HumansPosted by David Sumpter Mon, March 01, 2010 10:11:53This week's
Economist has an interesting series of articles on the data explosion we have experienced the last few years. Quite alot of this explosion is related to human collective behaviour. As Mehdi pointed out in an earlier post, Google is leading the way here. But there are lots of other sources it seems that big businesses are very interesting in spotting patterns in collective behaviour.
Its funny reading an article like this because first I got the impression that there was lots we can learn from these companies. But later on I came to the conclusions that they just get massive datasets load them in to R and look for correlations between things. Some of the correlations they were apparently surprised by were pretty obvious really, e.g. people mass buy easy to cook food before a hurricane is going to strike. I'm not sure I can do better than this, but it seems the idea of dynamically modelling lies a long way off.
HumansPosted by David Sumpter Thu, January 07, 2010 17:53:55Dirk Helbing has set up a webpage for something called
VISIONEER: Envisioning a Socio Economic Knowledge Collider. The idea is to collect together ideas which can shape future research of collective behaviour of humans. He is putting together a report for the EU on this theme, hopefully resulting in long term funding for this type of research.
I thought it was quite cool to use collective distributed intelligence to find the best way to study collective intelligence.
HumansPosted by Mehdi Moussaid Tue, December 08, 2009 19:08:26Recently, Google has released two nice tools that could be used to observe nice patterns of humans activity.
The first one is
Google trends. It allows you to see how often a specific keyword has been searched on Google over time. One can easily observe some nice 'herding' phenomena where everybody is suddenly paying attention to one specific event. For example, the search curve for keyword '
Harry Potter' shows a sudden peak during summer 2007. By cumulating the search volume day after day, you find a nice S-shaped curve, typical from phase transitions. Here, the transition is from a state where nobody cares about Harry Potter to a state where most people have paid attention to it. The same kind of curves can be observed for keywords 'tsunami' in 2004 (with a sharper transition), or 'bejing' in 2008 and many others. I've been talking about that during the last ECCS conference (see
here).
Similarly, I was surprised by the astonishing regularities of search patterns over years. Try to compare the curves for keywords such as
'snow' , 'beach' , 'football', or
'mothers day' , 'fathers day': Each year displays the exact same search pattern, same peaks , same slope... and even the same mistakes, such as a small amount of people searching for 'fathers day' during August... (See also the nice pattern for the keyword '
science').
The other tool is
Google flu trends, which displays data about the flu epidemics over the world. Again, there are many interesting things to say and to observe, such as the regularities of the epidemics over years, the spreading of the flu among neighboring countries, or the great correlations between these curves and the search volume for the keyword 'flu' each year... Therefore, it's not a surprise that a team from Google Inc. has made a Nature paper out of it:
Detecting influenza epidemics using search engine query dataFor sure, there's a lot to learn just by observing the web.
