During our weekly SoNet internal research meeting, my colleague Napo presented the paper “Predicting the Future With Social Media” by Sitaram Asur and Bernardo A. Huberman, archived on arXiv in March 2010. Using Twitter posts, they are able to forecast box-office revenues for movies, outperforming market-based predictors. They also do sentimental analysis on Twits by asking Mechanical Turk to tag few twits as positive, neutral, negative and then they train LingPipe to predict the positiveness of all the other millions of twits. Read it! Very interesting paper!
Today I attended an amazing presentation by Bernardo Huberman, director of the Information Dynamics Laboratory at HP Labs, titled “Social Dynamics in the Age of the Web”. Below the roughly editing notes I took during the amazing presentation. They are not intended to represent what Bernardo said but just to give you (me!) some pointers.
His first slide was just a painting by Bruegel “Village Feast” and discussed about “how sociologists would try to understand the situation, the relationships in the village” and how this would take a lot of time! He also cited some research about 4 widows in a village in Norway who were “studied” by a researcher for 5 years.
The alternative we have now is called Web/Internet and allows to collect tons of social information very easily, but beware, you lose a lot of details!
Then he brought up the issue of “attention” (You put things there (Youtube, Flickr, …) because you think someone is going to read it…)
Cited his paper “Assessing the Value of Coooperation in Wikipedia”. Question: “How good is Wikipedia (on average)?” Not interested in controversial articles. Predicting the number of edits a wikipedia page would get. Log-log distribution of edits on wikipedia pages.
Question: “the articles with lots of edits are the best ones?” “How to measure quality?” First approach: pagerank of page URL but not too representative of quality. Then got the Wikipedia featured articles.
Attention: the economics of attention. Information is now plenty available. Information has now no value. Some time ago “give me information about a hotel in New York”, and you paid for this info. Now go on the web and find lots of info about hotels in New York. You cannot ask to be paid for this information now, but few years ago yes!
Instead attention has a lot of value now! Attention is what is scarse (at least online).
“Economics is actually the science of distribution of scarse resources.”
“All this scientific citation game is an attention game”.
“How do people compete for attention? Imagine if now there are other 3 people giving talk, then things becomes interesting!!!”
“Spam is a phenomena of trying to get attention.”
He studied Viral marketing on Amazon.com and showed us two very different recommendations networks. For a medical book, recommendations network very spread (almost random graph). For a Japanese graphic novel (manga), recommendations network very power-law (4 people (hubs!) recommended this book to everybody!!!)
Temporal dimension: How long it takes for recommendations to propagate in the social network?
Stimulation: novel stimula fades (if you get a spike in the finger after some time you feel a sort of basic pain but not the pain you felt in the beginning. You get used to it. When novelty fades we search for novel experiences. Paper: “Novelty and collective attention” (2007)
Hypo: attention decays in time as a strecthed exponential
There is a phase transition (Bernardo is a physicist): a critical value in which prioritizing by novelty is ok but then prioritizing by number of diggs (he studied novelty and ranking on digg.com)
The second part was about opinions. Q: How epinions form and evolve?
Salomon Ash: group polarization. The basic idea at that time was “discussion in the agora will bring toward the middle, not towards the extremes”. But his experiment invalidated this common perception.
And then Cass Sunstein (which I love!), read Republic.com!
Is there group polarization online or not? It depends … on cost of providing your opinion.
On Jyte.com it is just a yes or no: costless! –> extremes! group polarization!
Amazon reviews are costly (take time to write them!) –> no group polarization.
Why? Possible explanation: over time, you give reviews only if you have a possibility of impacting the average, only if you disagree with the consensus already reached! He studied Amazon and IMDB.
Suggestion he gave to Amazon: “Want to produce good reviews for a book? Start putting bad reviews for that book and people will flock to give their (opposite) opinion!!!”
He is staudying right now ratings on Youtube over time.
Comparing ratings with reviews and ratings without reviews on Amazon!!!
If you write reviews (costly), you tend to disagree with the average (low ratings for movies with high average, and viceversa)
Question: people behave differently in online environments?
Yes but the new media is sucking in a lot of people, especially young people … we’re moving there, wikipedia is a phenomena you cannot deny.
“We are going back to the village, but the village is not defined by physicality” (tribes).
He told us that now you can selling your friends on facebook, on ebay!!!! You can go on Ebay and sell there your Facebook friends!!!
My personal comment: “this is what I call Social capital!!!” ;)
A quick search gave me sellyourfriends.com, the Facebook application which I added to my Facebook profile or read what blonde2.0 has to say about (blonde2.0, great! ;-)
“Most people in myspace don’t even know about wikipedia”
“Quality and attention are not correlated”. Paris Hilton is an attention genius, she nows what is novel in people’s mind. The only reason for which paris is famous is that she is famous. So attention is not correlated with quality.
I asked him how he does see science in 30 years. There will be no more scientists by profession because everybody will be a scientists (blogger baiscally)? He said he believes there will be even more need for scientists because specialized knowledge is hard.
Well, persoanlly I think that, since before we were saying people compete for attention and you are more likely to get attention for short, catchy, bursty sentences (possibly with the word “sex” in it), a stupid idea just flashed into my mind so why not writing it in this chaotic post? “Is there space for a sciencetwitter?!? Divulgate your research in less tha 150 characters!” At least it could be fun! ;) Spreading it to my sci.bzaar.net pals.
Somehow Orkut became a trash site for brazilians (not his words), Google woould have liked to make it a facebook of course, so, if google was not able to control the eveolution of a social network, well, this means it is not easy at all!!!
Last note, I think I overexxagerated with the tags in the title, would “sex”, “virgin”, “Paris” make this post the most accessed? We’ll see …