December 4, 2009
Human Signaling: Competition and Cooperation in Everyday Communication

A couple weeks ago, I sat in on a lecture by Judith Donath, who is an Assistant Professor of Media Arts & Sciences at MIT. She also works in the MIT Media Lab, where she founded the Social Media Group, and is a Fellow at the Berkman Center for Internet & Society at Harvard Law School.

Her talk, entitled Human Signaling: Competition and Cooperation in Everyday Communication, was one of a weekly seminar held by the Cooperation group at the Berkman Center. The talk introduced concepts of signaling, which draws from theoretical biology, and connected them to cultural practices of behavior, language, and even fashion. Judith's abstract explains more, below:

It can be quite beneficial to deceive - to indicate that one is smarter, nicer, or possessed of better genes than is actually the case. Yet if deception was rampant, communication would cease to function. Signaling theory provides an economic model that shows how enough honesty is maintained to keep communication working.

This model, which was developed in the field of theoretical biology, has also been used to understand a variety of human behaviors, e.g. cooperative hunting and religious rituals. Yet human communication differs significantly from animal signaling: we can rely on cheap
conventional signals because we have coordinated sanctioning; our cultural evolution allows for rapidly changing signal vocabulary; we can imagine other minds and deliberately manipulate impressions; we have internalized morality; and we are very creative - there are no signals, no matter how seemingly reliable, that we will not attempt to fake.

In this talk I will introduce signaling theory and describe how to adapt it for modeling human communication. I will then discuss two examples of applying this theory - first to analyze the social phenomenon of fashion (in clothing, arts, academic concentrations) and second to design new interfaces for online communication.

Unfortunately, while the presentation applied so well to many of the ideas floating around here at the Consortium, there's currently no recording of it. The best I could do for you wonderful readers was embedding Judith's similar talk from 2007 that she delivered at Google:

However, if you'd like more insight into the 2009 version of the talk, I've appended my personal notes after the jump below. Enjoy!

Judith Donath,


receiver benefits only when the signal is honest

signaler benefits only when receiver believes signal

but: communication must be beneficial to both signaler and receiver

(cooperation as a signal of status)

mathematical equations to create evolutionarily stable systems

Q: how narrative are all of the signaling elements in biology? Eg., butterfly colors: to what degree do they tell us information?

cue: anything used to infer some quality or predict a behavior
signal: cues intended or evolved for communication (intentional, unlike cues)

fashion: as signal or wealth; as signal of taste

receiver costs: if you want to listen to signals, how many consequences result? eg., loss of time spent eating, higher chance of encountering predators

ideal signal: costly to produce, free to assess

conventional signals: honesty via social sanctions
sparrows example: can hack signals, but eventually honesty wins out (when false discrepancy is discovered)

human examples:
- police badge; false wearing: not worth cost opportunity in being dishonest
- resume: judging the truthfulness of your self-presentation
- phrasal: "I'm /so/ glad to see you!"

A signal will be reliable if it is beneficial to produce honestly and unaffordably costly to produce dishonestly.

If the signaler gains a benefit (eg., pleasure) from producing the signal, it can take away from the honesty of it.

Receiver benefits:
Signaling benefits: improved knowledge
Functional benefits: enjoyment

v. Costs:
Deception/ of being deceived

Signal ambiguity: amplifiers, attenuators, & signals of need

Peacocks: tail display, v. vocalization; presentation: observation dependent


Ellison, N., R. Heino, and J. Gibbs. 2006. Managing impressions online; self-presentation processes in the online dating environment. Journal of .....?

Online dating: dates --> restaurants: Ability to signal assessment of the other, versus how much you want to say about who you are;

Risk v. reward: 2 minutes to check message, v. 15 minutes to have coffee --> adds up to high value of dinner date

Signals: fancy restaurant: value of wealth v. devotion (eg., poor person who saves up for expensive food for date)


Signals: as indicators of time (spent doing X)
Status displayed in online games: in terms of time spent playing games

Evolution of religion
or other cultural events
If you understand the cultural indicators/references, the reception is easier/more pleasurable.
Genuine member of the community: pays lower cost, v. someone who goes just to look cool (dishonest participation?).


Human signaling is different...?

- Creativity (trying to get around/hack social cues/norms)
Tigers jumping v. cosmetic surgery.

Pale v. tan; around beginning of 20th century; royalty stayed inside --> then poor went into factories (while wealthy went to tropical climates) --> tanning beds (has it lost its meaning though? --> deception of having had a relaxing time, since tanning beds are dangerous and boring)

* What looks nice is highly culturally dependent.

- Cultural sanctions/conventions
Eg., accents, or linguistic preferences/trends: show elements of origins/personal history
Value of making dishonest signals: not very high when sanctions are in place (eg., "My child is an honors student at X school.")

- Reputation rewards and sactions
Eg., eBay feedback (***ratings of reputation & reputation are not the same thing)

- Cultural evolution
Some signals don't always mean the same thing to the same people, (especially generationally)

- Internalized morality: in relation to dishonesty/lying


In clothing, technology, architecture, discourse, linguistics (eg., slang).

Fashion: signal about your knowledge (what clothes to wear, what trends you participate in): your cultural affiliation within a subcultural group.

Eg., how people move from one social networking site to another: type of fashion.

* Fashion is accelerating.
Eg., music tastes: Top 10: songs went from 30+ weeks on billboards to only 1-3 weeks

* Relationship between how fast information moves and how fast fashion accelerates.

(Imitation differentiation model.)

Fashion: risk in adopting what is new.
Because: at the forefront: everyone might not eventually adopt it.

Initially, Bluetooth wireless phone microphones: used to be shiny & blue; now that every cabdriver has one, cultural value has changed drastically.


Rogers, "The Diffusion of Innovation."
Value is orthoginal to innovation.
Technology: follows innovation curve, but value as signal decreases over time.

Grant McCracken: on goth culture
How trends in fashion change based on hierarchy: not by class, but by authenticity.
Also: "Club Cultures" (book): how songs in dance clubs change based on hierarchies

Return from aside:

Market for avant-garde art: taking a risk for being at the beginning.
Art: limited amount of paintings: very risky to choose certain pieces that might not return value/investment.
But: there's the potential to gain entrance into a culture by participating in the market.

* Understanding signal theory: a lot of what we interact with and do has a high level of communicative qualities to them.

High risk: because it might be anti-utilitarian.

Fashion; in Second LIfe: in terms of risk.
Institutions & businesses: had to question championing the space.
Have to understand trends to inform clients: silly, or valuable?

Language: Tower of Babel
LOLcats: picking up on the dialogue of the pictures: could pick up on trends
eg., "Invisible X": becomes an entire phrase with meaning: then arrive at play: Invisible Hand
creates instant subculture of people that understand the references

Serious note: devotion to change is costly
eg., discarded cell phones: unsustainable level of fashion consumption
People are markedly uncomfortable carrying around objects that signify cultural meaning (eg., old laptop --> outdated style)


Online representation of human behavior
One thing that people like about Second Life: it shows your face
Face shape: promotes certain levels of trustworthiness and knowledge
So, between 3 sliders, you can change from trustworthy to probably not trustworthy at all: still difficult to measure identity online

Facial recognition & trustworthiness
Morph your own face into a presidential candidate's: makes them seem more trustworthy (even if it's GW Bush's)


Judith's personal interest:

online identity: don't use faces, but data portraits
small snapshots of data: impress quickly on others what sort of person you are
eg., Twitter profile: left side: what user has said recently; right side: what user has said over time

[Q: Facebook: able to tie identity to things that are more 'real' and therefore more costly (?)]

words that people use: the signal of self-presentation
therefore: assessment: easier for the receiver, since the signal is already made publicly over time by the user

"Personas" project: takes Google information and creates a portraiture of you by attempting to categorize your words