The authors of this paper are from Google, IBM Research, and Utah State University. These are smart folks, and the study shows it.

Their premise was that one could predict, based on tweeters’ past behavior and how they used language, who would be most likely to retweet public service messages. These “information propagators” are more likely to be willing and ready (within a set time frame) to spread information on request.

The authors used two tweets, one location based and the other topic based, for their experiment. The location-based “public safety” tweet reported a shooting in the San Francisco Bay Area, and the topic-based “bird flu” tweet said that bird flu was expected to evolve in nature. Both tweets were obtained from news media sites.

Their retweeting system succeeded when they sent tweets to people with these characteristics:

• They have tweeted on the topic before.
• They have many followers (100 was the bottom threshold).
• They retweeted within a set period of time (possibly six or 12 hours, although the optimal time frame is not stated).
• Retweeters’ word usage in their own tweets tended toward inclusiveness, conscientiousness, and openness on the Big5 and Linguistic Inquiry and Word Count scales.

The authors provide copious information about the statistical analyses in which they engaged, so anyone with an interest in the project should be able to reproduce the experiment.

At the end of the paper, the authors state:

We found that our approaches were able to at least double the retweeting rates over two baselines. With our time estimation model, our approach also outperformed other approaches significantly by achieving a much higher retweeting rate within a given time window. ... In a live setting, our approach consistently outperformed the two baselines by almost doubling their retweeting rates. Overall, our approach effectively identifies qualified candidates for retweeting a message within a given time window.

Emergency management groups could really use a system like this. However, there is no uniform resource locator (URL) or other information about how to access the authors’ system (which was funded by the US Defense Advanced Research Projects Agency, or DARPA). Nevertheless, I suppose that any smart information technology (IT) person would be able to reproduce the system from the paper.