Hypertext '08: Bernardo Huberman, Social Dynamics in the Age of the Web

Today’s keynote:

Brughel painting showing the social
dynamics of a village festival.

Grounded the talk with a presentation
of statistics on user-generated content (Facebook, MySpace, etc.),
noting that whether those users are interacting with one another is
another question. Noted that his research is observational rather
than experimental, and that he won’t be able to go into detail,
because rather his overservations will focus on what we can learn
from the large number of users.

Noted that until recently, most content
was created by a few and consumed by many. Noted the “remarkable
inversion” in the creation of content.

Wikipedia – great example of
something created from the bottom up. Invoked Nature’s experiment
comparing Wikipedia articles with print encyclopedias.

Interested in the overall quality of
Wikipedia as a reliable source of information. Study how Wikipedia is
produced and generated. Correlation between the number of edits and
its quality.  With D. Wilkinson in First Monday, published on clear correlation between the number of edits and the quality of the article.

Transitioned to the importance of the scarcity of attention. Information that people used to pay for is now freely available, but what is rare is our attention.

Not psychological attention, but social attention. Citing a source so that a reader can attend to it.

How does the phenomenon of attention pay in the importance of information? What is the role of novelty? How do you maximize the amount of value you get from a website given the fact that we have a limited amount of attention.

Noted that very few people are publishing “real scholarly work” on this subject.

In areas of low information density, attention is not a problem. (Showed a desert scene with a stopsign… “information poor environment”).

A picture of Howard Rheingold in Tokyo, staring up at the insanely detailed advertising displays.  In an information rich environment, attention is valuable, ephemeral, and difficult to obtain.

Two ways to gain the attention of a group.  One: to broadcast.  At least initially it gets the attention of the people.  Another way is propagating the information virally.

Included graphic from a study on Amazon.com recommendations.  Most things do not propagate very far, but every so often you have a long chain.  Propagation of recommendations of a medical book has many shallow nodes, but the network for a Japanese graphic novel has far fewer nodes, each of which gets much more activity.

People are more likely to buy a DVD recommended by many people than they are to buy a book.  Two recommendations leads to a spike in recommendations for book purchases, but after that it drops off.  The DVD graph rises steadily with more recommendations, with the value approaching a higher figure.

As novelty fades, we pay less attention and search for more.  Novelty interactes with collective attention in a highly nonlinear but predictable way.

Prediction 1: the attention among all items is distributed in a log-normal way.
Prediction 2: attention decays in time as a stretched exponential (long tail, invoked radioactive half-life of about 69 minutes, roughly the amount of time a news story is on the front page of a website).

Suggested dynamically reconfiguring a website based on the number of hits the stories receive.

Noted that, while we do attend to novelty, we are also social beings, so we will attend to something simply because others are attending to it.  How does popularity affect attention?

[I note that Huberman defines “novelty” as “recency” rather than anything to do with the content. It seems then that it would be easy for a computer to notice when a link has unusually intersting or unusually boring content when the measured clickthrough rate differs from the predicted natural decay… I suppose that’s what he means by the “popularity” of an item.]

Public opinions are another attention structure. How do opinions form and evolve?  Noted psychological study on group deliberation and group polarization, demonstrating that people tend to move towards extreme views. On the web, it is costly (in terms of time) to write a review. Conjecture — people will only write a review if their opinion differs from the dominant opinion.  There’s a softening of reviews over time.

Voter’s paradox — why do people bother to vote when your individual vote is so insignificant in a country of millions of voters?

In larger groups, individual contributions have to be larger in order to have a significant effect on the end result.

When it is costless to express an opinion, polarization occurs. When it is costly to express an opinion, a softening takes place. [What implications does this have for multiple-choice or short-answer quizzes, and papers?]

The group opinion does not change, but the selection process governing who expresses an opinion does change.

IMDB — I think I missed something here; did he compare the effects of assigning a rating (a low-cost operation) with writing a review (a high-cost operation)?  He rushed through this point in order to end on time

Q and A.

Q How does free will enter into the picture?

A. There’s a huge difference between people and the objects that physicists study. Rocks falling and photons behaving can be described by an equation. People, unlike rocks, molecules and atoms, take into account the future when they make each action. “I’m not here as a result of Brownian motion.”   Using physics-like metaphors to discuss phase transformation is just that — a metaphor.  Not forcing the world of people to be like something that we know from physics.  Calls for “a certain amount of humility” when facing the study of human actions.

Q: Clarification of “novelty” — is it just chronology?

A: An idea is novel if it is not repeated; if you insert something in a familiar piece of music, people will attend to it.

Q: What implications for diversity of opinion?

A: Paris Hilton keeps finding novel ways of attracting attention, so he gives her credit for a certain amount of intelligence. There’s a competitive component — attention research actually can be liberating because distinct and diverse opinions attract attention.  The fact that whole world is collapsing to the point where we all see the same information is a different phenomenon.

Q. In the current issue of Atlantic, there’s an article about traffic management of automobiles in UK vs US.  The article says our road system is oversaturated with too many signs.

A: That theory is about a collective attention phenomenon, but not about perceptual attention to individual elements.

Q: Invitation to speculate about tension between the familiar and the new, and the “sweet spot” that attracts… movie genres is an example of people liking things they already know.

A: We haven’t done much on that. A personal home page or newspaper is familiar, but has to have content that changes.

Q: Note old comparative psychology study… what kind of animal would a rat want to play with most? Rats liked to play with rats best, but they also liked the human hand. Hypothesis — moderate unpredictability was good. The guinea pig always had the same behavior (hiding), but the hand was unpredictable. (The speaker noted that this was a comment not a question.)