The effects of social recommendations on network diversity(blog.acolyer.org)
blog.acolyer.org
The effects of social recommendations on network diversity
https://blog.acolyer.org/2018/05/24/algorithmic-glass-ceiling-in-social-networks-the-effects-of-recommendation-on-social-diversity/
6 comments
It seems like their model is too simple to draw any firm conclusions? It essentially treats all types of content the same. That is, what about this causality flow:
group -> type of content generated -> audience reach
People in different groups will prefer to author different kinds of content at different rates, and this could entirely account for audience reach.
For example, suppose we classify content by the language it's written in? One would expect people who write content in English or maybe Chinese to reach a larger audience than people generating content in a less well-known language.
But there are plenty of other dimensions that will affect audience reach.
group -> type of content generated -> audience reach
People in different groups will prefer to author different kinds of content at different rates, and this could entirely account for audience reach.
For example, suppose we classify content by the language it's written in? One would expect people who write content in English or maybe Chinese to reach a larger audience than people generating content in a less well-known language.
But there are plenty of other dimensions that will affect audience reach.
Even with a perfectly time-ordered feed with sharing/retweet functionality, a pareto distribution will emerge among content that bubbles up. That's how networked economies seem to work.
http://www.labs.hp.com/research/idl/papers/ranking/adamicglo...
http://www.labs.hp.com/research/idl/papers/ranking/adamicglo...
The problem is that the popularity of products in a networked economy follows the power law:
http://www.aspeninstitute.org/policy-work/communications-soc...
You might recognize the power distribution curve, as it's the same as the graph of the long tail:
http://en.wikipedia.org/wiki/Power_law#mediaviewer/File:Long...
What happens is that you have a handful of people who make a LOT of money / bust out virally, a small few who do OK, and the vast, vast majority starve. Since our economy is increasingly becoming an online, globally networked one, these effects are becoming stronger, and are a contributor to economic inequality.
Power Law phenomenon on the net has been observed for ages. Here's a post from 2003 where Kottke notes the distribution in the popularity of blogs on Technorati: http://kottke.org/03/02/weblogs-and-power-laws
I imagine the curve fits similarly for things like app store rankings, Reddit and Hacker News post popularities, top Steam sellers, Amazon rankings, etc.
> The world still rewards value, even if it takes some time. The people clamoring that it doesn't are doing so because they want to believe that it isn't their fault they didn't succeed.
The funny thing is, this is not entirely true. Quality is only important up to a certain threshold, after which you're at the mercy of what are essentially chaotic network effects early in the lifecycle of your product.
Salgankik, Dodds, and Watts performed an experiment that begins to provide some empirical support for this intuition [359]. They created a music download site, populated with 48 obscure songs of varying quality written by actual performing groups.
Visitors to the site were presented with a list of the songs and given the opportunity to listen to them. Each visitor was also shown a table listing the current “download count” for each song — the number of times it had been downloaded from the site thus far. At the end of a session, the visitor was given the opportunity to download copies of the songs that he or she liked.
Now, unbeknownst to the visitors, upon arrival they were actually being assigned at random to one of eight “parallel” copies of the site. The parallel copies started out identically, with the same songs and with each song having a download count of zero. However, each parallel copy then evolved differently as users arrived. In a controlled, small-scale setting, then, this experiment provided a way to observe what happens to the popularities of 48 songs when you get to run history forward eight different times. And in fact, it was found that the “market share” of the different songs varied considerably across the different parallel copies, although the best songs never ended up at the bottom and the worst songs never ended up at the top.
Salganik et al. also used this approach to show that, overall, feedback produced greater inequality in outcomes. Specifically, they assigned some users to a ninth version of the site in which no feedback about download counts was provided at all. In this version of the site, there was no direct opportunity for users to contribute to rich-get-richer dynamics, and indeed, there was significantly less variation in the market share of different songs.
There are clear implications for popularity in less controlled environments, parallel to some of the conclusions we’ve drawn from our models — specifically, that the future success of a book, movie, celebrity, or Web site is strongly influenced by these types of feedback effects, and hence may to some extent be inherently unpredictable.
http://www.cs.cornell.edu/home/kleinber/networks-book/networ...
http://www.aspeninstitute.org/policy-work/communications-soc...
You might recognize the power distribution curve, as it's the same as the graph of the long tail:
http://en.wikipedia.org/wiki/Power_law#mediaviewer/File:Long...
What happens is that you have a handful of people who make a LOT of money / bust out virally, a small few who do OK, and the vast, vast majority starve. Since our economy is increasingly becoming an online, globally networked one, these effects are becoming stronger, and are a contributor to economic inequality.
Power Law phenomenon on the net has been observed for ages. Here's a post from 2003 where Kottke notes the distribution in the popularity of blogs on Technorati: http://kottke.org/03/02/weblogs-and-power-laws
I imagine the curve fits similarly for things like app store rankings, Reddit and Hacker News post popularities, top Steam sellers, Amazon rankings, etc.
> The world still rewards value, even if it takes some time. The people clamoring that it doesn't are doing so because they want to believe that it isn't their fault they didn't succeed.
The funny thing is, this is not entirely true. Quality is only important up to a certain threshold, after which you're at the mercy of what are essentially chaotic network effects early in the lifecycle of your product.
Salgankik, Dodds, and Watts performed an experiment that begins to provide some empirical support for this intuition [359]. They created a music download site, populated with 48 obscure songs of varying quality written by actual performing groups.
Visitors to the site were presented with a list of the songs and given the opportunity to listen to them. Each visitor was also shown a table listing the current “download count” for each song — the number of times it had been downloaded from the site thus far. At the end of a session, the visitor was given the opportunity to download copies of the songs that he or she liked.
Now, unbeknownst to the visitors, upon arrival they were actually being assigned at random to one of eight “parallel” copies of the site. The parallel copies started out identically, with the same songs and with each song having a download count of zero. However, each parallel copy then evolved differently as users arrived. In a controlled, small-scale setting, then, this experiment provided a way to observe what happens to the popularities of 48 songs when you get to run history forward eight different times. And in fact, it was found that the “market share” of the different songs varied considerably across the different parallel copies, although the best songs never ended up at the bottom and the worst songs never ended up at the top.
Salganik et al. also used this approach to show that, overall, feedback produced greater inequality in outcomes. Specifically, they assigned some users to a ninth version of the site in which no feedback about download counts was provided at all. In this version of the site, there was no direct opportunity for users to contribute to rich-get-richer dynamics, and indeed, there was significantly less variation in the market share of different songs.
There are clear implications for popularity in less controlled environments, parallel to some of the conclusions we’ve drawn from our models — specifically, that the future success of a book, movie, celebrity, or Web site is strongly influenced by these types of feedback effects, and hence may to some extent be inherently unpredictable.
http://www.cs.cornell.edu/home/kleinber/networks-book/networ...
That is just an amazing comment, with wide reaching applicability.
If this is true then it seems to be the important piece of content here. Recommendation algorithms are generally designed to show the user what they want to see, rather than an equal, fair, or diverse set of options, so it sounds like they're working as intended here. This type of analysis could be done with any characteristic, not just gender, leaving me unsure as to what the specific purpose of this article is - if users want X more than Y then it should be no surprise that X is presented more often or receives more traffic.