Make Your Own Pie: Google's Personalization Merely a Taste of Things to Come
By Andrew Goodman, 3/29/2004
Personalization is one of those hip online topics like weblogs, RSS, and streaming
music that no commentator has felt the power to resist over the past couple of
Following a period that generated vast expectations for search engines (and
an Internet in general) that will do a better job of giving you exactly what you want,
personalization-punditry finally fell on hard times. The domain Personalization.com
(a former hangout for Christopher Locke and other luminaries) is currently
up for sale. Personalization is no longer as chic as it was in part
because it's become a fait accompli in the world of e-commerce. (It's
become so uncool and so "everyday" that I had to use two French terms
in that last sentence to jazz it up.)
Today, Google has released a
modest attempt to show how personalization might work for a search engine
user. It seems hardly worth the trouble, but I argue there's a lot more
to this than meets the eye. It has a lot more to do with meeting the evolving
needs of advanced search engine users than it does with anything more broadly-conceived,
such as "creating better and better search results that will come closer
to guessing exactly what the user wants," or the type of personalization
that goes on for profit at e-commerce sites like Amazon.com, which can remember
their user's preferences and make useful suggestions for additional purchases.
What I'm trying to argue, then, is that this Google effort at "personalization" is
really a modest but powerful attempt to show how an advanced way of using the
Google index might work not just in function, but in form. The type of personalization
that Google's prototype allows, after all, is pretty minor -- it allows you
to check a few areas of interest, and that's it. No demographic preferences,
no learning your habits, not even very many subcategories of interest. In the
geographical categories, when I looked, North America wasn't available, although
the US and many subcategories of it were, which might mean that if you're Mexican
or Canadian, you're out of luck. In spite of Spring's arrival, I took the path
of least resistance and checked "Polar regions." So basically, someone
who wanted an advanced tool here would be out of luck.
But let's put aside somewhat the broader discussion of personalization and
futuristic mind-meld that seems to get pundits so glassy-eyed. Yes, I'm sure
that services like Eurekster which rely on social networking as one more way
of generating relevant results are doing something pretty innovative, but it's
also far from new (why aren't they calling it "P2P" search anymore?
did the VC pipeline dry up for P2P so now it's all about "social networking"?).
Cory Doctorow's group had high hopes for a service called OpenCola, for example,
but for whatever reason, that didn't take off. Maybe because there is only
room in the marketplace for so many stunningly good ideas. And the Googles
and the Microsofts of the world are going to take above average ideas and actually
execute them and get people interested enough for long enough for them to become
part of the mainstream.
How easily we forget that the Internet has been, in many ways, many times
over, the quintessentially good place for groups, storage of information, peer
review, and judgments of relevancy, and that a great many quasi-search services
have had a shot at cracking this "personalization" nut but also failed
to catch on. Even the online bookmarking services like Backflip and HotLinks (the
latter founded by the same man, Jonathan Abrams, who would go on to found social
networking service Friendster)
had a great idea that wasn't so different from P2P search, or social-networking-as-search,
or whatever the next generation of jargon might want to call it.
Speaking of domain names for sale, HotLinks.com, too, is on the block.
So back to Google's experiment. After entering my preferences and dutifully
typing the query "stanford" to use as an example, I could see exactly
how this worked. The "dial" or "slider" is employed to
set personalization between "min" and "max." Results that
count as "yours" are denoted by the little Google "molecule
balls" logo. For example, if I expressed interest in computers, an upcoming
public lecture on computing might rise higher as I increased the personalization
setting. The cute little logos look like reasonably authoritative "relevancy-denoting
Nerf balls" as they bounce up and down my screen.
Of course the actual service doesn't really work too well at all. Typing "john
stuart mill," I got several "personalized" results that moved
up the page as I increased the personalization. This was likely because I expressed
interest in "philosophy" and "government." Well with Mill,
it was always either about philosophy or government, so in this case, the exercise
was purely, shall we say, academic. No search engine could have helped me unless
I fed it the entire contents of my brain.
That isn't the point, though. I'm convinced that the elegance of this functionality
lies in the "slider" or dial, which allows the user to watch the
impact of changing settings on how the search results are ranked. Now imagine
several useful "sliders" that would allow users to tweak other things
about the search results. Would you rather see the search engine place more
emphasis on keyword density or less? How about pure link popularity vs. the
quality of links? Do you tend to search more often for products and services,
or more often for invisible web and government resources? Would you like extra
cheese with that?
For the past couple of years (such as in this article, Search
Engines are Still in the Model T Era, published in a European magazine
called Internet Markets) I've been hoping that search engines would consider
installing such "dials." Finally someone seems to have read my
Advanced users could play around with these knobs until they were satisfied
with the quality of results; or, they could tweak the dials every day if they
wanted, like an audiophile with an excessive fascination with his equalizer.
The idea of lightly prompting the search engine to respond better to your
personal interests could grow into a situation whereby advanced users would
be given enough control over the search engine's ranking methodology that they'd
almost feel like they were honorary Google scientists tweaking the algorithmic "mix." Smarter
users would have things tweaked so well that they'd see less spam. Search engine
spammers have more incentive to reverse-engineer an algorithm that they know
is being applied to the results that every user sees. If some users have "keyword
density" maxed and others give more weight to metadata, it would be like
trying to spam ten thousand different search engines. (And you thought that "ten
thousand search engines" thing was a myth.) And yes, why not bring Orkut
into the mix, too. Highlighting resources enjoyed by others who share your
interests or in your social circle could be another optional setting to improve
the relevancy of search results.
None of that's going to happen overnight. It introduces potentially horrendous
complexity to the management of the Google product. But we'll look forward
to whatever sliders, dials, and lab experiments Google sees fit to work on.
So that's the future as this glassy-eyed pundit hopes to see it: a search
engine that works like a sophisticated flight simulator, with a bunch of dials
and instruments formerly available only to classified personnel. But to the
extent that your settings become comfortable to you, it would be a flight simulator
operated largely on autopilot. Now that would be one sweet ride!
Andrew Goodman is Editor-at-Large of Traffick.com and the author of "Winning Results with Google Adwords".