Log in

No account? Create an account

Joshua Rhys Taliesin O'Madadhain

the "filter bubble": a response

Joshua Rhys Taliesin O'Madadhain


the "filter bubble": a response

Previous Entry Share
This has been making the rounds recently:


The basic premise is that the information that you see on the web is becoming more personalized all the time:

these engines create a unique universe of information for each of us -- what I've come to call a filter bubble -- which fundamentally alters the way we encounter ideas and information

and this is problematic for the following reasons:

(1) "You're the only person in your bubble. In an age when shared information is the bedrock of shared experience, the filter bubble is a centrifugal force, pulling us apart."

(2) "The filter bubble is invisible...Google's agenda is opaque. Google doesn't tell you who it thinks you are, or why it's showing you the results you're seeing. You don't know if its assumptions about you are right or wrong -- and you might not even know it's making assumptions about you in the first place."

(3) "You don't choose to enter the bubble...personalized filters...come to you".

Google is not the only organization that comes under fire in this article--Facebook and Amazon also are mentioned--but Google is mentioned most prominently and consistently, presumably because it is the entity through which many people derive most of their information.

I'll start off by saying that much as I think that Google is cool, I hope that it never gets a monopoly on providing search results. It's important that there be viable choices--four or five good ones is a good start--for this service. We can make mistakes, and we won't always do the best job with every kind of information retrieval task. And if there's just one option, it makes it much more feasible for organizations to get access to information that they shouldn't (IMO). Defense in depth, memetic diversity, and all that--these things are critical to the healthy functioning of society.

I've also made specific proposals at work to suggest that we do some random perturbations of our search results in situations where our ranking algorithm doesn't really differentiate much, score-wise, between results, precisely to give our results a bit more diversity when it doesn't cost us anything (as far as we can tell).

All that said, I don't think that Mr. Pariser has really either (a) thought this through all the way nor (b) done his homework, at least so far as Google is concerned.

Before I respond to his specific points above, I'll point out that personalization is an incredibly useful approach to making any tool more useful, and search engines are no exception. If Google can determine that when I search for "latex" I am looking for information on typesetting rather than rubber fetishwear, so much the better. There's a universe of information out there and I don't have time to drink the whole thing. By all means, please do your best to give me the most relevant stuff.

Responding to his points in order:

(1) This is ridiculous. It's like saying that I am the only one who occupies this particular location in space-time and therefore I am isolated. Even if we suppose that I am the only person that has my particular combination of traits that Google measures, there are millions of people out there whose characteristics overlap strongly with mine. It's even stronger than that: customizing my information flow to my interests can help me discover people and communities and information to which I have existing implicit affinities that I might never have been able to run across otherwise.

(2) Dude. Do a search on "how to turn off personalized search". This will give you information on both how to do this and how to find out what kinds of factors are involved in personalization. This is not a new topic.
Could we be surfacing this better? Probably. But we do have to make some decisions about what information to give prominence to, or the search page gets incredibly cluttered. But the process (essentially, click on "view customizations" at the bottom of your search) is not difficult.

(3) There is some justice to this claim: I believe that search personalization is the default and while one can opt out, I don't know that we tell you up front when you create a Google Account that your results will be personalized.

That said: Google is not, and does not aspire to be, the _source_ of all information. We don't, for the most part, create content. We curate it, organize it, and try to make it accessible and useful. You, as a sentient being, are responsible for managing your inputs. We may decide that if you search for "rainforests" that you're more interested in organizations that want to preserve them unchanged than in organizations that want to extract the maximal resources from them at minimal cost, but nothing prevents you, or should, from searching for "rainforest preservation" or "rainforest resource extraction" instead. I don't really think that it's Google's job to help you break out of your rut.

reply at http://jrtom.dreamwidth.org/302129.html ( comments)
  • Does Google "curate" its news search results in the same way that a newspaper does? Does it have a team of editors who make judgement calls on publishing important news items? Is there anyone in charge of news.google.com who says, "You know, this news story isn't popular, and it's not relevant to a lot of people's personalized searches. But it's important, and we have an ethical obligation to bring it to people's attention."

    I would be very surprised if that were the case. And because I'm a savvy news consumer, I don't get my news exclusively from Google. The danger here is when people don't question the content they're seeing online. As more and more people get their news online, more and more of them will make the assumption that they're getting the same kind of story selection that they'd find in a newspaper. They're not.
    • Oh and also: Hey, it's really good to hear from you! Glad the family's doing well!
    • I don't know what criteria that Google uses to decide which articles to show either in the default view (no search terms) or in the case that you've specified search terms. Google has experimented with using human editors to highlight specific articles (http://www.niemanlab.org/2010/06/google-news-experiments-with-human-control-promotes-a-new-serendipity-with-editors-pick/) but I don't know what the status of that experiment is.

      Google does say a few things about how they choose stories here:

      Google has many decisions to make (which stories to cover, which (default) categories to offer, how to place stories, which sources to pay attention to, etc.) that have nothing to do with personalization, though. So I'm not sure that personalization is really the primary thing that one ought to be concerned about in the context of news.

      That said, I hope that I didn't come off above as suggesting that people should get all their information through Google. I don't, and I don't think that anyone should use a single source. I do believe that Google's news search does a pretty good job of giving you a reasonably broad spectrum of news (for unspecified searches) and of presenting news result that are relevant to a particular query. But it's still each person's job to ask the right questions.

      Side note: is there any newspaper whose editorial policy you know?
  • I know I've found it annoying that Facebook does not really give me any useful way to help train it on what I would consider "top news" for the top news feed. I appreciate the idea that I can filter down from the zillions of pseudofriends to the stuff that's actually important, but it seems to present me with a lot of stuff in Top News that I don't care about. I sometimes go over to Most Recent and hit Like on a few things that seem more relevant, in the hope that this will eventually teach it to work better, but I'm not terribly optimistic; it seems like the chief driver of Top News is number of comments and likes, which means that popular Pages get driven to the Top News feed frequently. Eventually I may just start stripping out all of the various random things I "like" (movies, bands, political figures, etc), and de-friend the countless mild acquaintances, in the hope of seeing more content about my actual friends...
    • I don't use Facebook, so I don't have any personal experience of what you're saying, but it's kind of what I would expect given the data that Facebook has (i.e., demographic and social).

      (Interesting that you are coming at this from the other direction, i.e., that it's not personalized _enough_ for you. :) )

      It's possible that Facebook is more susceptible to the echo chamber than (say) Google for that reason. Most people don't make a point of cultivating friends who have substantially different opinions and interests from their own (and it's more difficult than seeking out diverse sources of information), and if your friends _are_ your source of news...
      • Well, I don't really use Facebook for news at all; for that, I use Slate, the NYTimes, and a variety of political / econ / finance bloggers, whose links I will generally at least skim.

        What I want from Facebook is, basically Shorter LiveJournal -- I want it to round up status updates and photos from people whose lives I care about, and only occasionally show me something from outside that pool that gathers unusually large numbers of likes and comments. The judgement of "unusually large" should be relative to the number of likes and comments that particular stream gets, not compared to other unrelated things. Posts from, say, the official Rachel Maddow Show page, tend to get more comments than posts from my friends, across the board. If Rachel posts something that generates double the comments of her usual post, then maybe I want to see that.

        In any case, again, my complaint is fundamentally that hiding posts I'm not interested in, and going to the Most Recent page and liking posts of the kind I'd like to see more of (that had not been showing in Top News), does not seem to change what it shows me. Their algorithm does not pay attention to its own failures.

        ETA: I'd also add that friends who barely ever post, should just automatically be regarded as Top News when they do, whereas folks who post twelve times a day should have to gather a few likes and comments. (There's at least one person whom I like a good deal, but got dropped from my daily-reading LJ filter because of ten posts a day of content-free chatter driving everybody else off my front page.)

        Edited at 2011-06-03 05:11 pm (UTC)
Powered by LiveJournal.com