You know how some services try to offer you products in a similar vein to the one you’re currently consuming?
We all know why they do it. For the money. That’s basically all corporations are about, getting the dosh printed. But how does this practice affect you, the consumer, on a slightly more personal level? Let me sketch up a few concrete examples first.
Most webshops feature sections at the bottom of item pages that show you similar products, or what products other people who also happened to buy this item bought. Comparably, music services (be it streaming or selling) will usually present you with a list of artists similar to the one you’re currently listening to, or that fit the general theme of your recent listening history.
This whole “check this thing out, you’ll probably like it” thing can roll two ways. I’ll illustrate using the “similar music artists” example, since that’s probably the easiest to grasp. I’ll be referring to a “spectrum” a couple of times. When I say that, just visualize a color spectrum, but replace the colors with genres. Of course, similar ones are close together, while the extremes are on opposite sides.
The first is actually slightly grim and can on some levels be compared to the “bubbling” search engines and social media do when they take your opinions into account and only display things you agree with. By only recommending artists that fall into a certain range of the musical spectrum (that is, the user’s current preference), the user is pretty much stuck with that kind of music, assuming they don’t make an effort to listen to new things themselves.
The second way is a bit brighter, as it results in users slowly but surely expanding the range of music they listen to. By recommending artists that are similar, but not exactly the same, the central position of the user’s musical spectrum changes ever so slightly, leading to the range of recommended artists shifting with it. Because the recommended artists are so similar to what the user is currently listening to, it’ll be a much easier transition than jumping from ambient right into hard rock, for example. Over a longer period of time, this leads to the user broadening their musical taste, or at least being able to say they gave a lot of things a fair shot.
On the flip side of that, what if they get to a point where a chunk of the spectrum doesn’t fall into the user’s “I could listen to this” category? What if those chunks are on both sides? It is very likely something like this happens, and when it does, users will no longer listen to the recommended artists because hey, that’s not the kind of music they like. Broadening of taste stagnates and we’re now in a similar situation to the first scenario we talked about.
So, what can be done to fix this? Recommending similar products is fine because, if done right (keeping it comparable, but not too much so) it can encourage the user to broaden the genres or types of products they’ve experienced in a natural way. However, to prevent this from eventually halting, a “here’s something you probably haven’t though about yet” recommendation may be a good idea. I’m not sure how this would work for products you buy in a webshop, but I can definitely see it being useful for finding new music. An algorithm will take a look at your current musical taste and shift that up or down the spectrum randomly, until it finds a genre different from but not opposite to what it knows you like, and then picks an artist in a genre that is popular with a lot of different people. That last bit is important, because you don’t want to introduce someone to jazz by having them listen to the most abstract free-form stuff out there. That’s be too radically different from what they’re used to, as well as generally appealing to a smaller group of people. Instead, pick what’s popular with a broad range of people, from metal-heads to folk-folks.
Music distribution networks, you can mail job offers and signed checks to my email address.