Difference between revisions of "Filter bubbles"
| Line 1: | Line 1: | ||
| − | + | [[File:Filterbubbles.jpg|right]] | |
Filter bubbles or echo chambers are what happens when the information you come across reflects your history so well that you stop learning anything new. They're a big problem for [[Knowledge Management]]. | Filter bubbles or echo chambers are what happens when the information you come across reflects your history so well that you stop learning anything new. They're a big problem for [[Knowledge Management]]. | ||
| Line 9: | Line 9: | ||
At least part of the reason for relatively weak mixing processes on the internet is that algorithmic methods tend to chase marginal improvements in precision, using relatively fixed social and taxonomic data with low degrees of freedom as a conduit for mixing processes, and build highly complex systems that can't keep up with a rapidly changing flow. But there are relatively simple ways around some of these things, more in line with balanced natural systems (for example, better use of clickstream data), so I expect that filter bubbles are not the end of the story. | At least part of the reason for relatively weak mixing processes on the internet is that algorithmic methods tend to chase marginal improvements in precision, using relatively fixed social and taxonomic data with low degrees of freedom as a conduit for mixing processes, and build highly complex systems that can't keep up with a rapidly changing flow. But there are relatively simple ways around some of these things, more in line with balanced natural systems (for example, better use of clickstream data), so I expect that filter bubbles are not the end of the story. | ||
| − | + | [http://www.pontneo.com/ctf/ {!ctf}] is all about increasing mixing feedbacks to balance convergence. | |
Revision as of 16:49, 26 October 2018
Filter bubbles or echo chambers are what happens when the information you come across reflects your history so well that you stop learning anything new. They're a big problem for Knowledge Management.
They reflect self-reinforcing positive feedbacks, which accelerate towards convergence. They act to split areas of convergence from one another and generate heterogeneity in information flow. In natural systems, these positive feedbacks are held in check by opposing mixing processes, which cause divergence instead of convergence, spreading flows out instead of concentrating them. Filter bubbles emerge when those spreading processes are relatively ineffective and so convergence comes to dominate the flow.
Up to a point, you could regard filter bubbles as a natural response to enormous flows passing through parts of the internet. But filter bubbles are not necessarily the most productive or efficient way to accommodate that flow, and clearly they have other important consequences too (e.g. political).
At least part of the reason for relatively weak mixing processes on the internet is that algorithmic methods tend to chase marginal improvements in precision, using relatively fixed social and taxonomic data with low degrees of freedom as a conduit for mixing processes, and build highly complex systems that can't keep up with a rapidly changing flow. But there are relatively simple ways around some of these things, more in line with balanced natural systems (for example, better use of clickstream data), so I expect that filter bubbles are not the end of the story.
{!ctf} is all about increasing mixing feedbacks to balance convergence.
