A personalized product needs a visible off switch
by•
Personalization is useful right up until the user cannot tell why the product changed. Then it starts feeling less intelligent and more haunted.
My current rule is simple: show the signal, let the user correct it, and make disable or forget controls immediate. A model can infer, but the person should remain the authority on their own goals and preferences.
Where have you found the line between helpful adaptation and unwanted automation?
4 views
Replies
We ran into this with AI matching in our product. When recommendations started shifting and users couldn’t see why, the reaction wasn’t “oh it learned something” - it was “did something break?”
The trust doesn’t just wobble, it flips.
What ended up working for us was showing the signal at the moment of change, not hiding it somewhere in settings.