Sunday, November 25, 2007

Social Cognition Made our Heads Explode

Ross Mayfield from Social Text gave a talk recently as part of ASC's new PARC Forum series and he mentioned a bunch of “Numbers” relevant to network effects (e.g., Metcalf's Number). He mentioned Dunbar's Number, which is one of my favorites because it highlights how social awareness and reasoning improves our fitness, and may explain why the human head exploded (at least on the scale of evolutionary time). The September 7 issue of Science included a whole section of social cognition, included an article by Dunbar on his recent work in the area.

So what is Dunbar's number? Well, it's 150. That's the theoretical limit of the number of people that you can “know socially” in the sense that you know them as individuals and know something about their relations to one another (and you), and the reason why it's an interesting number is that it seems to be related to the size of that melon sitting on your shoulders.

As an undergraduate I joint-majored in anthropology and psychology, so I'm generally interested in brains, but even more so because it has been a bit of a mystery as to why ours (homo sapiens) got so big. Second only to the heart, the brain consumes a massive amount of our energy intake, and those big heads make childbirth more problematic than our primate relatives. Most people probably would tend to believe that bigger brains mean greater intellectual capacity, and that somehow that provides a a substantial increase in out ability to, say, forage, to offset the costs of big brains. But as Dunbar notes in his Science article, there is no reason for a chimp to have a brain so much bigger than squirrel when they basically solve the same foraging problems.

It turns out that neocortext ratio for various species is strongly correlated with the average size of the social group for members of that species (and for humans that number is 150 at the limit), Recent evidence suggests that it is more specifically correlated with pairbonding. But even more importantly, it appears that increasing sociality increases reproductive success. So social cognition increases fitness.

This frames all sorts of interesting questions about social technologies. James Surowieki's “The Wisdom of Crowds” argues that one can get more accurate and “wiser” judgments from large-scale aggregate behavior in things like electronic markets or voting systems. But if was just the size of the herd that mattered then cows would be wizards. The claims made by the comparative biologists studying brain size is that our ability to maintain awareness and reason about complex social relations buys us something important. So, assuming that things like Twitter and Facebook (or the Wikidashboard) and the rest give us greater social awareness and reasoning--what exactly does it buy us?

Monday, November 12, 2007

How social tagging appears to affect human memory?

Three weeks ago, I was at the ASIST 2007 Annual conference in Milwaukee, which had a special theme on Social Computing and Information Science. During one of the panels on Social Tagging, a question was raised on how tagging really affects memory and retrieval. I mentioned that the ASC group here at PARC has been doing some experiments on this, and briefly talked about the results, and many attendees at the conference (over 10 people) had asked for the pre-print, so here I'm blogging about it.

Raluca Budiu, who is a post-doc working in our group, has conducted some very interesting research with us on how tagging appears to affect human information processing. She studied two techniques for producing tags: (1) the traditional type-to-tag interface of typing keywords into a free-form textbox after reading a passage or article; (2) a PARC-developed click2tag interface that allows users to click on keywords in the paragraph to tag the content.

The experiment consisted of 20 subjects and 24 passages in a within-subject design. Participants had to first study passages and tag them, and then they performed memory tests on what they had actually read and tagged. The memory tasks were that, after tagging the content, they have to either (a) freely recall and type as many facts from the passages as possible; or (b) answer 6 true/false sentences in a recognition task.

As reported in the paper, the results suggest that:

  • In the type-to-tag condition, users appears to elaborate what they have just read, and re-encoded the knowledge with keywords that might be helpful for later use. This appears to help the free-recall task (a) above. In other words, users seem to end up with a top-down process and induces them to schematize what they have learned.

  • While in the click2tag condition, users appears to re-read the passages to pick out keywords from the sentences, and this appears to help them in their recognition tasks (b) above. In other words, users seem to use a bottom-up process that simply picked out the most important keywords from the passage.

Click here to download the technical report and pre-print (the highlights in the paper are mine).

Monday, November 5, 2007

PARC and ASC to host a special Web2.0 speaker series

Starting next week on Nov 15th, PARC and ASC is hosting a special Web2.0 speaker series as part of our normal PARC Forum talks. We have a very exciting and evolving list of speakers. This series will run for several months throughout the winter season, right here at PARC.

The first week we will have Ross Mayfield from SocialText speaking to us about Wikis in the Enterprise. Following that, we have speakers from various places ranging from startups, industrial research labs, and academia. Topics will range from social practices of online communities, startup excitements, mashup techniques, and academic studies. The talks will be recorded and published on the web. Here is the preliminary announcement:

PARC Forum -- special speaker series on going beyond web 2.0
Thursdays at 4 pm, Palo Alto, CA

more info at

Upcoming confirmed speakers:

*November 15 -- Ross Mayfield, SocialText
*November 29 -- Garrett Camp, Stumble Upon
*December 6 -- Charlene Li, Forrester Research
*December 13 -- Guy Kawasaki, Truemors, Garage Ventures
*January 10 -- Bernardo Huberman, HP Labs
*January 17 -- Chris Anderson, Long Tail
*February 7 -- Premal Shah,
*February 21 -- Andrew Mc Afee, Harvard Business School
*March 20 -- Lisa Petrides, Amee Evans; OER Commons
*March 27 -- Ed Chi, PARC Augmented Social Cognition

To subscribe to future PARC Forum announcements and/or our e-newsletter,
please visit: