The results of my content analysis of BusinessWeek's Second Life coverage

I mentioned earlier that I was in the midst of building a modified General Inquirer negative dictionary to analyze BusinessWeek's coverage of Second Life in the past year. The results are in. I used Yoshikoder, LexisNexis Academic, Excel, and a modified General Inquirer Negative dictionary to perform the analysis. Here are the negative rates that I obtained:

A couple of notes on the methodology, before I give a quick-and-dirty interpretation of the data. First, the study did not use sampling methods. Rather, all of the BusinessWeek and BusinessWeek Online articles in the LexisNexis database that mentioned "Second Life" were used (articles that used "Second Life" in some other context were discarded.) Still, there weren't many articles to use, as this table shows:

Jan-06 0
Feb-06 0
Mar-06 1
Apr-06 1
May-06 3
Jun-06 0
Jul-06 3
Aug-06 1
Sep-06 2
Oct-06 3
Nov-06 6
Dec-06 1
Jan-07 2
Feb-07 2
Mar-07 2
Apr-07 8
May-07 0
Jun-07 2

Moreover, many of the articles were briefs of just a few hundred words, meaning there not only was a lack of text to analyze, but also the chance of the results being clouded by a single negative sentence were quite high.

Because of the lack of text available for my use, I did not perform a KWIC (keywords in context) analysis. Instead, I performed a full Yoshikoder dictionary report on all of the concatenated articles from each month of the survey, using the giant GI Negative dictionary that I had created. There are a few breaks in the data -- this was because BusinessWeek didn't mention Second Life in June 2006 and May 2007.

So, what does the data mean? The BW articles that were published in the latter part of 2006 generally had a lower percentage of negative terms than those published in the first four months of 2007. This agrees with the anecdotal observations by myself and a few other sources that BW hyped Second Life in late 2006.

However, the negative rates from the early part of 2006 were surprisingly high. In May 2006, the rate approached 5%, and that was the same month BusinessWeek made the famous pronouncement that "Virtual worlds abound in useful business applications." The analysis suggests that there was actually a stronger negative thread running through the BW coverage during this time, although that apparently dropped away during the summer, when the negative rate dipped to about 2.5% in August.

Later this week, I'll try to catch up with other topics that I wanted to blog about on I Lamont, including my trip to Singapore for State of Play V.

State of Play: Networking and a Second Life documentary

The State of Play V conference got off to an excellent start. The first evening of the program featured a fascinating documentary about Second Life and some of the early activists/inhabitants/builders who have helped make it what it is today (sorry, I can't recall the title). There is a lot of interesting history relating to the development of this virtual world, and the film also descibes a fair amount of conflict. Many of the interviewees were willing to frankly discuss these issues, which include tax revolts, the rise of corporate outposts and SL "land barons", message board flamewars, and allegations of slander. Interviewees include Prokovy Neva, Anshe Chung, Guntram Graef, Philip Rosedale, Cory Ondrejka, Wagner James Au, Beth Noveck, Edward Castronova, and lots of other insiders and observers (In fact, I'd say the documentary was a little top-heavy with experts and insiders -- it would have been nice to have more perspective from ordinary residents or newer residents, who have a different view of SL). One of the most delightful aspects of the film was the extremely clever machinima, which complemented the live-action interviews and footage, and was used to recreate various historical issues/episodes. The documentary makers apparently used Second Life to create these scenes, but were somehow able to skillfully incorporate movements, close-ups, and "camera angles" that I did not know know were possible within the standard SL client. The editing was superb, and the music they chose worked well.

The other part of the evening was the dinner/networking event, and that was a great opportunity -- I was able to talk with people whom I would probably not be able to meet in person that easily, including Graef (based in Wuhan, China) and Xu Hui, the founder and CEO of HiPiHi (a Chinese-language-based virtual world that is still in limited beta).

I also met a bunch of very enthusiastic folks from the Singaporean government, which is taking a very proactive approach to encouraging the development of next-generation computing industries. From talking with them, it is clear that virtual worlds and related technologies are on the radar screen of the Infocomm Development Authority of Singapore and other agencies. One person from a local health bureau even made a point of coming to the conference on her own personal time -- she is convinced that VW-related applications will be very important in the future, and she wants to hear and be a part of the discussions that are taking place now.

Preparing for a Computer Content Analysis of Second Life media coverage

For my master's thesis at the Harvard Extension School, I performed a Computer Content Analysis of news articles from the New China News Agency in the 1970s, 1980s and 1990s. Later this week, I am going to use some of the same CCA skills and tools to perform an analysis of Second Life-related news coverage. To that end, I've been spending my free moments this weekend creating a Yoshikoder-friendly version of the General Inquirer negative dictionary used for computer content analysis of political texts. It entails adding wildcards, which Yoshikoder recognizes, but the original GI program did not. This means that the dictionary will be far more sensitive to variations of common negative terms. The creators of the GI dictionary got some variants -- for instance, "exasperate" and "exasperation" -- but missed many other obvious ones, such as "exasperates" and "exasperating". Using "exasper*" will catch these terms.

Of course, wildcards don't work for every word. For instance, "envy" and "envious" could be replaced by "env*", which would get variants such as "envies," but would also catch unrelated words with neutral or even positive meanings -- "envelope," "envision", etc. In this case, I simply added "envies" to the list, rather than using a wildcard.

Converting the General Inquirer dictionary is no easy task. There are 2000 words in the original dictionary that my thesis director gave me (although I see another version contains 2291 words), and each one requires manual review to ensure that wildcards are effectively used and don't introduce unwanted terms into the content analysis that I am planning -- a review of press coverage of Second Life in the past 18 months. Although the GI dictionaries were originally created to examine political texts, I believe they can be used to evaluate other types of text content as well. The GI negative dictionary doesn't contain some of the terms that one typically sees in American or British media articles about new technologies, but it does have a very solid baseline list of negative terms that one might see anywhere.