Social network users tend to tell some of their personal secrets there, even without knowing it. A group of researchers at the University of Pennsylvania has created an algorithm capable of deducing whether Facebook users are ill and what their ailment is by analyzing their publications.
In the study, researchers asked hospital patients if they could analyze their Facebook postings and compare them to their medical records. Crossing the data, they investigated the words each user used on Facebook. The algorithm determined which words and word combinations were most likely to be related to certain diseases. The researchers then began testing some of the data collected and used the algorithm with the rest of the data to try to figure out what diseases people posting on Facebook had.
For some disorders such as depression, the Facebook data did not provide the information needed for the algorithm to guess, but for diabetes, anxiety or psychosis, the Facebook posts were enough for the algorithm to discover. The authors say that “words expressing hostility (‘tont0’, ‘shit’, ‘motherfuckers’) were the main characteristic of people with drug abuse or psychosis problems”, while “the words most associated with depression were related to somatization (‘my stomach hurts’, ‘my belly hurts’) and emotional distress (‘pain’, ‘crying’, ‘tears’).
The researchers warn that it is difficult to find explanations for the results because, for example, some of the words associated with users with diabetes were related to religion (‘god’, ‘family’, ‘pray’), which may be a mere coincidence with the age of the patients or other demographic factors.