma Xi, The Scientific Research Society. Reproduction

with permission only. Contact [email protected].

has an important effect on participants:It increases how much they empathizewith and trust each other. Not surpris-ingly, negotiations with lots of mimicrytend to be more successful, no matter

which party starts copying the othersgestures first.

Each of these signals likely has rootsin the biology of our nervous system.Mimicry is believed to be related tocortical mirror neurons, a distributed

brain structure that seems to be uniqueto primates and is especially promi-nent in humans. Mirror neurons reactto other peoples actions and providea direct feedback channel betweenpeople. Newborns, for instance, mimictheir parents facial movements despite

their general lack of coordination. Simi-

larly, our activity level is related to thestate of our autonomic nervous sys-tem, an extremely old neural structure.Whenever we need to react more vigor-ouslysay in fight-or-flight situations

or when sexually arousedthis systemincreases our activity level. On the oth-er hand, we tend to be listless and lessreactive when our autonomic nervoussystem is blunted, as during clinicaldepression. The relationship betweenautonomic nervous system functionand activity level is tight enough thatwe have used it to accurately estimatethe severity of depression.

The Habitual and Attentive MindHow do social signals interact with

language? Evolution rarely discards

successful working parts. It general-ly either builds additional structureswhile retaining the old capabilities orsubsumes old structures as elements ofthe new. When our language capabili-

ties began to evolve, our existing sig-naling mechanisms most likely wereincorporated into the new design. Thequestion is then: How has modern hu-man society been shaped by our an-cient signaling mechanisms, and towhat extent do these mechanisms stillgovern our lives?

A partial answer to this question canbe found in the work of psychologistDaniel Kahneman and artificial intelli-gence pioneer Herb Simon, both NobelPrize winners. Each embraced a model

of a human mind with two parts: a

Figure 1. What President Lyndon Johnson was saying to U.S. Senator Richard Russell, a Georgia Democrat, no doubt mattered during what ap-peared to be a tense White House meeting in 1963. But non-verbal aspects of their exchangeso-called social signalswere likely important too.

Observe the way each man holds his arms, mirroring the other. The author and his collaborators use electronic devices to capture the influence ofsuch signaling between individuals and within groups. Their reality-mining research is contributing to a growing body of evidence finding that

signaling plays significant roles in everyday persuasion and decision-making. (Photograph by Yoichi Okamoto, courtesy of the Lyndon BainesJohnson Library and Museum.)


4/9

206 American Scientist, Volume 98 2010 Sigma Xi, The Scientific Research Society. Reproduction


habitual, automatic and largely uncon-scious mind, along with an attentive,reasoning and largely conscious mind.It is likely that the habitual mind repre-sents an older system and contributesmental capabilities similar to thoseheld by early humans: being fast, good

at complex trade-offs and associations, but not very adept at what we thinkof as abstract reasoning. Correspond-ingly, the communication abilities ofthis early human mind would likely belimited to signaling and simple signs.Although the habitual mind is quite

capable of learning new behaviorsthrough experience or mimicry, suchlearning likely is limited to associa-tions among perceptual features.

The ability to go beyond association-based learning may be the key contri-bution that the attentive mind makesto the fitness of our species. There areinherent limitations to learning using

associational mechanisms and Kahne-man has speculated that these limita-tions probably spurred the evolutionof the attentive mind. In addition, thelinguistic capabilities of the attentivemind allow much faster spreading ofnew behaviors among a population.

Of Kith and KinOne of the surprising conclusions fromour studies of social signaling in ev-eryday situations is that the attitudesand actions of peers, rather than logic

or argument, often dominate peoplesbeliefs and actions. It seems that ourforebears understood this intuitivelyand more than a 1,000 years ago hada name for it: kith. Kith and kin is anarchaic phrase that still rings familiar,

but most of us dont know the mean-ing of kith. The word derives fromold English and old German wordsfor knowledge, and it means a moreor less cohesive group with common

beliefs and customs. These are also theroots for couth, which means possess-ing a high degree of sophistication, aswell as its more familiar counterpart,uncouth. Thus, your kith is the circle ofpeers (not just friends) from whom youlearn sophisticated habits of action.

Figure 2. Specialized brain cells likely enable human social signaling. Mirror neurons,believed to reside in the inferior frontal gyrus and in the inferior parietal lobule, may be in-

volved in our tendency to mimic one another. Marco Iacoboni at the University of Californiaat Los Angeles and other neuroscientists argue that mirror neurons allow people to inter-

nally experience the observed actions and feelings of others, which facilitates empathy.

Figure 3. Real-time observation of face-to-face interactions requires the right tools. Custom-designed electronic badges, (top left), measure apersons tone of voice, body posture and gestures, along with proximity and posture relative to other people. Smart phones also can track who

is near whom. The resulting data, depicted in orange above, can be used, along with e-mail communication, depicted in blue, to map exchangeswithin an organization. This example comes from observations in a German bank where employees were developing an advertising campaign. It

exposed limited face-to-face contact by many on the team with customer service staff, who likely had useful insights to contribute to the campaign.

(Photographs courtesy of the author.)

Customer

ServiceSupportSalesDevelopmentManager


5/9



It appears that the English had theright idea 1,000 years ago about howpeople learn. We are ruled by commonsense, the habits our kithmates have incommon. This social learning works bymodifying us through social pressure(usually mediated by social signaling),instead of through critical reasoning.The use of kithmates to form common-

sense habits is another clue to howearly humans may have leveraged thesocial signaling mechanisms to make

better decisions.Imitative learning from kithmates

may be why allowing more time aroundthe water cooler or coffee pot may bethe simplest way to increase workersproductivity. Why? In our studies ofmore than a dozen organizations, wehave found that cohesion among peeremployeeskithmatesis one of thelargest factors in both productivity and

job satisfaction. In these instances, cohe-sion is defined as how connected kith-mates are with each other. That is, dothe people you talk to also talk to oneanother? How tightly woven and inter-connected is your peer network?

In one study in Chicago, we usedelectronic badges to monitor the socialsignaling and conversational patternsof information technology specialists.The badges were fitted with infraredsensors, Bluetooth location measure-ment and accelerometers to measure

body movements, and recorders thatcaptured the pitch and pace of voices.We found that peer-group cohesionwas a central predictor of productivity.In fact, workers whose group cohesionwas in the top third had more than10 percent higher productivity whencompared to the mean. This result un-derscores the extent to which we aresocial animals and that our connec-tion with our peers at a local level isvitally important. With increased co-hesion comes an increase in sharingthings such as tacit knowledge, atti-

tudes, work habits and social support.In other words, much of the importantinformation about how to be success-ful and productive at a job is likely to

be found in the break room.

Tapping Collective IntelligenceBut is it good for people to be confined

by common sensethat is, the beliefs ofthose around them? To answer that, itis important to understand how social-signaling mechanisms help people de-cide when to be guided by kithmates

and when to follow a separate path.

From a theoretical point of view, per-haps the simplest, mosteffective wayto integrate individual preferences intouseful common sense is through an ideamarket. Idea markets resemble voting,

but instead of building on a single voteper person, we allow people to expresstheir expectations of the returns associ-ated with multiple courses of action.For instance, how much food will wefind if we go over the hill? How muchwill we find if we go across the river?And so forth for each alternative. One

can think of these expectations as bets

and use standard probability theory toweight the bets in proportion to theirexpected payoff. In this way we canselect the action that maximizes the ex-pected return and minimizes the risk.

People naturally create idea marketsusing social signaling. Everyone betson each suggested action by signaling alevel of interestsay with the energy intheir voices or their movements. Thengroup members add up the signalingto pick the option with the most posi-tive signaling. This method of decision-

making doesnt require language. In

mimicryreflexive copying of one person by another

activitymovement that indicates

interest and excitement

influencethe extent to which

one person matches

the other's speechpatterns

consistencyfluidity of speechthat is perceived

as expertise

Figure 4. Until very recently, it was impossible to record social signaling in natural settingsover any length of time. But advances in electronics, battery size and computational analysis

make it possible to record signaling in both small and large groups. The author and his col-laborators frequently focus on signaling that predicts cooperation among individuals, includ-

ing qualities such as mimicry, increased activity levels, adoptions of anothers speech patternand consistency, or fluency, in speech. Using data about such signals, social scientists can now

predict outcomes in interactions with surprising success.


6/9



order to pick the winning course of ac-tion, each participant must only signalto the rest of the group how interestedthey are in each alternative and then

be able to read the groups combinedsignaling. Animal-behavior researchsupports the idea that this is what bothsocial bees and apes do when decid-ing about group movements. It also

is similar to the initial-reaction signal-ing seen in business meetings. Thoseums, ahs and hmms so commonin conference rooms, along with theanimated or slack body postures, sug-gest how our modern decision-makingprocesses preserve and leverage theseancient mechanisms.

Still, the challenges individuals facechange dramatically over time. As a con-sequence, social signaling mechanismsmust allow us to quickly select the rightkithmates to help solve the newest prob-

lems. This really is a question of identity:The character of the problem determineswho will be the best kithmates for learn-ing new, effective actions. This poses aproblem for decision making by socialsignaling, however, because when peo-ple are faced with new, important deci-sions, they need to quickly form peergroups that are relevant to the problem.Thus, we need to determine whetheror not people dynamically form prob-lem-defined kith groupings in moderndaily life.

To test this idea, we monitored thesocial signaling and patterns of inter-action for 81 residents in an MIT dor-mitory during the 2008 presidentialelections, giving them smart phonesthat could track who spoke face-to-facewith whom. What we found is thatwhen politics became especially prom-inent, as during a presidential debate,the students shifted their groupingsand began to selectively spend timewith others who shared the same ideo-logical position, excluding those hold-ing opposing opinions. This was not

true of communication channels withno face-to-face contact, such as phonecalls; those remained unchanged, per-haps because they are less effective atconveying social signals.

Further, the extent to which studentsformed a cohesive kith with peoplepossessing similar opinions predictedtheir level of interest in the presiden-tial race, their liberal-conservative bal-ance and even their eventual voting

behavior. For first-year students, themagnitude of this effect was similar

to the effect detected in experiments

evaluating political advertising andmedia exposure. The finding rein-forces the view that when people arefaced with important decisions, theytend to form cohesive, reinforcing peergroups, allowing social signaling to ce-ment shared ideas and attitudes.

Whence Creativity?

We have seen that these signals havea major effect on person-to-person in-teractions and on group behavior, butdo they influence even our most so-phisticated abilities? As it turns out,the humble honeybee has much to tellus about the flow of information in so-cial species. It is common knowledgethat worker bees search for good foodsources and then return to the hive anduse waggle-dance signaling to commu-nicate the distance and direction of thefood source. Less well known, though,

is that bees use this same mechanismas the basis for an elegant approach togroup decision-making.

One of the most important group de-cisions made by a bee colony is whereto locate a nest. Bees seem to use akind of idea market to guide their dis-covery: The colony sends out a smallnumber of scouts to survey the envi-ronment. Returning scouts who havefound promising sites signal their dis-

covery with an intense, active dance.As a result of this social signaling,more scouts are recruited to the bet-ter sites. This cycle of exploration andsocial signaling continues until, even-tually, so many scouts are signalingin favor of the best site that a tippingpoint is reached and the hive moves enmasse. (See, for example, Group Deci-

sion Making in Honey Bee Swarms,American Scientist, MayJune 2006.)

The bees decision-making processhighlights information integration aswell as information discovery. Thesetwo processes are crucial to every or-ganization but each has different re-quirements. The solution suggested bythe bees is to alternate between star-shaped, one-to-many networks that are

best for discovery and a richly connect-ed, many-to-many network that is bestfor integration. Networkswhether

apian or humanthat vary their com-munication structure as needed areable to shape information flow to opti-mize both discovery and integration.

Our studies at MIT have shown thatthis same sort of oscillation betweendiscovery and integration seems to

be characteristic of creative teams ofpeople. In one study we tracked em-ployees in the marketing division ofa German bank, capturing informa-

Figure 5. Social organisms can use different social signals and different patterns of com-munication for different functions. For instance, honeybees will perform a waggle dance to

spread news about a promising new food source for their community. This is an example ofa star-shaped, one-to-many pattern of communication. The dance can also assist with group

decision making. Intense dancing recruits more scouts to examine a site until many scouts endup signaling to each other about one location. When enough scouts signal the same location,

a consensus is reached and the hive moves en masse. This is an example of a richly connected,

many-to-many pattern of communication.

ScottCamazine

/PhotoResearchers,Inc.


7/9



tion about their social signaling duringeach encounter. Analysis of the datashowed that teams charged with creat-ing new marketing campaigns oscillat-ed between two communication pat-terns. In one they placed themselvesin the middle of multiple streams ofcommunication, a centralized commu-nication pattern that is associated withdiscovery. In the second, they engagedin a densely interconnected pattern of

communication where most conversa-tions were with other team members.In contrast, members of productiongroups showed little oscillation, speak-ing mostly to other team members. Asecond study demonstrated that cre-ative teams not only had more varia-tion in the shape of their social com-munication networks, but also that therange of variation in network shapecorrelated with how creatively produc-tive the groups judged themselves to

be. Oscillation in the shape of thesenetworks, in other words, can predictcreative productivity, at least as de-fined by the people in the networks.

Why might this pattern promotegreater creative output? One way tointerpret these findings is that thispattern of oscillation brings new infor-mation to a group for integration intopeoples habitual minds. Because thehabitual mind uses association ratherthan logic, it can more easily makeintuitive leaps and find new, creativeanalogies. It can take the experienceof a new situation, let it soak in for

a while and then produce an array ofanalogous actions.

There is considerable literatureshowing that unconscious cognitionis more effective than conscious cog-nition for complex problem solving.The habitual mind seems to work bestwhen the more logical attentive mindisnt interfering, such as during sleepor when we are turning it over in the

back of our mind. In contrast, the at-tentive mind provides insights into ouractions, helping us detect problems

and work though new plans of action.

Figure 6. Charisma is one quality that helps

individuals bring new ideas to a group.Many influential people have displayed it,

including Mahatma Gandhi (right) the cham-pion of nonviolence and leader of Indias

independence movement. While in Englandin 1931, Gandhi visited workers in the mill

town of Darwen in Lancashire, where Indiasboycott of cotton goods was taking a toll on

local industry. Still, people there responded

warmly to him.

Figure 7. By measuring the expression of qualities associated with charisma, the authorsresearch team was able to predict which business executives would convince a group to sup-

port their business plans during an exercise hosted at MIT. Among the qualities measuredwere active patterns of socialization on the part of speakers, meaning they interacted with

many people; high influence on the patterns of the conversations they participated in; and

high activity levels, which can convey enthusiasm.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

team

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.00

predicted

rejected

chosen

probability

Underwood & Underwood/CORBIS


8/9



The Power of CharismaAlthough using social-signaling mecha-nisms for making decisions appears to

be good for combining action alterna-tives and interests, it is likely not to begood for learning new behaviors. Thisis because the idea market combinationmechanism tends to select only consen-sus views and is unfriendly to new orunusual alternatives. It leads to a verystable, conservative social group. Thisresistance to change raises the impor-tant question of how social-signalingmechanisms might have facilitatedlearning of new action habits from ex-amples outside the community.

One possible mechanism is thephenomenon of charisma. Althoughno one has fully defined charisma,research subjects reliably agree on its

characteristics. In particular, most re-port that charisma is much more than

just word choice or argument. We canunderstand at least a pedestrian sortof charisma if we define it by its op-erational characteristics: an unusualability to convince others to try outa new behavior. Under this defini-tion, people who are good at pitching

business plans, building high-perfor-mance teams and succeeding at simi-lar activities demonstrate the quality.Importantly, many of these charisma

qualities seem to involve social signal-

ing. In our studies, we have observedthat there is a certain style of socialinteractionone that we can identifyquantitatively and automatically bycomputer processing of voice and ges-turethat is highly predictive of suc-cess at influencing others behavior ina variety of situations.

To illustrate, consider our study on business-plan pitches. In that study,a group of rising-star business execu-tives gathered at MIT for an impor-tant task. Each executive presented a

business plan to the group, and thegroup then chose what they concludedwere the best ideas. The executiveswore our badges, which captured theirstyles of social signaling. By analyzingthat signaling, we were able to predictwith a high degree of accuracy which

business plans the executives wouldchoose. Our executives, it seems, were

busy measuring the social content ofthe presentations, quite apart from thespoken, informational part.

To understand why this makessense, consider the situation in moredetail. Imagine you are listening toa business plan pitch on an unfamil-iar topic. Although you dont knowmuch about the subject, the speakerspresentation is fluid and practiced.Also, the speaker is noticeably ener-

getic and clearly excited. Your habit-

ual mind says to itself, ``Well, I maynot know much about this, but sheis clearly expert and she is excited so I guess it must be a good plan.This successful presentation style ischarismatic by our definition, becauseis it effective at convincing people toconsider new behaviors.

Similarly, another recent study from

our research group focused on execu-tives attending an intensive one-weekexecutive education class at MIT wherethe final project, again, was pitching a

business plan. This time we used ourelectronic badges to observe the execu-tives during a mixer on the first eve-ning of the course. We found that theirsocial styles at the mixer were predic-tive of how well their teams businessplans would be perceived at the end ofthe course. The most successful styleis what we call the charismatic con-

nector. These people circulated in thecrowd, practiced intense listening, hadfluid speaking styles and tended todrive conversations with questions.

The more charismatic connectors agiven team had among its members,the better the team was judged dur-ing the business-plan pitch. The rea-son seemed to be that the membersworked together better. In teamswhose social style was dominated bythese charismatic connectors, team dis-cussions were characterized by moreeven-handed turn-taking, high levelsof engagement and higher productiv-ity. These two characteristicscha-risma and connectorusually go to-gether. We have found that the peoplewho have the most consistent andinfluential style of speaking are alsothe people who are the greatest con-nectors. People whose social networkscross many different groups are ex-actly those people who display a char-ismatic style of interaction.

Under the Signaling Influence

Our research suggests that peoples behavior is much more a function oftheir social network than generallyimagined. Humans truly are social an-imals, and individuals are best likenedto musicians in a jazz quartet, forminga web of unconscious reactions tunedto exactly complement the others inthe group. These various studies frommy research group all serve to dem-onstrate that this immersion of selfin the surrounding social network isthe typical human condition, rather

than an isolated example found in ex-

Figure 8. At an MIT Media Lab event for its corporate sponsors in Tokyo in 2008, graduatestudent Ben Waber explains a social-network visualization produced by tracking people at

the event who were wearing sociometric badges. Such patterns of interaction now can beobserved in real time in large and small groups. The technology offers promise but also intro-

duces ethical issues that must be addressed regarding ownership and fair use of the resultingdata. (Photograph courtesy of the author.)


9/9


with permission only Contact perms@amsci org

ceptional circumstances. Our ancientreflexes for unconscious social coordi-nation fuse us together into problem-oriented peer groupsour kith. Andthose groups strongly influence ouractions every day.

What practical conclusions can wedraw from this? These results tell usthat individuals should consciously

work toward having a cohesive, en-gaged set of kithmates, helping themadopt more effective habits of action.There is solid evidence that peopleinvolved with such kithmates are not

just more productive and creative.They are also happier, more resilientand more satisfied. And how can onego about collecting this set of valuablemates? The charismatic connectorstyle of signaling we have uncoveredmay be the single most importantfactor in promoting the success of

group activities, by creating a conta-gious positive mood, increasing trustand encouraging more even, sociallyaware participation. It may be time to

begin training people to become morelike these connectors.

Reality mining offers promising in-sights such as these because its largedata sets reveal social patterns thatonce were invisible, showing us pic-tures of hundreds, even thousands, of

people working together. Of coursethis method raises ethical issues thatmust be addressed. Such data alsopose a potential threat to individualprivacy. Because of that, it is impor-tant that individuals rather than cor-porations own the data used for realitymining. To my mind, that would placecontrol of the datas use with the ob-

served individuals, where it belongs.And it would also allow the individualowners to derive personal value fromthis important knowledge source aswe strive to understand how we can

better work together.

Acknowledgments:I would like to thank the students and fac-ulty who collaborated on the research dis-cussed in this article. Tanzeem Choudhury,Taemie Kim, Daniel Olguin Olguin andBen Waber helped create the sociometric

badges and conduct the related experi-ments. Nathan Eagle, Anmol Madan andHarvard University Professor David Lazerhelped create and conduct the smart-phoneexperiments. New York University As-sistant Professor Sinan Aral, MIT Profes-sor Erik Brynjolfsson, Dr. Tracy Heibeck,Dr. Peter Gloor and Lynn Wu helped withthe analysis of the experiments. Research

papers discussing the resulting researchcan be found at http://hd.media.mit.edu.

BibliographyAmbady, N., and R. Rosenthal. 1992. Thin

slices of expressive behavior as predictorsof interpersonal consequences: A meta-analysis. Psychological Bulletin 111:256274.

Barsade, S. 2002. The ripple effect: Emotionalcontagion and its influence on group be-havior. Administrative Science Quarterly47:644675.

Buchanan, M. 2009. Secret signals: Does a

primitive, non-linguistic type of commu-nication drive peoples interactions? Nature457:528530.

Buchanan, M. 2007. The science of subtle sig-nals. strategy+business 48:110.

Couzin, I. D. 2009. Collective cognition inanimal groups. Trends in Cognitive Sciences13:3643.

Iacoboni, M., and J. C. Mazziotta. 2007. Mirrorneuron system: basic findings and clinicalapplications.Annals of Neurology 62:213218.

Pentland, A., with T. Heibeck. 2008.Honest Sig-nals: How They Shape Our World. Cambridge,Mass.: MIT Press.

Zahavi, A., and A. Zahavi. 1997. The HandicapPrinciple: A Missing Piece of Darwins Puzzle.Oxford: Oxford University Press.

For relevant Web links, consult this

issue ofAmerican Scientist Online:

http://www.americanscientist.org/

issues/id.84/past.aspx

Documents

pentland-amsci