Judith Donath: When do we care if you are human?

Judith Donath of the Socialble Media Group at MIT’s Media Lab, is one of this year’s new fellows at the Berkman Center. Her lunch talk today is titled “Designing Society” and looks at the ways that communications technology can be designed to enhance human interaction and make communication more expressive.

Donath’s current work is on signaling theory in human communication – she has a forthcoming book at MIT press called “Signals, Truth and Design.” She shows a table of contents, which includes topics like “signals, cues and meaning”, “deception” and “gossip, ratings and reputation”. She explains that most critical social information isn’t visible. “Do you like my gift?” “Would you be a good partner and care for my offspring?” It’s hard to see the answers to these questions, so we read social signals instead. A key question is how reliable these signals are as a way of detecting what lies beneath. This turns into a question about economics: “What are the costs of being deceptive?”

When we raise children, a good part of socialization is teaching children not to tell the truth. A three-year old wants to tell you that you look funny, that he hated Grandma’s gift, that he thinks a stranger is ugly. A good part of parenting is telling children when it’s appropriate to be truthful and when you need to be deceptive.

Much of her talk focuses on projects that her students have built at the Media Lab and the questions that they’ve raised. One set of projects focuses on improving conversation in online spaces. A project called Chat Circles was a response to the first generation of online chat systems that included avatars, like The Palace. She argues that “putting up a picture of a sofa as a way to encourage people to talk seems a bit heavy handed.” This is an example of a tension between abstraction and legibility – it’s possible to recreate very accurate, very legible spaces in virtual worlds, but it tends to miss some of the power of abstraction that can come from the digital realm.

In Chat Circles, each user is represented by a colored circle. The circle expands when you’re speaking and contracts when you’re silent. The words of people near you in the virtual room are visible – if you’re further away, you can see a person speaking, but can’t comprehend the words. The interface therefore uses space in a way that makes intuitive sense – if someone bothers you, move away and you won’t hear what he’s saying.

Donath observes that people have an odd tendency to fill physical needs in virtual space, despite the fact that these needs are non-existent. Why provide chairs in Second Life? Avatars don’t get tired and want to sit down. Yet people build meeting rooms with chairs, conference tables and fluorescent lighting – the goal is to signify a meeting space by providing acoutrements that aren’t strictly necessary. (One audience member observes that she gets tired with her avatar standing in Second Life and feels better when it sits down. John Clippinger suggests that this may be due to mirror neurons, which percieve someone standing and evoke a feeling of strain or exhaustion in the viewer.)

Rather than recreate a meeting space by rendering chairs, Donath is interested in building a space that shows subtle cues. In a real-worl meeting, we can see if people are paying attention, nodding off, or looking away from the speaker. A new space she and students are working on allows users to show agreement and disagreement with an idea being discussed by moving to different parts of a room. The conversation is archived in small balloons above the speaker’s head. This makes it easy to review the conversation, but also to see who’s been dominant and who’s been silent. (In the slide she shows, the people disagreeing with an idea are more outspoken, a phenomenon that I suspect is not uncommon.)

Another piece focused on making conversation visible is RadioActive, by her student Aaron Zinman. The project, a master’s thesis, was a response to the problem of people making phonecalls because they were bored, not because they had something to say. Zinman responded by designing a phone-based, audio interface to something a bit like usenet newsgroups – persistent, asyncronous threaded conversations in audio. The system tracked the behavior of users encountering the audio, and allowed for simple annotations, including whether people listened to an entire post or gave up halfway. A visual system used three factors – the size of a colored circle, the size of a ring around the circle, and color saturation to communicate message length, perceived value of the contribution and age of the post, making it very simple to evaluate multiple axes of a post in a single glance.

Knowing Berkman’s fascination with poker – see our founder Charles Nesson talking with Stephen Colbert about poker – she shows us a project by Scott Golder. Golder was interested in making online poker more like real-world poker, so his program communicates a number of factors usually invisible in online poker games. It shows you when another player looks at her cards, and forces people to bet by setting sliders, which makes it possible to see how certain their bet is. It adds information to the equation, letting people’s actions be visible to others… which can be a technique for bluffing as well.

Donath leaves this section of the talk with the question, “How much deception do we desire?” In partial answer to this question, she tells us about an experiment to enhance online social interactions with gloves that measure galvanic skin response. These are very crude tools for measuring human arousal – they don’t distinguish between excitement, anger and stimulation, but they do give a basic sense of whether something has captured someone’s emotional mind. Adding this sort of feedback to chat situations revealed high group interest when someone told jokes, stimulating everyone in the room. But she’s not sure it would be a good addition to all online systems – “we haven’t really evolved to use hand sweat as a form of communication”, and we’re bad at controlling inappropriate responses, or using this form of communication with care and restraint.

To show us the world of social media visualizations, Donath begins with a visualiztion that she thinks was a failure – though a very popular one – Rebecca Xiong’s visualization of threaded discussions as flowers. A message board is represented as a garden of flowers. Each flower’s stem length is determined by the length of that person’s involvement in the conversation. The bloom of the flower includes a petal for each post, color-coded for whether the posts are original or responses. It’s a beautiful visualization, clearly a very compelling metaphor. But the boards she was tracking included some very hateful neo-nazi speech, and the most active posters were some of the most virulent authors… though they were represented by the prettiest flowers. From this, Donath observes, “the semantics should be correct in a visualization.”

ThemeMail, by Scott Golder and Fernanda Viegas, was an attempt to solve the semantics problem. The project helped users visualize their stored email. Most users keep a huge set of mail on hand – ask them why, and they’ll tell you they might need it. Most of us don’t. But that mail is a fascinating representation of our relationships and seems to have a great deal of sentiment for most of us. The ThemeMail visualization showed portraits of your correspondents based on histograms of word usage in your exchange with that person. The researchers found that they could often identify who a person was just based on the histogram. And many users chose to print out their histograms and keep them as an art object that embodied their personal relationships, almost like photo snapshots of friends. Donath asks, “How do we handle the end of ephemerality? What do we do when history no longer fades away?”

The rise of complex new online social tools gives researchers a whole new set of spaces to study. Aaron Zinman has been studying spam in social networks. In a space like MySpace, it’s possible for people who don’t know each other to “friend” one another. Sometimes this behavior is harmless – often, it’s the precursor to commercial spam. It’s possible for humans to identify spammers by certain subtle patterns – someone who has 100 friends, but all friends were made on the same day, is likely a spammer. To tell real friends from spammers, it may be neccesary to look at more “expensive” forms of interaction – it’s cheap to “friend” someone, but more expensive to leave comments on their profile. When analyzing these networks, you need to consider, “what is the significance of a connection?”

Robin Dunbar, a British academic, has offered an interesting theory in his book “Grooming, Gossip and the Evolution of Language.” He observes that apes maintain social ties through grooming, which is time consuming, and keeps groups small, as only a small band can reinforce ties by grooming each other. With language – and specifically with gossip – humans can maintain ties within a much larger band. One of Donath’s interests is how digital spaces can make socialization more efficient, allowing people to meaningfully expand their pool of friends.

She suggests that technology can serve as a social catalyst, giving people a group experience that can cause people to engage with one another. Increasingly, people in public spaces have their attention on personal-sized screens – technology can force people out of that private space and into common space. She shows off Agoraphone, a project by Kelly Dobson, which allows users to phone into a number and speak, via loudspeaker, to a public passing by. In the six weeks the piece was present at MIT, one user started scheduling a weekly call-in radio show, broadcast over the speaker. Other users called it while they could watch it, using mobile phones. The technology became a very effective icebreaker to encourage interaction in public spaces.

Other projects designed to serve as social catalysts include “Chit Chat Club”, which turns empty chairs into avatars – of a sort – for people participating in conversations remotely, displaying text or images on monitors embedded in the chair. She closes by showing us Foosball Land, a foosball game that’s played simultaneously by real-world players and players in Second Life. The goal of the designer was to create a game that requires interaction from the physical and the virtual to achieve victory.

She closes with yet another provocative question: “When do we care if you are human?”

This entry was posted in Berkman, Geekery, Media. Bookmark the permalink.

3 Responses to Judith Donath: When do we care if you are human?

  1. ““signals, cues and meaning”, “deception” and “gossip, ratings and reputation”. She explains that most critical social information isn’t visible. “Do you like my gift?” “Would you be a good partner and care for my offspring?” It’s hard to see the answers to these questions, so we read social signals instead.”

    It’d be to see this in a cross-cultural framework. It could give a fresh perspective on Chinese notions of face and trust, concrete ways of distinguishing it from social signals in, say, American society.

  2. Atanu Dey says:


    Dunbar’s number (wikipedia) comes to mind. Also “What is the Monkeysphere?”


  3. Pingback: SearchCap: The Day In Search, February 6, 2008 at World Flat

Comments are closed.