Last week I made mention of poor recommendations by way of suggesting that we should maybe stop trying to draw conclusions about everything all the time. This time I want to point out another facet of the problem.
I’ve written before about Big Data and the general meaninglessness of that term. Still one of my favourite quotes from 2013 was (paraphrasing here because I can’t seem to find it anywhere), “Saying ‘I’ve got loads of data, that must be worth a lot of money’ is like a sewage treatment plant saying ‘I’ve got loads of shit, that must be worth a lot of money'” (If that was you, please ping me so I can give you credit!)
Not only is data like feces in that it’s a byproduct of our online consumption (and other behaviour), it’s also not necessarily worth all that much. What is (at least putatively) worth something is the meaning that can be made from the data. And as of now, it seems there’s not a whole lot of that around. Take facebook as an example. Facebook has loads and loads of data about us, its users. It is valued at over $200 billion, in large part because of that data. It employs some very clever people who are presumably working on very clever algorithms to extract meaning from the data. It performs experiments that collect more data to assess the impact of subtle changes in the way it presents content. And yet, meaning remains elusive. Try the following experiment: open a web browser you normally don’t use. Make sure the cache is cleared and you’ve deleted all your cookies. Now log onto facebook and have a look – notice anything? It’s not personalised like it was in that other browser that you use a lot. Hmm. That would seem to indicate that most of the data that’s used to customise your facebook experience comes from your browser and not from all those volumes of data that facebook has about you, sitting on their servers. So what, exactly, is the value of that data? What kind of meaning can facebook extract from it? Who benefits from that meaning, if anyone does?
Here’s another one: Amazon knows what you’ve bought. It will, as previously discussed, try to sell you a thing based on whatever the last thing was that you bought, no matter whether that’s appropriate or not, and correcting it is a significant effort. And yet it’s not very good at suggesting things that are actually like the things that you like – especially if the things you like are a bit off the beaten path. One friend of mine has been indulging a minor obsession with the Easter Rising of 1916. He’s bought maybe 8 books on it, all on Amazon, yet Amazon has never recommended another book on the topic. Curious, no? Or Spotify. I listen to a good deal of classical music, lately a lot of Chopin and Debussy. Using this as a rationale, Spotify recommends Stravinsky. That’s like saying because you like Britney Spears, you’ll like Nina Hagen. Yes, Stravinsky and Chopin are in both composers of ‘classical’ music, but their music is wildly different. Just as Britney and Nina are both female singers, but that’s where the similarity ends.
Shouldn’t we be doing better than this by now? There is so much data out there, especially so much metadata about music and books, that it seems absurd that we’re getting something so basic so wrong. And it’s not improving, which is a little alarming. Are my (and my friend’s) interests unimportant because they’re outside the mainstream? Do I deserve what I get in the world of music recommendations because I’m not listening to the same things as the majority?
One of the great things about the internet is that it (in theory at least) levels the playing field – anyone with access to it has just as much right to consume or create a piece of it as anyone else. But the products we use the most don’t seem to honour that principle – they still cater to the majority, even though including outliers is (at least in the cases I’ve listed here) easier than ever. That’s lazy, and it’s also bad business – good recommendations lead directly to increased sales.
I know it’s not as sexy as a self-driving car, but I’d really like to have a book or music buying experience that’s at least half as good as it used to be when there was a bookshop on every second corner, with staff who actually read. Or maybe I’ll just wander down to the Wellcome Collection instead. Their bookshop is killer.
To be continued (part 2: curation)….