Blog: Richard Hackathorn

« Sybase Enters Data Warehouse Appliance Market | Main | IBM Expands InfoSphere Branding »

Microsoft Research Confirms a 'Small World' for IM Chats

In an outstanding research article, Jure Leskovec of Carnegie-Mellon and Eric Horvitz of Microsoft Research analyzed 30 billion IM conversations among 240 million people during the month of June 2006. This resulted in a network graph with 180 million nodes (a person who sent a message during that month) and 1.3 billion undirected edges (two persons exchanged messages) - by far the largest data set for social networking. In fact, the authors described their investigation as being 'planetary-scale'!

A key result was the confirmation of Stanley Milgram's popularized hypothesis that any two persons in the world are, on the average, separated by six degrees. That is, only four other persons are needed to form a path between those two persons. Leskovec and Horvitz found that the average degree of separation was 6.6 with a median of 6. The longest paths had lengths of under 30, which is still amazing. In contrast, if one person wanted to find another person among a group of 180 million, you would expect to search through half of that group. Right? Correct if you were searching randomly for that person.

This is called the Small World phenomenon where the interconnections of a social network are much more strong than one would expect based on random networks. In other words, we as social beings have a strong affinity for selectively connecting to others, even using Microsoft Messager.

The above result about six degrees was only one of dozens. The results of the research article go on and on! The reader is forewarned that the article is slow reading, but only 10 pages.

I wrote an article entitled The Link is the Thing in August of 2003, in which I applied the Small World phenomenon to enterprise data warehouses. We may have small worlds lurking in our EDW! I encourage you to apply network analysis techniques to EDW as I suggested. I obviously would be very interested in what you find.

Share: