Matt posts on the role of humans in Google search, prodded by a NYT story on the topic:
If you ask an average techie about Google, you’ll hear that we use lots of computers and algorithms. Indeed, the title of the New York Times article is “The Human Touch That May Loosen Google’s Grip.” But (in my opinion), it would be a mistake to think “Google is nothing but cold algorithms and computers; there’s no room for humans at all.” I’ll give you a few examples of the role of people over the years at Google:
– PageRank is fundamentally about the hyperlinks that people on the web create. All those people creating links help Google formulate an opinion of how important a page is.
– Google News looks at a wide variety of news sources; the decisions of human editors at thousands of news sites help Google estimate whether a particular story is significant.
– Google introduced voting buttons on the toolbar back in 2001. They look like happy/frowny faces and they let regular people send thumbs-up or thumbs-down votes to Google.
– Google has allowed users to remove results that they don’t like from Google.
– For more than five years, we’ve allowed users to report spam to Google. We’ve said for years that we reserve the right to take manual action on spam (e.g. if someone types in their name and gets off-topic porn as a result).
And of course, it’s not as if Google’s search engineers drive into the Googleplex in the morning and then spend the whole day sitting around doing nothing while the computers do all the work. 🙂 Instead, Google researchers and engineers spend our days looking for deeper insights that will let us create the next generation of search. I believe Google’s approach to search has always been pragmatic: if an approach will improve the quality of our search, we’re open to it.
He also refers to the interview we did together last year.
Automation also helps Google integrate human intelligence in real-time, something no individual person could do. Yesterday, for example, my wife searched for “Silver Surfer” on Google (on her Treo) and the first thing she got was show times. Linguistically, it was a very ambiguous query, but by being able to integrate the behavior of millions of people in real time, Google was able to successful infer that we were interested in the recently released movie.
I’m also very curious how the human-powered engines (not the neo-directories, but the operator-driven engines like Cha-Cha) will be able to provide consistent service. The human mind is an amazing thing, but your search results are only going to be as good as whatever operator you happen to get. Having called customer service reps frequently in the last 5 years (both domestic and outsourced), I’m not optimistic.
I am very happy that slowly but surely people seem to be catching on to how community search might work (even though there is still a long way to go).
Of course I would not interchange the community of sciam.com subscribers with those of del.icio.us and/or digg.com (or whatever). Neither would I try to compare sciam.com with harvardbusinessonline.hbsp.harvard.edu (and/or marketing.org with movie.de) — that would be comparing apples and oranges.
(part 1/2)
Most of the stuff being touted as “social” and/or “community” search is in fact hardly “focused” at all. I have no idea whatsoever what I might have in common (which might make me a member of the same “community”) with millions of other people that might be users and/or members of this/that 2.0 bite-sized platform.
(part 2/3 [oops!])
And most of the present-day “one-size-fits-all” algorithms will become a quaint reminiscence that we can tell to our grandchildren (and then they will shake their heads and think we must have been living in caves or something like that).
Then again, perhaps there are other things that might be even more humorous to reminisce about.
(part 3/3)
Collarity still believes in your “data base of intentions” philosophy. We believe the best people-powered search feedback is buried in Apache log files in the form of implicit, anonymous behavioral data. Human-driven search relevance will be better reflected through what people do, and less by what they say/write.