Bing released its top searches of the year today, continuing the trend of presuming the year ends before December begins (watch for Yahoo and Google’s lists in the next week or so). Once again, the data is utterly uninspiring and shallow. I mean, did we really not know that the US is fascinated with celebrities and iPhones?
This is becoming something of a trope for me, but given all the data to which search giants like Microsoft and Google have access, I’d love to see some real data science being applied – find us the conceptual scoops, the insights, the second and third order trends. Is that too much to ask?
Sadly, most of the audience seems to prefer the things that you and I would consider boring.
“Is that too much to ask.”
Actually, I think it _is_ too much to ask. I haven’t commented on your blog in a while, but your frustration about the data being “utterly uninspiring and shallow” resonates with the core of many of the comments I’ve made over the past 5-6 years here.
Here’s the problem as I see it: Bing, Google, Yahoo, etc work very well at returning answers to your questions that are the “head of the head” of the distribution. The largest box store on the biggest main street. That’s just what these algorithms do. All the signals they use.. from inlink popularity to training on click popularity, their machine learning learns the head of the head of the distribution. Forget the fat belly, and especially even forget the long tail. You won’t find those results at ranks 1-3, 4-10, or even 11-40. And especially not beyond 40.
Now, because these search engines don’t do very well at fat belly and long tail needs, I think what has happened is that users themselves have subconsciously learned to ask certain types of questions that they know will work, and no longer ask certain types of questions that never seem to work. I know you know what I’m talking about: for certain types of questions, these search engines can do no wrong and return answers as if my magic. But for other types of questions it doesn’t matter how many different ways you phrase it or through how many dozens of results pages you browse, you’re just not going to find the deeper, more complex information that you seek.
People (searchers) notice this dichotomy. And I think they stop asking those sorts of questions that don’t yield easy, snappy answers. That require more in-depth exploration and pattern discovery. That have more than one right answer. Over the last dozen years, searchers have subconsciously been training themselves to NOT ask what doesn’t have quick and easy, head of the head answers.
And then, of course, the more that people ask easy-answer, head-oriented questions, the better the Google learning algorithm becomes at answering them, and the worse it gets at answering anything else. It is a vicious loop.
That’s a long way of me saying that conceptual scoops, insights, and second and third order trends are too much to ask for. The head of the head “tails” off very quickly, and there just isn’t much there to play with, no matter how big the data is, because of how the system and the users have mutually adapted themselves to each other.
Interesting, JG. Where do you go to get those insights then? Any suggestions?
I think the problem (diversity) is the solution (diversify) is the problem (diversity). The solution is all around us. And by that, I mean, all around us. Aka, source diversification. I usually solve the problem by going to multiple sources and not always asking one source, no matter how hard they’re trying or smart they are. I’ll ask the question of my friends on FB, check G, Y, Bing, and even Ask, and even (yes still!) go to the library and browse the shelves. I’ll also incorporate word-of-mouth into my information seeking processes. And I’ll keep up to date on all the professional blogs and tweeters.
That’s not too different from what all of us do, anyway. But the point is that I consciously try and diversify my sources as much as possible, and get as many different viewpoints as possible. I make it a conscious choice, an active decision. I consciously try to not just use one search engine, but to rotate and vary my usage as much as possible. I just wish that there were tools to automate this rotation, because what I really want is a personal tool that keeps track of from which sources I get the best answers for different types of questions. A personal monitor and tracker, locally installed on my own machine even, that automatically federates my querying to multiple sources, and then lets me inspect and make decisions about how to deal with the diversity of results that come back from each of those sources.
I remember a decade ago, Google used to make it very easy for one to rotate one’s usage. They had links at the bottom of the first SERP page to try the same query on a competitor’s site. Those are gone. And they also make it difficult for me to manually search multiple search engines at once, and make my own diversification choices in a post hoc manner. Go read the Google terms of service, and to this very day it still says, “You may not “meta-search” Google.” You can’t get more anti-web 2.0 than that. But if I were allowed to do that, if I weren’t locked in to using it exactly and only how they want me to use it, if Google’s SERPs were really as free as open and Web 2.0 as they should be (translation: re-usable, share-able, reproducible, re-mixable), then I think that would go a long way toward alleviating the current dominance of the monoculture mindset.
Note that this isn’t a specific-to-Google problem. The problem arises when any person sticks to any single source for the brunt of their information seeking needs. The same problem arises when one only sticks to Y!, or only sticks to Bing. Google’s just a great example of the problem, because they’ve been doing it for the longest and have the deepest systematic entrenchment if that vicious loop. Somehow we have to escape the mindset of “it’s better because more people click, and more people click therefore it’s better”.
Perhaps I am naive, but I hold the notion that if we all had personal tools that helped us maintain diversity in our personal information seeking, rather than everyone only going to one tool/source, it would create an environment in which a diversity of vendors could arise to address different aspects of everyone’s information behavior. Google could handle the vapid, Britney Spears queries, Amazon and related competitors the transactional queries, and some new as-yet-to-exist, more librarian-oriented competitor the “what were the historical antecedents of the 2008 financial collapse” queries. That search engine that really lets me dig deep and find statistics and other sources on whether the collapse was caused by the government forcing banks to make loans to poor people (as libertarians and conservatives claim) or whether it was caused by corporate greed and fiscal irresponsibility (as liberals claim). That latter search engine would really let me compare and contrast sources, link various pieces of information together and train the engine myself on which items are related, etc., so as to really be able to dig in and let me chew on my question.
But since there are not a diversity of vendors, because there are no tools that allow all vendors to simultaneously exist, we all end up going to a single source (whether it be G, Y, or B) and therein lies the seeds of vapidity.
My analogy is to a farmer’s market vs. a supermarket. Google is a supermarket. You can get everything, but it’s controlled by one company, often with that company’s own competing brand next to national brands. And nothing else.
A farmer’s market, on the other hand, let’s multiple small vendors thrive. I get my honey from the honey guy. My berries from the orchard lady. My root vegetables from the organic farm. Different needs, different desires on my part, and different vendors to meet those needs. Diversity of sources.
We need a tool that acts as a farmer’s market, a tool that creates that information marketplace. Not a tool and funnels and amplifies everyone until only the head of the head of the distribution (the vapid results) rise to the top.
Love the farmer’ market metaphor. I think we need to apply that concept to a lot more than search….
Agreed that we need to apply it to a lot more than search. But many of those movements already exist. Alice Waters has been around for decades. Farmers’ markets and CSAs are on the rise, nationally. But a “slow food” search engine? When is that going to appear? And more importantly, where is the grass-roots behavior that needs to first exist, to call for a slow food search engine? I rotate my engines on a regular basis, using many different engines over the course of even a single day. But how many other people, even tech savvy people, do the same? A farmers’ market needs customers, people who will actually go there and shop at multiple vendors. How many of us are behaving in our search engine usage like good farmers’ market patrons? Very, very, painfully very few. We’re all still addicted to our informational Burger Kings and McD’s. Fast and easy. (Any surprise, btw, in how much Google emphasizes speed, speed, speed? See the parallel to fast food?)
Hmm.. these comments appear in reverse order. Read from the bottom, first.
“your frustration about the data being “utterly uninspiring and shallow” resonates with the core of many of the comments I’ve made over the past 5-6 years here.”
See for example my comment about the Found conference: http://battellemedia.com/archives/2008/10/oreilly_launches_found_conference.php#comment-335309070
As search drives more people to main street, main street adapts to more people, which makes search drive even more people to main street, and so on. The thousands of ma and pa grocers just around the corner in your own neighborhood, where you used to stop and get eggs and milk, is now a 1950s relic, replaced by immense mega-grocery stores with football field-sized parking lots.
This shift has been happening in the physical world for decades, and now search engines are causing it to happen at exponential speed in the digital world.
The data is uninspiring and shallow, because that’s where the algorithms drive the searchers. Then the searchers in turn only try to drive to where the data is uninspiring and shallow.
It would be strange and a bit sinister if a list of the year’s most popular searches wasn’t predictable. It’s a non-story. Bing’s year-end summary is like being ceremoniously presented with a map of your own household furniture: yup, there’s the sofa, in the same place it’s been all along …
Only this year Justin is on it, rather than Britney.