Back in a 2010 blog post, I noted the kinds of “echo chamber” effects that can result from personalization and targeting of various types of information on the Web. That particular posting concentrated on search personalization, but also noted the impact on Internet-based discussions, a situation that has become dramatically more acute with the continuing rise of social media. Given the current controversies regarding how “filter bubbles” and other algorithmically-driven information surfacing and restriction systems may impact users’ views and potentially increase political and religious radicalization — particularly in relation to the 2016 elections here in the USA — I believe it is relevant to republish that posting today, which is included below. Also of potential interest is my recently reposted item related to Internet fact checking.
Search Personalization: Blessing and Trap?
(Original posting date: September 16, 2010)
Greetings. Arguably the holy grail of search technology — and of many other aspects of Internet-based services today, is personalization. Providing users with personalized search suggestions, search results, news items, or other personalized services as quickly as possible, while filtering out “undesired” information, is a key focus not only of Google but of other enterprises around the world.
But does too much reliance on personalization create an “echo chamber” effect, where individuals are mainly (or perhaps totally) exposed to information that only fits their predetermined views? And if so, is this necessarily always beneficial to those individuals? What about for society at large?
Diversity of opinions and information is extremely important, especially today in our globally interconnected environment. When I do interviews on mainstream radio programs about Internet issues, it’s usually on programs where the overall focus is much more conservative than my own personal attitudes. Yet I’ve found that even though there’s often a discordance between the preexisting views of most listeners and my own sentiments, I typically get more insightful questions during those shows than in the venues where I spend most of my time online.
And one of the most frequent questions I get afterwards from listeners contacting me by email is: “How come nobody explained this to me that way before?”
The answer usually is that personalized and other limited focus information sources (including some television news networks) never exposed those persons to other viewpoints that might have helped them fully understand the issues of interest.
An important aspect of search technology research should include additional concentration on finding ways to avoid potential negative impacts from personalized information sources — particularly when these have the collateral effect of “shutting out” viewpoints, concepts, and results that would be of benefit both to individuals and to society.
Overall, I believe that this is somewhat less of a concern with “direct” general topic searches per se, at least when viewed as distinct from search suggestions. But as suggestions and results become increasingly commingled, this aspect also becomes increasingly complex. (I’ve previously noted my initial concerns in this respect related to the newly deployed Google Instant system).
Suggestions would seem to be an area where “personalization funneling” (I may be coining a phrase with this one) would be of more concern. And in the world of news searches as opposed to general searches, there are particularly salient related issues to consider (thought experiment: if you get all of your information from FOX News, what important facts and contexts are you probably missing?)
While there are certainly many people who (for professional or personal reasons) make a point to find and cultivate varied and opposing opinions, not doing so becomes much easier — and seemingly more “natural” — in the Internet environment. At least the possibility of serendipitous exposure to conflicting points of view was always present when reading a general audience newspaper or magazine, for example. But you can configure many Web sites and feeds to eliminate all but the narrowest of opinions, and some personalization tools are specifically designed to enhance this effect.
As our search and related tools increasingly focus on predicting what we want to see and avoiding showing us anything else (which naturally enough makes sense if you want to encourage return visits and show the most “attractive” ads to any given individual), the funneling effect of crowding out other materials of potential value appears to be ever more pronounced.
Add to that the “preaching to the choir” effect in many Internet discussions. True, there are forums with vibrant exchanges of views and conflicting opinions. But note how much of our Twitter and Buzz feeds are depressingly dominated by a chorus of “Attaboy!” yells from “birds of a feather” like-minded participants.
I am increasingly concerned that technologically-based Internet personalization — despite its many extremely positive attributes — also carries with it the potential for significant risks that are apparently not currently receiving the research and policy attention that they deserve.
If we do choose to assign some serious thinking to this dilemma, we certainly have the technological means to adjust our race toward personalization in ways that would help to balance out the equation.
This definitely does not mean giving up the benefits of personalization. However, we can choose to devote some of the brainpower currently focused on figuring out what we want to see, and work also toward algorithms that can help determine what we need to see.
In the process, this may significantly encourage society’s broader goals of cooperation and consensus, which of necessity require — to some extent at least — that we don’t live our entire lives in confining information silos, ironically even while we’re surrounded by the Internet’s vast buffet of every possible point of view.
– – –
–Lauren–
I have consulted to Google, but I am not currently doing so — my opinions expressed here are mine alone.
– – –
The correct term is “Internet” NOT “internet” — please don’t fall into the trap of using the latter. It’s just plain wrong!