Apple Music uses a number of signals to learn your musical tastes. The most obvious one is that you can explicitly Love or dislike a particular track by flagging it in iTunes or the Music app. You can also do this on HomePod, by saying ‘Hey Siri, like this track’ or ‘Hey Siri, dislike this track’ while it is playing.
But it also uses a number of algorithms to automate its learning. For example, if you play a particular track often, it will assume you like it. And simply playing a track and listening to it all the way through (as opposed to skipping it) has some degree of influence.
All of which gets problematic when more than one person controls it …
My partner and I have rather different musical tastes. We do have shared genres and artists we both like, but I have music Steph dislikes, and vice-versa. We also let guests control it when they visit, and their musical tastes may again vary from mine.
HomePod has a solution for this: you can tell it to either use or not use your listening history. But that’s not a great solution.
If I leave listening history on, then my Apple Music account gets influenced by the tastes of my partner and our guests. If I switch it off, then my account no longer learns from my listening habits on HomePod, and since that’s most of my listening these days, that’s a big problem too.
In theory, there’s a third option, which is to switch it on and off repeatedly, depending on which of us is controlling it at the time – but that can swap back and forth multiple times a day, so isn’t really practical.
The ideal solution, of course, would be for HomePod to offer multi-user support. Associate two or more different Apple IDs with it, and identify ourselves by voice. Then have HomePod recognize the voice and update the appropriate listening history. In that way, it would also do things like read the correct text messages and check the correct calendar when anyone asks.
But if Apple can’t do that much at this stage, then a good first step would be simply to recognize the voice of the main user. If that user plays something, update their listening history; if anyone else does, don’t.
Distinguishing all voices may be difficult in some homes. You could, for example, have two sisters with very similar-sounding voices. But for the simplest possible case of a male/female couple, it ought to be trivial. And not too difficult to distinguish same-sex couples, and adults from children.
Is this a feature you’d like to see? Please take our poll, and share your thoughts in the comments.
