Slashdot Mirror
Data Mining Shows How Down-Voting Leads To Vicious Circle of Negative Feedback
KentuckyFC writes: "In behavioral psychology, the theory of operant conditioning is the notion that an individual's future behavior is determined by the punishments and rewards he or she has received in the past. It means that specific patterns of behavior can be induced by punishing unwanted actions while rewarding others. While the theory is more than 80 years old, it is hard at work in the 21st century in the form of up- and down-votes — or likes and dislikes — on social networks. But does this form of reward and punishment actually deter unwanted actions while encouraging good behavior? Now a new study of the way voting influences online behavior has revealed the answer. The conclusion: negative feedback leads to behavioral changes that are hugely detrimental to the community. Not only do authors of negatively-evaluated content contribute more but their future posts are of lower quality and are perceived by the community as such. What's more, these authors are more likely to evaluate fellow users negatively in future, creating a vicious circle of negative feedback. By contrast, positive feedback does not influence authors much at all. That's exactly the opposite of what operant conditioning theory predicts. The researchers have a better suggestion for social networks: 'Given that users who receive no feedback post less frequently, a potentially effective strategy could be to ignore undesired behavior and provide no feedback at all.' Would Slashdotters agree?"
Don't feed the trolls?
********************
I object to Intellect without Discipline.
That's the criticism, well maybe that's too strong a word... That's the critique I've had of Slashdot's moderating system. By allowing both up-votes and down-votes, you create a system where the voice of the majority can drown out the voice of the minority. I've often seen people here express the mistaken belief that if a minority viewpoint is introspective or informative, it will survive the unfair downvotes and rise to the top. It doesn't work that way.
The average ranking is not rank = up - down. It's rank = p1*up - p2*down. Where p1 is the size of the population which would rank it up, and p2 is the size of the population which would rank it down. A minority viewpoint consequently gets a disproportionate number of unfair downvotes simply because it's a minority viewpoint, and thus has to garner a lot more upvotes just to obtain an equal ranking to a majority viewpoint.
For an apolitical, non-religious example, consider Windows vs. Linux. Say Windows users outnumber Linux users 50:1. Now imagine if a search engine let you rate search results based on whether they were useful or not useful, which is then used to prioritize subsequent search results. In every population, there's going to be an idiot segment who votes stuff down simply because they don't like it, not because it was inaccurate or irrelevant it was to their query. Consequently, if a search for hard disk repartitioning brings up four Windows sites and one Linux site as the top results, the Linux site is going to have 50x as many downvotes from those idiot users who never specified Windows in their search but were upset that an "irrelevant" Linux site was included in the search results. If the idiot segment of the Windows population exceeds 2% (numerically equivalent to 100% of the Linux population), that Linux site will end up with a negative rating regardless of how useful or informative it is.
I say "criticism" is too strong a word because neither way is the "right" way to do it. They are just different. A moderating/ranking system which only allows upvotes simply generates different results from a moderating system which allows both upvotes and downvotes. Sometimes the former is more useful; sometimes the latter is more useful. The important thing is to understand the limitations of both and how it will bias the rankings, and not fall into the mistaken belief that a minority viewpoint has just as easy a time reaching +5 on Slashdot as a majority viewpoint. If a contrary viewpoint reaches +5 on Slashdot, it must be making a helluva good point.