Invitations and Becomings

I was lucky enough to attend a demonstration of some of Vignette's latest offerings, beyond their more well-known content management system, VCM. One of these is Vignette Recommendations.

Recommendations is best described as a behaviourist's wet-dream. It aims to anticipate the behaviour of a visitor to the site, and provide the content that the visitor will ultimately find most valuable. The thing is that this content may not actually have anything in common with what the visitor thought they were looking for - e.g. it may not share any keywords with the search that brought the visitor to that site.

What Recommendations offers are the pages or products that other visitors who arrived at the site in a similar fashion, or who took a similar trajectory through the site, ended up spending most time on.

In general, personalisation is about skewing the proximities between each bit of content to ensure that your visitor is always as close as possible to what they are looking for. It's about customising your IA on the fly. What differentiates forms of personalisation is what they use to determine these proximities, what defines the topology of the site for this person.

Traditional search functions are a form of personalisation, because the results page effectively re-orders your content according to its similarity to the search phrase entered by the user. What differentiates Recommendations is that it doesn't allow the user's articulation of what they are looking for to define the topology of the personalisation. Instead, it treats that articulation as a sign, or better a sympton, of where they are likely to end up, and then reconfigures the IA to minimise the time it takes them to get there.

It's not what you say you're after that matters, it's the behaviour of people who made similar reports to yours that matters. Even these reports themselves are just treated as another bit of behaviour to factor it. Recommendations is interesting because it treats searches as just another move in the game, with no special significance. What people say they want is no more important than any other information you can gain about them. So, arriving at the site from a Google search and arriving at it from some other site are entirely equivalent. Your keywords only count insofar as they help to group you with other people.

This all reminds me of philosophy lecturer of mine who used to joke about two behaviourist bumping into each other on the street: "Hi there," one says to the other, "I see you're fine. How am I?"

But perhaps there's another image that better captures the experience of using recommendations (and yes, I'm aware that talking about "experience" here would be heretical for behaviourists). It comes from Dirk Gently's Holistic Detective Agency by the late, great Douglas Adams. To paraphrase: Want to get somewhere, but don't know the way? Follow a car that looks like it knows where it's going. You may not get where you were trying to go, but you're certain to end up somewhere you needed to be. Recommendations purports to be that car. So, I suppose it's the perfect kind of personalisation for fatalists: it doesn't so much satisfy your wants, as it offers a shortcut to your destiny.