Sunday, December 01, 2013

Filter Bubble

Companies like Google, Facebook, Axiom and Blue Kai know a lot about you. They can use this to taylor websites and advertising to individual users. This can have serious impact on our society.

Huge amounts of information are produced every day. Determining what is relevant is a huge challenge. That is where algorithms come in. Google taylors search results to the individual users. Facebook's news feed is based on what it thinks users are interested in seeing. Advertisers and retailers also customize their online promotions based on what they think users will be interested in seeing. This can cause people to reside in a "bubble" where they only see things that they like. They may not see opposing viewpoints or anything that allows them to think. This is even more worrisome since most of this data filtering is done transparently without users knowing they are living in a bubble. (Perhaps this is a form of "mind control" where the big companies can gradually nudge people in a direction they would like.)

Personalization is treated as a "black box" by most companies. The companies themselves may not even know exactly what results are being returned. They can tweak algorithms based on feedback. However, they probably could not tell exactly what type of results would be returned. The algorithms can use various different data points to personalize. Thus, even if a user is not logged on, their location, computer or web browser could identify them and provide personalized results.

This book rambles on to sound the alarm against "filter bubbles" that allow individuals to live in their isolated worlds filtered to provide what they want. This will keep out information from opposing viewpoints. It will also tend to stock them up with the most "sensational" junk-food content rather than the "good for you content." This could result in a dumbing down of society as people don't work their brains to get around new thoughts. Also, it can make it more difficult for new thoughts and media to get out there. (Since discovery is often based on "likes", only things most like what already exist will tend to get more exposure. Thus, even though it is theoretically easier for new things to be discovered, it is actually much more difficult to find innovation.

The filter bubble is often transparent to users, and can gradually steer content to more extreme viewpoints. This can lead to highly cantankerous partisan discussions. (Since each side is not exposed to the opposing view, they not even understand how others can feel that way.) Even worse, people wont realize they are living in the filter bubble and assume that everybody else is receiving the same information. Discovering "new" ideas can actually be more difficult than it was in the days when everybody was force-fed the same broadcasts.

I was expecting this book to be a discussion of the difficulty we have today of "filtering" through all the information out there. Instead, it focussed on the danger of a personalized web. The two are tightly related. However, this book seemed to spend a lot of time rambling from bullet point to bullet point. It contained plenty of good ideas, but the connections where not very strong.

The amount of information out there is enormous. Discovering useful information is becoming more and more difficult. The quantity of "junk" out there seems to be growing at a faster rate than the amount of useful data. Fifteen years ago, it was easy to put a website out there and get visitors interested in the content. You might get a few random spammers or bots, but most traffic was legitimate. Similarly, if you wanted to search for something, you could use one of the numerous search engines and find relevant sites. The search results may contain a bunch of sites that you were not interested in, but this was more a result of bad algorithms than bad sites.

Today, however, there is so much junk out there. You may have to wade through numerous spam and junk sites to get to the site you want. People are much less likely to find a site that somebody just put up. I see more traffic on this blog than on my proto-blog from 1996. However, the quality of traffic seems to be much worse (at least judging from the ratio of real comments to spam comments.) It is harder to discover quality new content. And it is harder for quality new content to be discovered.

I find myself spending more time on "established" sites. They may occasionally guide me to independent sites. However, they are more likely to simply direct to other well-known commercial sites. With so much information out there, curation has to be done somewhere. I don't have the time to do it (or even to create an algorithm to do it.) I'm dependent on somebody to do it for me. Since this is a huge undertaking, these "somebodies" will likely be large corporations that need to earn money. Since I am cheep, putting up with advertising is my most likely "payment". This puts me in a position vulnerable to being influenced by whatever the corporation or the advertisers desire. (Ironically, at the same time the internet is giving away unlimited content for the price of advertising, broadcast media has become more reliant on "subscriber fees" as part of its business model.) Thus, we become subject to whatever whims the big algorithms have. Is this really much different from being beholden to the broadcaster's desires? At least with the broadcasters, we were likely to find something new we liked. With personalization we can find ourselves further ghettoized. (I'm often finding that problem with online radio. I can create a station that plays songs I like within a very narrowly defined range. However, I get sick of the same type of music and want more variety. However, it is difficult to get variety without a bunch of junk. I'd almost prefer to have a DJ picking the music for me.) I still haven't found a recommendation engine that does a really good job. With the glut of information out there, one of the big challenge is filtering the unique from the derivative. Perhaps now is the time to reinvent the web.

No comments:

Post a Comment