News:Science Article On Social Influence Bias On Voting Patterns ("Herding") With Impact On Biostar.

Question

[Michael]

An interesting article on social herding behavior in voting patterns with impact on QAs sites like BioStar has just come out in Science:

Lev Muchnik, Sinan Aral, and Sean J. Taylor, Social Influence Bias: A Randomized Experiment, Science 9 August 2013: 341 (6146), 647-651. [DOI:10.1126/science.1240466], http://www.sciencemag.org/content/341/6146/647

Quote from the abstract:

Whereas negative social influence inspired users to correct manipulated ratings, positive social influence increased the likelihood of positive ratings by 32% and created accumulating positive herding that increased final ratings by 25% on average. This positive herding was topic-dependent and affected by whether individuals were viewing the opinions of friends or enemies.

Herding effects were topic dependent.

The authors showed significant positive herding effects exists for topics like "politics", "culture and society", and "business". No significant impact was shown for e.g. "IT" (I am so relieved ;)), "economics", and "news". They observed a general trend to "correct" negative votes by up-votes. No negative herding effect was observed.

Answer 1

[Istvan Albert]

Reminds me of one of the first papers I participated in about 10 years ago - one that looked at manipulating user's ratings by showing them altered (inflated/reduced) recommendations: Is Seeing Believing? How Recommender Interfaces Affect Users’ Opinions . Turns out one can elicit higher and lower ratings on items the user knows well just by showing higher/lower predictions. In the end it all seems to have a logical explanation, we are social creatures that trust one another.

I also would offer a more logical explanation on why some fields show less bias. The more objective the topic the less likely to be polarizing. That's why say discussions on IT will be radically different than politics.

Then there are "ego centric" sites, FB, Twitter, G+, reddit where the majority of content is about personal experiences, opinions, achievements, feelings etc. Then there are "altruistic" sites like Biostar, where most content is created with the goal of helping someone else. The usage patterns will differ there as well.

Answer 2

[Mary]

Sure, I think there's a similar psychology. I remember in the early days I gave essentially the same answer as someone with higher ratings than I had--and I gave it earlier--and I came back to see that person had 5 uprates and I had none.

However, I don't think it matters that much. The answers here are frequently not generally purely opinions--they are actually things you can try out, visit, or otherwise verify*.

But you also have to care about whether something is popular or not. I don't. I just care if it's right, useful, or effective.

*Ok, but this answer is opinion.

[KCC]

I have seen situations (on stack exchange type sites) where someone with a very high score makes a trivial comment saying they agree with somebody else's comment, and the high-ranked individual gets upvotes for this, even when the person who made the original comment does not.

Answer 3

[KCC]

When I saw article, I thought it might have some implications for Biostars as well. However, one of the important differences with Biostars is each person has a reputation. I suspect the reputation might have a vastly stronger effect than the actual vote count on any individual answer or comment. Although, you could think of reputation as a way of voting for all the future comments that a person will ever make on Biostars, with each past vote for the individual, not counting as much as a vote for the individual comment. However, maybe there is an equivalence like 10k of reputation is like having all your answers or comments getting 1 upvote.

I also suspect that where a comment falls on the fact vs. opinion spectrum would determine how susceptible it is to being voted up or down, with opinion based comments more prone to herding behaviors.

[Michael]

I think it is hard to say anything about the impact of a feature that was not investigated. One might hypothesize about a small positive "donation effect" ("poor guy has no reputation yet, let's donate an up-vote for a warm welcome"). Anyway I, personally, believe (believe == I don't know anything about it) the impact of existing reputation is much smaller than you think.