Algocreep — a phenomenon of shared use of digital devices and their algorithmic recommendation systems where multiple users’ inputs create undesirable algorithmic results. One person’s algorithmic recommendations creep into the other’s feed.
To anyone who has ever used Spotify while having kids, this phenomenon will be intimately familiar as suddenly generated playlists will contain children’s movie soundtracks or other songs among their own preferences. It can be funny or jarring, but handling multiple users’ input is not something our current generation of algorithmic recommender models are well equipped to do. It’s like algorithmic split personality: One minute, some Beyoncé or 90s hip hop; then suddenly the Paw Patrol theme song.
A more narrow, somewhat milder version is the confusion music streaming services develop between work and entertainment playlists: Many knowledge workers listen to completely different music while working — for example, only instrumental, or focus playlists, or classical piano — while for their enjoyment in their free time they prefer completely different vibes.
Less entertaining and a little more invasive — at least it feels more invasive — is ever using someone else’s device and just being exposed to their completely different targeted ads or sometimes personalized search results. This feels truly weird at times, and it’s not pleasant for anyone involved. It’s like stepping into someone else’s bedroom, only 10x more so. There was never a more awkward time to fix your parents’ computer.
Or maybe it’s just a reminder that algorithms aren’t all that powerful or sophisticated yet. They’re just confused DJs at our party, desperately trying to please everyone — and pleasing no one in particular. In a way, they even more need the world to be orderly and structured than we do. And there’s some solace in that.
