A novel use for a social media tool

CUPERTINO, California (AP) – An online tool to help you get pregnant? Surely, do not let the April 1 dateline fool you.

While dating and romance sites have helped bring couples together for years, there is now an online service that can actually help existing couples find the right time to conceive. Which is not at all what the creators had in mind.

Twitter.com is the breakout web community of 2007. Based on a simple premise, “What are you doing right now?”, hundreds of thousands of “Twitterers” have been sharing their most mundane thoughts and actions for the past several months. Earlier this year attendees at the influential South By Southwest Conference in Austin spurred the site to a new level, bringing together a massive community of users.

It was exactly the sort of data pool Dr. Phil Larkin needed to test his theory.

Larkin and his team of graduate students have spent the last 18 months analyzing language, and in particular how mood affects word choice and syntax. The problem they had in reaching a meaningful framework was the lack of a standard baseline. The first attempts at cracking a pattern were centered around weblog posts. However, not enough is known about the typical patterns of a new blogger.

Enter Twitter

The team’s discovery of Twitter.com changed the dynamic considerably. “The thing we loved about Twitter was the lack of friction,” Larkin said. “Since all of the posts are short, and there’s no need to worry about the inspiration of the author, we now have a blogging rhythm that is more periodical and predictable.” Unlike regular science, Larkin had discovered a fantastic data source. All he needed now was a hypothesis, and even that took a while to find.

“We had been studying Twitter for months, just waiting for an explosion in traffic,” research associate Darla Turner said. “We knew it would be an easy match for something, as long as we found the right baseline and the right problem to solve. Right before South By Southwest, we figured it out and started soaking in the data stream. The timing was perfect.”

And as it turns out, the problem they solved was one of timing as well.

The Birth of an Idea

As the volume of the Twitter information started to slowly increase, Larkin saw a need for a new measurement. As Twitter-posts (or “Tweets” to fans) are limited to 140 typed characters, the content would always be limited. But Larkin did see the potential for measuring the volume and frequency of the posts themselves. “All we had to compare this data to was itself, so we needed a way to tie it back in a statistically sound way.”

No one on the team is sure about the real origin of the idea, but Turner gets credited for it indirectly. “The other members of the team used to laugh at me when I’d complain about my menstrual cycle. I simply did not feel like writing. Then we went back and looked, and sure enough there was a cycle where I wrote less often, fewer words, and fewer syllables.”

The idea was almost dismissed as too controversial from the outset. Was it possible to tell which women were having pre-menstrual syndrome (PMS) simply by reading their Twitter.com logs? The first hurdle was to determine if any such pattern was detectable. “The early volume of Twitter was much smaller, but easier for us to pull from,” Larkin said. “New users, as we have found, take a couple of months to settle into a regular pattern of posting their Tweets. There are so many new users now we have to be careful about how we check the formula, but so far it holds up.”

So exactly what unintended service is Twitter now able to offer?

“The First Web-based Basal Thermometer”

“I almost peed my pants when the thought popped into my head,” Turner admitted. “If we can look at female Twitter user who has been on the system for awhile and tell when she is suffering PMS, we can make a good guess about when she’ll be fertile.” Larkin added “It was the last thing we thought we’d ever do with a website.”

The previous research by Larkin’s staff has not gone to waste. Word choice and simplicity also play a role in identifying the ovulation window more accurately. “Now that we can find that one week in the big cycle, we can now examine the more subtle patterns in the smaller ovulation cycle. When the data matches up and reinforces itself, we can make a fairly accurate prediction.

In hindsight, Larkin is not surprised by the results of this unusual data-gathering. “For this to work, we needed lots of people. The men were important too, because we were able to show they didn’t have the same one-week-of-four anomaly. We also needed people who didn’t know they were participating in a study. Self-selection in a sample group can sneak unintended biases into the equation.”

[tags]Ike Pigott, Occam’s RazR, Twitter, humor, parody[/tags]