The Facebook Effect, by David Kirkpatrick 发布的博文
NY Times' flu story re Google--fb may ultimately predict better
Yesterday's New York Times story by Miguel Helft told a fascinating tale of how analyzing search queries using a new tool from Google.org called Google flu trends may be able to accurately identify regional outbreaks of flu. Here's a quote from the article-- "The material published on the Web amounts to a form of 'collective intelligence' that can be used to spot trends and make predictions." Then at the story's end MIT professor Tom Malone says "we are just scratching the surface of what's possible with collective intelligence."
Fair enough. However, using the fact that someone types "thermometer" into a search query as an indicator that there may be a flu epidemic underway is rather roundabout. The entire approach is, in effect, inferential. It infers something from something else--i.e. that a person has the flu because they do a search query for "flu symptoms."
What about a place where people actually say what they are doing, for example explicitly writing that they are sick? That place, of course, is facebook.
All the things the article said may be possible in terms of making predictions and exploiting collective intelligence are even more true about facebook. To apply analytical tools to the collective conversation inside facebook is likely to yield fascinating and highly useful fruits for both commercial and non-commercial purposes over time.
Facebook already operates a service called Lexicon that can give a crude sense of what's possible. Lexicon can chart, for example, the sentiment inside the service toward one idea as opposed to another. Take a look here to see a comparison of facebook sentiment over time towards Obama and McCain. More sophisticated algorithms will eventually no doubt yield far more complex insights.
Ideally, Helft in the Times piece would have noted that Google is far from the only place on the Internet where insights highly useful to society may potentially be gleaned.
Fair enough. However, using the fact that someone types "thermometer" into a search query as an indicator that there may be a flu epidemic underway is rather roundabout. The entire approach is, in effect, inferential. It infers something from something else--i.e. that a person has the flu because they do a search query for "flu symptoms."
What about a place where people actually say what they are doing, for example explicitly writing that they are sick? That place, of course, is facebook.
All the things the article said may be possible in terms of making predictions and exploiting collective intelligence are even more true about facebook. To apply analytical tools to the collective conversation inside facebook is likely to yield fascinating and highly useful fruits for both commercial and non-commercial purposes over time.
Facebook already operates a service called Lexicon that can give a crude sense of what's possible. Lexicon can chart, for example, the sentiment inside the service toward one idea as opposed to another. Take a look here to see a comparison of facebook sentiment over time towards Obama and McCain. More sophisticated algorithms will eventually no doubt yield far more complex insights.
Ideally, Helft in the Times piece would have noted that Google is far from the only place on the Internet where insights highly useful to society may potentially be gleaned.