The availability of search engines like Google provides an easy way of checking on factual claims you may find questionable – just enter the relevant keywords into a search engine and see what comes up. If such a search produces nothing to support the claim, or evidence to refute or qualify it, then it’s time to start demanding evidence.
This started me thinking about a more general problem with search engines. Using search engine results in the way I suggest rests on the assumption that a given query will produce given results. The same is true if I want to say “Site X is the top result on engine Y for query Z”. But what happens if, as is already possible, search results are personalised, based on, say, previous search history and choice among search results. The same search, undertaken by someone else, might produce completely different results.
Personalisation has some obvious benefits. if I’m searching for bus routes in Brisbane, I probably don’t want results about Brisbane, California. But it undermines the usefulness of search engines results as evidence in analysis or argument.
Full-scale personalisation might get us to the point feared by writers like Cass Sunstein. Dogmatic leftwingers or rightwingers, supporters and opponents of the Iraq war, and so on, might be presented exclusively with search results that confirmed their prejudices, and might never realise that they were looking at a completely different Web to that seen by someone with different views. This process would work only for people who usually don’t follow search results that lead to views contrary to their own – personalisation would reinforce this tendency until it became automatic.
{ 1 trackback }
{ 13 comments }
Seth Finkelstein 11.21.05 at 5:44 am
In terms of a beautiful theory slain by an ugly fact, the answer is going to be “Just Google-General it”.
Yes, there will be mistakes where people confuse Google-Personal with Google-General. But Cass Sunstein’s alarmism is based on a flawed model of extremism. It’s a model very appealing to his audience, since it basically tells them that they are the philosopher-kings who must educate the rubes for their own good (and I sympathize!). But in fact, the extremists usually understand full well their opposite counterparts, and vehemently think they’re wrong (e.g. white supremacists often know far more about African-American history than apolitical people).
That is, when personalization becomes widespread, people will incorporate that assumption into their response, IF IT IS RELEVANT AND THEY ARE AWARE OF IT. There’s of course some play both those factors. But the specific scare-mongering over political personalization is more punditry than problem.
Ginger Yellow 11.21.05 at 6:04 am
“But it undermines the usefulness of search engines results as evidence in analysis or argument.”
Frankly I don’t think search engine results count for much in an argument, so I’m not sure we’re losing much on that front. Far more worrisome would be the further fragmentation of the web into ideologically homogeneous enclaves. But I think the likelihood of this outcome is overstated. I don’t use Google to find things I might agree with politically; that’s what blogrolls are for. Google is for generic info-hunting.
g 11.21.05 at 7:47 am
I’m sure it’s true that white supremacists often know *more facts* about African-American history than apolitical people. But knowing more facts doesn’t equate to *knowing more*, if the facts are misleadingly selected — as I suspect they often are in this instance. (But I’m willing to be corrected by those who know more about white supremacists than I do.)
Eszter 11.21.05 at 8:22 am
Yes, there will be mistakes where people confuse Google-Personal with Google-General.
I suspect this will be the default not the exception. I don’t know of systematic studies of this particular question, but based on related studies, I suspect most people do not and will not understand that the search results they are getting are based on their previous online actions and thus are personalized and not generalizable to other users’ experiences.
Eszter 11.21.05 at 8:28 am
Oh, and I do think that search results do count for something in an argument and may be used more and more. The title of this post is very appropriate. I suspect many of us have heard people say “just Google it” when referring to the answer to a question.
Ginger Yellow 11.21.05 at 8:46 am
Yes, but that’s almost always in reference to some piece of information that is easily Googlable, saving the respondent the trouble of explaining it in detail. Personalisation won’t change that. It’s got nothing to do with saying “Google turns up x hits when you search for y” or “The top hit for x is y”, which would be affected by personalisation but strike me as completely pointless factoids to bring up in an argument.
Henry 11.21.05 at 9:22 am
Seth – in fairness to Cass (whose polarization thesis I mostly disagree with), he says that problems of polarization, information cascades etc are “endemic among the Philosopher Kings”:http://papers.ssrn.com/sol3/papers.cfm?abstract_id=262331 too.
Richard Bellamy 11.21.05 at 9:29 am
I suspect many of us have heard people say “just Google it†when referring to the answer to a question.
Yes, but more often to questions like “Who the heck is Katie Holmes?”
Rarely to a question like “Which side is correct in the Israeli/Palestinian conflict?”, or “Was the war in Iraq justified?”
Jacob T. Levy 11.21.05 at 9:42 am
Rarely to a question like “Which side is correct in the Israeli/Palestinian conflict?â€, or “Was the war in Iraq justified?â€
Well, those questions are straightforwardly normative.
But there are both contentious (“Do increases in the minimum wage increase unemployment?” “What are the effects of school vouchers on test scores?”) and pseudo-contentious (“Did Israeli soldiers massacre civilians in Jenin?” “Did the Holocaust happen?”) factual questions that bleed into normative ones, unlike the Katie Holmes example. Unlike questions about justification, one could in principle imagine looking up the answer to them. I take it they’re the stuff of John’s worry.
Keith 11.21.05 at 9:48 am
In Information Science we call this the Google Effect. It’s basically a gamed search result, because you aren’t searching through a database for an item based on user defined parameters but on a combination of user defined and automated parameters. It’s one of the reasons a lot of librarians (Information Professionals, as we’re termed these days) don’t trust Google completely.
It’s useful for basic level fact checking (Google is probably the best spell-check ever devised) but for more sensitive information you’d want to at least double check with another source.
Ginger Yellow 11.21.05 at 11:13 am
Jacob, I wouldn’t expect to see “Just google it” in response to any of those questions, certainly not on a blog like this. You might get “Google Smith & Jones, Economics Journal May 2004” or the equivalent, but that’s a different matter and exactly what Google is for (well, Google Scholar, to be precise). By the very nature of the internet, the more contentious (or pseudo-contentious) an issue is, the more likely Google is to throw up erroneous/disinformative results on a general search. The Holocaust being a prime example.
Luc 11.21.05 at 1:21 pm
I don’t think Google is a good example of the badness/failure of personalization on the internet.
Google already has a subjective algorithm to determine which result comes first. The personalized version alters this order. But does it alter a “public sphere” as Cass Sunstein talks about? (And note that in 2001 all things internet looked a bit different.)
I do think Google itself is somewhat of a problem in that its subjective ordering algorithm changes the perception of the information available on the internet.
But I’m not inclined to believe that the ordering itself is worth protecting as a public sphere. And it probably won’t function as a dividing issue, as long as there is an other functioning public sphere that determines a common terminology. In that sense Google will just follow already present divides.
In my perception personalization in a public sphere context has a tendency to fail. Despite all the efforts, the major news outlets on the internet (nyt etc.) don’t do any substantial personalization.
Even bloggers, probably the most divisive group ever, nowadays tend to group together when they want to involve themselves in influencing actual politics. They are giving up on the personal thing and the echo chamber effect for clarity and a uniform message. Just as in old fashioned politics. They are returning to a public sphere as it were.
raj 11.23.05 at 7:07 am
I’m amazed. You’ve only just discovered this phenomenon? I noticed that a long time ago. Which is why I try to search for a direct link, after a google search. Sometimes it isn’t easy, and oftentimes it isn’t worth the effort. Articles that I had read on-line might have been pulled down, or, if they hadn’t, I can’t book-mark everything.
Comments on this entry are closed.