This is a big issue with a lot of ML algorithms - it causes issues with consistency. Prioritising apps in auto complete search should be done with a simple tall

This is a big issue with a lot of ML algorithms - it causes issues with consistency. Prioritising apps in auto complete search should be done with a simple tally of how many times the user has opened the program, perhaps as a rolling n days or something. Either way, that causes consistency. If you have some ML algorithm that takes into account how long the window of the app has been visible or in focus or a bunch of other weird features will cause poor UX the moment the baby projectile vomits across the room while the dog has diarrhea that the robot vaccuum spread all over the room. Suddenly you have some esoteric app called PhotosResizer2000 stay in focus for 8 hours while you deal with the emergency from which point onward the ML algorithm will suggest this app whenever you start to type Pho...