Spoofing music recommendation services and personalised radio

Music recommendation services and personalised radio stations like Last.FM depend on tracking the behaviour and preferences of their users, and building personal profiles on the basis of this. So what happens if the data you feed into these services isn’t a human’s preferences, but something else, like the programming of a traditional radio station or the output of another recommendation service?
At the risk of putting 2 and 2 together and making 5, there’s another link here with last week’s BBC Creative Future announcement. In the music section of the announcement there is a recommendation to “Enable people to create their own virtual radio channels out of the wealth of our existing output, channels reflecting their own personal tastes”. That doesn’t sound identical to Last.FM — which makes virtual radio channels out of a catalogue of music tracks reflecting users’ personal tastes — but it is kind of similar.

So it’s probably just coincidence, but nevertheless intriguing that, last week, BBC technical team member Matt Biddulph reported asking himself a question: “what happens when you plug behavioural data generated by an automatic process into social software designed for humans?” He pumped a year’s worth of playlist data from BBC 6 Music into a Last.FM profile (the profile seems to be updating in almost real time from 6 Music).
“Will this skew the data pool that underpins Last.FM’s recommendations?” Zac Johnson asked when passing on the Matt Biddulph link. With just one of their 1.5 million users doing it, probably not. But this isn’t the only time something like this has been done. There’s now a way of feeding music recommended by Pandora into your Last.FM account.
It reminds me of a possibly apocryphal story about the early days of Artificial Intelligence and expert systems. The London Stock Market installed some AI machines to help with trading. The machines were programmed to model and predict the way that human traders worked, and to adjust their output accordingly. But there was more than one such machine, and the machines were not programmed to model the behaviour of other machines. When the machines interacted with each other, they nearly went into a feedback loop and sent the market into meltdown!
Perhaps nothing quite so drastic as market meltdown will take place if people start hooking up the feeds from one music recommendation service to another, but it does provide one means for unscrupulous competitors to spoof each others’ systems.
Meanwhile, I can’t muster much sympathy for News International when they complain that the BBC’s plans may compete with MySpace, but Last.FM has less muscle behind it. I imagine the BBC probably wants to work alongside such independent players in the market, and Last.FM doesn’t have a patent on personalised radio software (thankfully such things don’t exist here).

Leave a Reply

Your email address will not be published. Required fields are marked *