LANGUAGE LEAKS LOST LOVE
thousands of reddit users' posts were analysed for signs of their impending relationship break-ups. they were a bit less analytical than usual, & said 'i' and 'we' a bit more. love hurts.
original article: Seraj et al., 2021 (Proceedings of the National Academy of Sciences of the United States of America), reported in: The Daily Mail by Shivali Best on 1st February 2021 & The Times by Rhys Blakely on 2nd February 2021 image source
this story was in episode 3 #breakup #male #reddit #dumped #loser
the error bar says
the Mail and the Times report on a study of the language people use when they are about to go through a breakup.
the study by psychologists in Texas, USA used computer text mining to analyse more than a million posts on the social media site Reddit, in a particular subsection about relationship break ups.
the text analysis identified a decrease in the use of analytical - or problem-solving - thinking, increases in more emotional cognitive processing, and increased use of the personal pronouns 'I' and 'we'.
the authors claim that such methods help understand the emotional upheavals of relationship breakups; the newspapers claim that people about to break up change the way they speak.
can the language we use betray our failing relationships?
while the headlines state that we can use our partner's language to warn us of an impending breakup, the evidence is not so clear-cut.
first, this study used the large, freely-available dataset of Reddit, a social media site on which all users and posts are public and can be fully anonymous. none of the users, relationships, or stories were, or can be, verified for truthfulness. and they were dredged for data.
second, the researchers targeted a section of the site called BreakUps. and they only looked at users who went to that site for the first time, presumably to post about their relationship break up.
third, the authors assumed that the time of the first post was also the time of the breakup. they only actually read two hundred of the 1,027,541 posts. of those 200, only 121 gave an indication of the break-up time, and of them, the majority - 98 - said it was within the previous 2 weeks. so, fewer than half of the relatively small sample of posts said that the breakup was recent. so any claims that language can predict break ups are impossible to verify.
fourth, while the headlines and claims are about 'people' in general, two thirds of Reddit users are male, and typically young. this is reflected in the very patchy and unreliable demographics of the study 'participants': an estimated 70% were male, 67% had been dumped, the average age was 23, and average relationship length was 2 years.
fifth, despite containing millions of datapoints, dozens of variables tested, and looking at each variable over every single 2-week period over 2 years, there is no proper treatment of the statistical problems of false-positives - due to many tests being done - or of auto-correlation - the fact that the same people are providing data at many different time-points, and the number of people changes at each time-point.
finally, even if the results stand up, the introduction contains contradictory predictions about the usefulness of the 'analytic' type of thinking - the authors predicted on page 1 that people deciding to leave a relationship should increase their analytical thinking, but on page 2 that people will decrease their analytic thinking as a breakup occurs. page 2 lists a long "set of hypotheses" that are all a little vague. who starts a study with a set of hypotheses.
overall: this study has the faint but familiar smell of a deep sea fishing expedition.
.jpg){kind=link}