us

The VICE Channels

    New Study: The Internet Is Not Entirely Assholes

    Written by

    Michael Byrne

    Editor

    I'm pretty sure I've given above-average effort in the online comforting of others. That's probably a bold, overly-hopeful claim, but (if true) this almost certainly has more to do with with where I choose to participate online--a couple of private messageboards populated mostly by people I consider friends, as well as a couple of condition-specific health messageboards--than me being an above-average dude. Am I good at helping internet people out with their woes? Probably not, but it turns out there's a new indexing method that could sort that out with some accuracy.

    Said method is described in a paper posted this week to the arVix pre-print server, courtesy of University of Iowa computer scientist Kang Zhao et al. Their research looks at 500,000 different posts on online health community forums--"80-percent of adult Internet users in the U.S. use Internet for health-related purposes," according to the paper--from 50,000 different threaded discussions over a 10 year period. The goal was to identify where in all of that noise users are successfully making other users feel better. Why does that seem so novel?

    MIT's Technology Review explains the method: 

    They began by manually tagging 300 randomly selected posts as either positive or negative. An example of a post with negative sentiment is:  “My mom became resistant to chemo after 7 treatments and now the trial drug is no longer working :(, ” An example of a post with positive sentiment is: “Hooray! The tumor is gone, according to my doctor!”

    They then extracted features useful in identifying the sentiment, such as the presence of smileys, exclamation marks, and words with positive or negative connotations. They used these features to train software classifiers to automatically spot positive or negative posts.  

    Once a method of quanitfication was established and, with it, a massive pile of data, the researchers were able to let their algorthims loose, so to speak. Zhao et al write:

    We conclude that the more positive the sentiment of responding replies, the greater the positive change in the originator’s sentiment. While this finding only establishes association, it does lend support to our thesis that the sentiment of thread respondents can impact that of originators. The analysis also establishes that social support from respondents generally influences thread originators in a positive way. 

    I'm not actually sure if that's surprising. It isn't to me. But it definitely opens a peculiar door in the world of online indexing compared to, say, a Klout score. It's also interesting to see this kind of cold scientific probe into the part of the internet that is intensely personal--and even just proof that this other side of the internet exists. I think that a fair amount of people don't even like to think of the internet as a source of social comfort, or, rather, don't trust it enough thanks to trolls, internet ultra-irony, or anonymity in general.

    So maybe it's that last little bit that's most interesting, that there's another side of anonymity that can do real good. Real-name advocates (like myself) tend to assume that masks bring out the worst in people, or even that anonymity breeds trolling. But, of course, it's not that easy.     

    Reach this writer at michaelb@motherboard.tv

    Topics: machine, learning, trolls, message, board, culure

    Connect To Motherboard

    Most Popular

    Comments
    comments powered by Disqus