Thursday, May 20, 2010

Google and Received Opinion

Brilliant satire from @newsbiscuit : New Google only searches for sites that match your preconceived opinions.

now so much easier to find exactly what you want to see

I have long complained that Google provides a systematically distorted way of finding out what is going on, and encourages what A.A. Milne called Thinking with the Majority. This is because Google's page ranking algorithms are basically designed for people who want to ask the same questions as everyone else, and get the same answers. Consequently, Google helps to amplify the circulation of Received Opinion.

The distortion is further amplified by massive duplication of material from a common source. If you search for a topical story, you will often find hundreds of popular websites repeating exactly the same version of events in slightly different words, and unless you are extraordinarily persistent you may never find a website that gets its information from a different source. See my posts on Google and Spin (1, 2).


@roygrubb raises a related concern - that Google remembers and is influenced by our previous searches. This is not just a privacy issue but also a context issue - our interests may switch from one project to the next. For example, let's say I'm working on a project using SAP, so when I'm searching for technical information on this project I may concentrate on material that is relevant to SAP. But I certainly don't want Google to put an implicit SAP filter on my searches, even if some Google engineer thought this would be helpful to me, because that could seriously prejudice my view of the available technology. Worse, this bias might persist (without my knowledge) when I'm working on a completely different assignment.

I can imagine that Google could build some kind of context-awareness into its search algorithms, so it somehow detects when I move to another project. And (to take a more controversial example) if I search for information about some deadly disease, it can try and work out whether I'm suffering from the disease myself (in which case it can sell me health insurance before it's too late) or enquiring on behalf of a friend or client, or whatever. But that's not the point. The point is the increasingly complicated relationship between our tools and our knowledge, which even many technologically literate people seem touchingly naive about.

No comments: