Think about the various tweets, posts and status updates your friends and acquaintances make about feeling sick or under the weather. Likewise, consider videos of street violence or CCTV footage of crimes taking place you might have seen with location details in the description. If you haven’t thought much of this behavior, be assured that others have and that they are looking for patterns.
Who are the people data-mining Twitter for epidemiological information? So far, researchers at the University of Rochester and Johns Hopkins University have made significant strides and have been able to to “predict who would get sick next, up to eight days beforehand , with 90% accuracy.” With this sort of information, government officials could use current data to guide their response to public health threats even before the CDC can draw up data formally. Likewise, the general public could use their awareness of infections to keep themselves away from hotspots, adopt safer practices and ensure that they have less exposure to the virus.
Of course, the system has not yet been perfected, while machine learning is used to teach the tracker the difference between people who were “sick of Mondays” from those who were actually tweeting about illness, other common messages can often causes spikes of perceived flu-like activity when there is none. One such incident has come up when Twitter-users focused on Kobe Byant’s symptoms during a basketball game which weakened the reliability of the computer models. What’s more, the general population is generally uninformed about the flu and has a high possibility of artificially inflating data if people are aware that they are currently in a zone of high-incidence for sickness.
On top of this, patterns of behavior seen in Twitter are also being studied to possibly predict events like genocide. According to the Genocide Watch, the first step leading to genocide is “Classification,” which leads to intolerance and cultural division. The Sentinal Project has taken this information to heart to develop Hatebase, a crowdsourced database for multilingual hate speech in order to aid researchers detect the early stages of genocide with quantifiable data. While results are still too early to be conclusive, positive correlations based on the information provided by Hatebase has made between anti-Baha’i statements made by Iranian officials and an increase in attacks on members of that minority.
How else could this information be used for researchers to learn about society? Well, the data already has political and marketing implications, but some universities have even gone to gauge the mood of countries based on geolocated Twitter content. Utility companies also often maintain their own Twitter accounts to not only address outages in real-time, but to track patterns in their services.
What other data can be tracked? Well, what else do we tweet?
Leave a Reply
You must be logged in to post a comment.