Herd effect of voters. Muuu ... MIT research
Muuu ... Internet users tend to follow the herd when it comes to rating comments on a news site.
“Wisdom of the crowd” became the mantra of the Internet era. Need to choose a new vacuum cleaner? Check out reviews on Amazon. Is this restaurant good? See what TripAdvisor or Yelp will tell you. But a new study shows that such online assessments do not always determine the best choice. A large-scale experiment by MIT ’s web users has found that such ratings are highly susceptible to irrational “herd behavior” and that the herd can be manipulated.
Sometimes the crowd is really smarter than you. Classical examples are the assumption of the weight of a bull or the number of candies in a can. Your guess will probably be far from the truth, while the average choice of many people is noticeably closer to the true number. But what happens when the goal is to evaluate something less tangible or material, like the quality or value of a product? According to one version, the wisdom of the crowd is still real. It is a measure of the totality of people's opinions and forms a stable, reliable value. Skeptics, however, argue that others can easily influence people's opinions. So pushing a crowd at an early stage of opinion formation, suggesting opposing opinions, for example, provoking a crowd to a very positive or very negative assessment, can influence the opinion of the crowd in one direction or another. To check which of the hypotheses is correct, you will need to manage a huge number of people, offering them false information, and determine how it affects their opinion.
')
The team, led by Sinan Aral, a scientist at the Massachusetts Institute of Technology in Cambridge, did just that. Aral secretly worked with a popular site where new posts were published. He says that this site is similar to Reddit, but keeps the name secret, as he has another experiment currently with the same site and doesn’t want it to be spoiled by the media (Habré got the info that there was News2Ru, but where does this information come from, if the author of the study is silent, remains a mystery). The website allows users to leave comments on the news and vote for each other's comments up or down. The counted votes are visible in the form of numbers next to the comment, the position of the comments of a chronological sequence ... (well, everything is straight, like on Habré) Posts on the site receive on average about 10 comments and about three votes for the comment. This is a continuation of his experiment with movie ratings, whose task is to determine how individuals influence each other online (result: strongly). This time, the scientist wanted to know how much the crowd influences a person, and whether it can be controlled from the outside.
For 5 months, each comment entered by the user randomly received a voice "up" (positive), a voice "down" (negative), or no voice at all for comparison. The group then watched users rate these comments. In total, over 100,000 comments were received from users, which were viewed over 10 million times and estimated over 300,000 times.
At least, when it comes to comments on news sites, the crowd behaves more "gregariously" than wisely. Comments that received fake positive votes from researchers were 32% more likely to get more positive votes than comments without any votes at all. By the end of the study, the positively manipulated comments received a total markup of about + 25% of the vote. However, the same is not true for negative manipulations. Ratings of comments that received fake negative votes are usually negated by positive votes.
" Our experiment does not reveal the psychology of decision-making by people," says the Aral. “But this is a clear explanation of the fact that people are more skeptical about negative social influence. They are more willing to support the positive opinions of other users. ”
Duncan Watts, a Microsoft Research scientist in New York, agrees with this conclusion. “But one question is whether the herd effect is relevant only for this site or the truth as a whole,” says Watt. He points out that the news category in the experiment had a strong influence on people, and how much they could manage. “The category of business is quite similar to the category of economics, but in the first category the influence is 50% more. What explains this difference? If we are going to apply these findings in the real world, we need to know the answers. ” Will the company increase its online sales by manipulating ratings on a massive scale? “It's easier said than done,” says Watt. If people find that the comments on the site are being manipulated, the herd can get scared and run away ...
Transfer. Original: Sciencemag.Org JOHN BOHANNON
The study also reports that friends are more likely to support each other. Considering that Habr is a party quite closed (registration only by invite, karma, rating, etc.) and this is also a site that generates posts on various subjects, then, logically, there is an even greater effect of supporting each other (again, according to the research in different hubs this may manifest itself in different ways, as emphasized by Duncan Watts). Of course, we are all educated here, well-read and with our own opinion, but against universal reflexes and MIT research, “but you will not argue ...
My subjective conclusions (everyone can make his own for himself, I do not claim to have no alternative):
1. Habr is a great lad who hid votes for posts before voting. The same thing needs to be done for comments, because this study describes them. I would even say that it is vital because Karma / rating is influenced by both posts and comments equally.
2. If you are the owner of a new customer-oriented resource, having received some positive feedback in the form of comments or likes before forming public opinion, you will be able to significantly predispose your audience to a new product.
3. Once, on the advice of a tripadvizor, I stopped at a hotel in Istanbul and went to restaurants with a high rating ... It was hell. It seemed that the ratings were formed in this application by poor students. It is understandable because They are the main users of mobile technologies and applications of this kind, but then you have to write a “rating made up of such and such an audience.” In general, personally, in your place, I would not trust the stars on the booking, tripodvizor and other similar services.
“Wisdom of the crowd” became the mantra of the Internet era. Need to choose a new vacuum cleaner? Check out reviews on Amazon. Is this restaurant good? See what TripAdvisor or Yelp will tell you. But a new study shows that such online assessments do not always determine the best choice. A large-scale experiment by MIT ’s web users has found that such ratings are highly susceptible to irrational “herd behavior” and that the herd can be manipulated.
Sometimes the crowd is really smarter than you. Classical examples are the assumption of the weight of a bull or the number of candies in a can. Your guess will probably be far from the truth, while the average choice of many people is noticeably closer to the true number. But what happens when the goal is to evaluate something less tangible or material, like the quality or value of a product? According to one version, the wisdom of the crowd is still real. It is a measure of the totality of people's opinions and forms a stable, reliable value. Skeptics, however, argue that others can easily influence people's opinions. So pushing a crowd at an early stage of opinion formation, suggesting opposing opinions, for example, provoking a crowd to a very positive or very negative assessment, can influence the opinion of the crowd in one direction or another. To check which of the hypotheses is correct, you will need to manage a huge number of people, offering them false information, and determine how it affects their opinion.
')
The team, led by Sinan Aral, a scientist at the Massachusetts Institute of Technology in Cambridge, did just that. Aral secretly worked with a popular site where new posts were published. He says that this site is similar to Reddit, but keeps the name secret, as he has another experiment currently with the same site and doesn’t want it to be spoiled by the media (Habré got the info that there was News2Ru, but where does this information come from, if the author of the study is silent, remains a mystery). The website allows users to leave comments on the news and vote for each other's comments up or down. The counted votes are visible in the form of numbers next to the comment, the position of the comments of a chronological sequence ... (well, everything is straight, like on Habré) Posts on the site receive on average about 10 comments and about three votes for the comment. This is a continuation of his experiment with movie ratings, whose task is to determine how individuals influence each other online (result: strongly). This time, the scientist wanted to know how much the crowd influences a person, and whether it can be controlled from the outside.
For 5 months, each comment entered by the user randomly received a voice "up" (positive), a voice "down" (negative), or no voice at all for comparison. The group then watched users rate these comments. In total, over 100,000 comments were received from users, which were viewed over 10 million times and estimated over 300,000 times.
At least, when it comes to comments on news sites, the crowd behaves more "gregariously" than wisely. Comments that received fake positive votes from researchers were 32% more likely to get more positive votes than comments without any votes at all. By the end of the study, the positively manipulated comments received a total markup of about + 25% of the vote. However, the same is not true for negative manipulations. Ratings of comments that received fake negative votes are usually negated by positive votes.
" Our experiment does not reveal the psychology of decision-making by people," says the Aral. “But this is a clear explanation of the fact that people are more skeptical about negative social influence. They are more willing to support the positive opinions of other users. ”
Duncan Watts, a Microsoft Research scientist in New York, agrees with this conclusion. “But one question is whether the herd effect is relevant only for this site or the truth as a whole,” says Watt. He points out that the news category in the experiment had a strong influence on people, and how much they could manage. “The category of business is quite similar to the category of economics, but in the first category the influence is 50% more. What explains this difference? If we are going to apply these findings in the real world, we need to know the answers. ” Will the company increase its online sales by manipulating ratings on a massive scale? “It's easier said than done,” says Watt. If people find that the comments on the site are being manipulated, the herd can get scared and run away ...
Transfer. Original: Sciencemag.Org JOHN BOHANNON
From the author of the post ...
The study also reports that friends are more likely to support each other. Considering that Habr is a party quite closed (registration only by invite, karma, rating, etc.) and this is also a site that generates posts on various subjects, then, logically, there is an even greater effect of supporting each other (again, according to the research in different hubs this may manifest itself in different ways, as emphasized by Duncan Watts). Of course, we are all educated here, well-read and with our own opinion, but against universal reflexes and MIT research, “but you will not argue ...
My subjective conclusions (everyone can make his own for himself, I do not claim to have no alternative):
1. Habr is a great lad who hid votes for posts before voting. The same thing needs to be done for comments, because this study describes them. I would even say that it is vital because Karma / rating is influenced by both posts and comments equally.
2. If you are the owner of a new customer-oriented resource, having received some positive feedback in the form of comments or likes before forming public opinion, you will be able to significantly predispose your audience to a new product.
3. Once, on the advice of a tripadvizor, I stopped at a hotel in Istanbul and went to restaurants with a high rating ... It was hell. It seemed that the ratings were formed in this application by poor students. It is understandable because They are the main users of mobile technologies and applications of this kind, but then you have to write a “rating made up of such and such an audience.” In general, personally, in your place, I would not trust the stars on the booking, tripodvizor and other similar services.
Source: https://habr.com/ru/post/195062/
All Articles