Continuing with Day 2 of the 3rd International AAAI Conference on Weblogs and Social Media in San Jose this week.

For tweets from other #AAAI / #ICWSM attendees, check the Summize results here. Follow Jessica’s tweets at http://twitter.com/jesstsai

Tue May 19 9:01:57    Speaking: Duncan Watts, Yahoo! Research, “Using the Web to do Social Science.”

Tue May 19 13:38:15   ”How does [system] design impact online community culture?” Panel Discussion

Tue May 19 15:05:13   Now Speaking: Tad Hogg, HP Labs, “Stochastic Models of User-Contributory Websites”

Tue May 19 15:31:42   Up Now: Christian Wolff, University of Regensburg, Germany, “Personal Information Mgmt vs. Resource Sharing”

Tue May 19 15:59:09   Next Up: Oded Nov, Polytechnic Insti. of NYU, “Motivational, Structural and Tenure Factors that Impact Online Community Photo Sharing

Tue May 19 16:55:57 Final Session: Michaela Götz, Cornell Univ., “Modeling Blog Dynamics”

[Details after the break...]

Tue May 19 8:59:45    Join the ICWSM group online at Crowdvine

Tue May 19 9:01:57    Speaking: Duncan Watts, Yahoo! Research, “Using the Web to do Social Science.”
Tue May 19 9:15:31    “My real interest is to think of the Web as a *tool* for social science research” – Watts
Tue May 19 9:16:36    dynamics of networks: Individual level has a lot of turnover and change, but evens out and stabilize at aggregate level – Watts
Tue May 19 9:18:13    Watts asks why are there “cultural hits” in the market – difficult to predict what will be popular
Tue May 19 9:23:19    Hypothesis: inequality and unpredictability derives from “social influence” – Why pay attention to others? Learn what u may not know
Tue May 19 9:25:34    Study: Created [social environment] Music Labs. Recruited 28K people for study. Presented with 48 songs by unknown bands
Tue May 19 9:27:58    Song in the Top 10 got more listens. Not surprised that indiv. influence each other, but what consequences are for market formation
Tue May 19 9:29:45    “Inherent unpredictability” – No matter how much information you have, you still won’t be able to predict effect of social influence
Tue May 19 9:32:35    Implications of Music Lab: it’s possible to study emergence experimentally – would be difficult to do in offline world
Tue May 19 9:37:42    Some argue that Americans are increasingly sorting themselves into like-minded comm. – Bad thing bc we don’t learn/exchange ideas
Tue May 19 9:40:06    Goel, Mason, Watts, develop “Friend Sense” Application – Quiz to see how well you know your “friends”
Tue May 19 9:40:54    Perception and reality not in line;Model ur friends & projection bias. Ppl are really terrible at knowing when their friends disagree
Tue May 19 9:44:19    Watts’ study raises the question of how much individuals really influence each other
Tue May 19 9:45:55    Watts’ next study: Performance-based pay better than fixed-pay? Can you get better work with better pay?
Tue May 19 9:50:49    Amazon Mechanical Turk (AMT) – current leading platform from crowd sourcing.
Tue May 19 9:53:09    People always think they’re being underpaid no matter how much they’re being paid.
Tue May 19 9:54:01    “Anchoring effect” responsible for lack of impact on accuracy/quality.
Tue May 19 9:55:42    Would you work harder if your employer doubled your pay? Most people think they’re already working hard enough.–Watts
Tue May 19 9:56:37    All platforms (Facebook, AMT, Email networks, Web) offer new tools to study social interactions and collective dynamics. – Watts
Tue May 19 10:49:03    “Networks have gone from being too small to too big in a blink of an eye” – Watts

Tue May 19 13:38:15   “How does [system] design impact online community culture?” Panel Discussion #ICWSM
Tue May 19 13:41:25    Panelists: Cathy Gill, Lukas Biewald, Dolores Labs; Rashmi Sinha, SlideShare; Cameron Marlow, Facebook
Tue May 19 13:42:38   Correction: Moderator – Kathy Gill, Senior Lecturer, University of Washington
Tue May 19 13:53:51   Q: Biggest predictor of your activity? a) Distribution; b) Feedback? c) What Your Friend’s Are Doing? A: C – Marlow
Tue May 19 14:06:42   Moderator Question: Tell us a time when a behavior was not what you expected, and what did you learn from that?
Tue May 19 14:07:30  ”You know when someone untags a photo, it’s probably not accidental.” -Marlow
Tue May 19 14:08:30   How do you balance intuition+data in developing new features?
Tue May 19 14:08:53   Sinha: Start with intuition, then become more data driven.On a SNS, so much opp to listen to ppl talk, see how they’re using it, react
Tue May 19 14:09:34   When do you know you should rollback a feature? The first minute people say they hate it? Or wait?
Tue May 19 14:13:49   Sometimes you get emotionally attached, hold onto features longer than should have. You should roll back pretty quickly – Sinha
Tue May 19 14:16:50   The advantage of SNS is that you can get qualitative feedback directly from users; run experiments on features that way – Biewald
Tue May 19 14:20:52   ”The best job we can do is making sure ppl in charge of making decisions are aware of all biases in [design systems]” – Marlow
Tue May 19 14:23:50   Hashtags on Twitter are “Distributed conversations” anyone can track. Really scalable convos, more than any other mechanism. – Sinha
Tue May 19 15:00:40  Startups and smaller orgs have a better chance of staying in touch with their user communities – Gill

Tue May 19 15:05:13   Now Speaking: Tad Hogg, HP Labs, “Stochastic Models of User-Contributory Websites”
Tue May 19 15:16:34   Hogg is focusing on the phenomenology of Digg (users submit/vote on stories they like & friend those they want to follow).
Tue May 19 15:30:37  Reasons users vote for story not available on Digg. Conjectures: story topic, novelty, popularity – Hogg

Tue May 19 15:31:42   Up Now: Christian Wolff, University of Regensburg, Germany, “Personal Information Mgmt vs. Resource Sharing”
Tue May 19 15:34:09  Focus of talk is on “tagging motivations – why do people tag and how does motivation differ from platform to platform” -Wolff
Tue May 19 15:38:35  Tagging – a fourth layer of indexing, behind 1) author keywords; 2) intellectual indexing by prof.; 3) automatic indexing
Tue May 19 15:40:02  Wolff finds tags are usually noun oriented, almost all tags reflect aspects of content, instead of time, task, emotion, sentiment
Tue May 19 15:42:09   Amazon’s Mechanical Turk seems to be a preferred platform for these social studies
Tue May 19 15:54:40   Not surprising: Youtube & Flickr more for Resource Sharing, while Delicious and Connetea are more for Personal Information Mgmt (PIM)
Tue May 19 15:58:16    More to be done in helping people to better organize tags (e.g., auto tagging suggestions, teach professional classification systems)

Tue May 19 15:59:09   Next Up: Oded Nov, Polytechnic Insti. of NYU, “Motivational, Structural and Tenure Factors that Impact Online Community Photo Sharing
Tue May 19 16:00:15   “Why do people share stuff in online communities?” – Nov
Tue May 19 16:00:56   “If you don’t have users sharing goods, you don’t have community.” – Nov
Tue May 19 16:01:44    Understanding *why* they share is essential to know how to sustain the community — Nov
Tue May 19 16:04:43   2005 paper by @dweinberger [David Weinberger] on “Tagging and Why It Matters”
Tue May 19 16:11:16     Content “creation” is independent of “sharing” – Nov
Tue May 19 16:15:32    The more contacts you have, the more likely you’ll do what is expected of you in context of community (e.g., share photos on Flickr)
Tue May 19 16:17:28    People share less the longer they’re in the community – Nov
Tue May 19 16:18:38    Enjoyment (r=.022) not statistically significant in sharing; conflicts with user self-reports that enjoyment is primary reason – Nov
Tue May 19 16:20:29   Enjoyment may lie more in the “creation” rather than the “sharing” – Nov
Tue May 19 16:25:13    Negative correlation btwn “self-development” (-.182) and sharing: Nov suspects that discrepancy comes from quality/quantity trade-off

Tue May 19 16:55:57 Final Session: Michaela Götz, Cornell Univ., “Modeling Blog Dynamics”
Wed May 20 17:00:57 Find Gotz’s research here
Wed May 20 17:02:17  Gotz proposes “Zero-crossing (ZC)” model that eval. inter-posting time, “burstiness” of posting activity, and popularity over time.
Wed May 20 17:04:21   Gotz says the model could be used for forecasting, viral marketing campaigns, and Web advertising, but admits it may be a “long shot”
Wed May 20 17:06:50   ”Poster-Demo Madness” Session – each presenter has ONE minute to whet our appetites