Will The Personalized Web Filter Out Diversity?

About a year ago, I wrote that I had received a book for Christmas called The Big Switch by Nicholas Carr.  I’m a little embarrassed to say that it was only two weeks ago, on my post-holiday flight back to California, that I finally finished it after ten months of letting it collect dust.

Carr’s attitude about the rise of cloud computing, social networking, and all the other web 2.0 buzztrends caught me off guard.  While he seems to marvel at and mostly celebrate the speed and scale at which this phenomenon has grown and subsequently become integral to modern life, he does so with a very cautionary and sometimes pessimistic tone.  I guess I should have expected as much after reading his piece in the Atlantic last summer (hat tip to Joel for passing it along) called “Is Google Making Us Stupid?”.  Though I don’t share all of his concerns, it is refreshing to find a voice like his among the generally over-enthusiastic technorati (myself included).

Carr highlights a particularly interesting threat in the chapter he titles “The Great Unbundling.”  While granting that the jubilation over the democratizing and barrier-lowering nature of the Internet is mostly well deserved, he also warns that these characteristics come with potential harm.  What harm?  In short, Carr argues that the speed and personalization offered by the Internet facilitate our natural human tendency to cluster into isolated groups that can deafen and blind us to differing opinions.

Carr’s main evidence is an experiment conducted in 1971 by a Nobel Prize winning economist named Thomas Schelling.  The experiment consisted of a grid of squares and a set of white and black markers.  After randomly distributing these markers to represent an “integrated” community, he moved the markers around based on the rule that no one marker could have more than 50% of its immediate neighbors be of the opposite color.  This rule, Schelling posited, mirrored a natural instinct that humans possess: to be closer to people similar to ourselves than we are to those who are different.  After moving pieces one at a time based on this rule, he ended up with one all-white cluster and one all-black cluster.  Self-segregation, Schelling concluded, is the natural result of our instinctual preference to be closer to those that resemble us than to those who differ from us.

Extrapolating from these results, Carr hypothesizes that because the Internet removes so much friction from the market for information, self-segregation will happen much more quickly and completely online than, say, in a once-integrated neighborhood that slowly segregates over many years into two or more distinct socioeconomic or racial groups.  In my opinion, however, his interpretation of Schelling’s experiment, and his application of it to the Web, is flawed from the start.

First of all, Schelling assumes we have a preference that “50% or more” of our “neighbors” be more like us than different from us.  Can we trust this initial assumption?  I don’t contest that it’s only human to feel more comfortable when we’re surrounded by people like us.  But does comfort level always translate into preference?  Perhaps sometimes or often, but can we say this is categorically true?

What of those who make a point, out of an intellectual or moral or other conviction, to surround themselves with people who are different and disagree with them?  We are not lab rats, after all; we’re people.  And who is to say that the number is 50%?  What would the results of the experiment have been if that figure was 40%?  20%?  Or what if there were three colors?  Or thirteen?  I haven’t done enough research to see if Schelling iterated on his experiment with different rules and variables; but even if he did, Carr certainly does not mention it.

My second critique of Carr’s interpretation of Schelling’s experiment is his confusing of physical proximity with what I’ll call intellectual proximity.  It’s one thing to have a desire to live, work, and learn in the same physical region as others like us; but does this supposed bias necessarily have an intellectual parallel?  Again, I do not deny that, anecdotally, it seems to be a human tendency to read and watch and converse with those who are of the same mind as us.  But who is to say this is a systematic preference?  Again, what of those who resolve to immerse themselves in opposing viewpoints in order, say, to challenge and strengthen their own?

Carr is also particularly troubled by the rising trend of “personalization algorithms and filters” that seek to “understand” us and deliver to us the information that they think we “want.”  Take the recommendation services of Netflix, Amazon, or iTunes as examples; these companies take your purchase and browsing histories, run them through complex mathematical formulas and models, and suggest products that it thinks you will enjoy.  The longer we use these services, the better they can build a detailed “profile” about us, and the more accurately they can assess, target, and satisfy our specific interests.  Carr notes, however, that this could lead to a disturbing snowball effect of recommendations-feeding-consumption-feeding-recommendations that isolate us within our own interests.  The recommendations are based on history, but who is to say that our history is always the best indication of our current or future desires?

All of these fears assume that humans do not have the mental presence or rigor to ensure that we provide ourselves with a healthy intellectual diversity in the face of recommended or automated personalization.  Like in Schelling’s experiment, it is assumed that we are merely complex machines that react to stimuli in a pre-determined way.  I have more faith in humanity than that.  This is not to say that there is not ample evidence to support fears of cybersegregation.  We saw it in the presidential election last year, for example, with e-mails being circulated among ultra-conservative communities suggesting that Barack Obama is a Muslim or even a terrorist, reinforcing a dangerous groupthink that ignored the widely publicized truth.  We — as humans but especially as Americans — are often lazy when it comes to fully exploring every facet of a situation.

But the hypothesis that the trend towards personalization and filtering could lead us all into — as Carr puts it — “cultural impoverishment and social fragmentation” seems a little far fetched.  While we might indeed have a tendency to congregate physically and mentally with those who resemble us, I believe that there is also a natural human inquisitiveness and skepticism that will prevail.  At some point, echo chambers always harmonize to such a precise frequency that at least one member will start to think it’s too good to be true, and they will seek new sources of information, and the cycle will shatter.

Though this natural inquisitiveness or skepticism may not always prevail on the individual level, I believe it will almost always end up happening on the community level.  Especially in democratic societies, there will always be individuals and groups who make it their mission to make sure that opposing points are heard and considered.  This, after all, is one of the benefits of freedom of speech and thre free press, and is another reason that while newspapers may die, I don’t think professional journalism ever will.  Carr also never mentions the boon that online communication brings to the forging and solidifying of groups that do good in our world, such as those organizations that focus on the environment or social justice or philanthropy.  While it is important to constantly seek the truth, I think it is also important and healthy to find and reinforce one’s identity within a community of truth-seekers.  The web facilitates our ability to do just that.

And so I arrive at my normal conclusion for debates like these.  Moderation.  Is.  Key.  Personalization and filtering are awesome features with epic potential for helping us discover more about the things we’re interested in.  And though I think our nature will never lead us to that point, I believe it would be a grand mistake for us to completely outsource our judgment of our own desires and preferences to machines.  Our ability to decide for ourselves what we like — and what we think is the truth — is part of what makes us human, after all.