I can confidently assure you that I've had that feature off for ages and could not possibly be happier about it.
And yes, some people, myself included turn it OFF because of privacy concerns. But we are in a really small minority.
A/B testing tends not to produce the result you want when the metrics you are going for are harmful.
Generally speaking there's no answer in principle to the question "is the time spent actually valuable". For the vast majority of people, time spent on Youtube cuts into time spent doing basically the same thing somewhere else. We're not all researching cures for AIDS or new sources of green energy. Most people just want to sit around being entertained.
If your non-personalised experience is stunted in functionality because the core UX of the product is biased towards personalisation, it will be unsurprising if people "prefer" personalisation.
But it's very unlikely the actual product experience would be worse with personalization enabled, measured by probability that a search includes something relevant near the top.
Somehow the Instagram ads algorithm has figured out exactly what I’m self conscious about with my own body; all I see are ads for invisible braces, skin cleansers, and hair treatments.
I don’t actually want to buy any of these things, and they make me feel bad to look at. A de-personalized Instagram experience would be far superior to me.
An unpersonalized IG ad experience would be teenage women's clothing and Netflix branding (assuming you're in the US). You probably don't want that either.
On YouTube, I highly recommend turning off personalization and watch history, especially if you find yourself spending too much time there clicking recommended videos. Without history, it will pretty reliably recommend the same videos again and again, which can help you realize when you have seen a lot and might want to stop.
It may be true that Google has over-invested in personalization-driven features, and under-invested in user-neutral features.
Personalization belongs in the agent. Specifically, the agent that I own and control.
Centralized services have built personalization systems in which they operate the personalization algorithms.
I don't have a problem with that, per se, except in that their interests (profit) and user interests (effectiveness) are only partially aligned.
The biggest reason, as near as I can tell, these system architectures were built is that control of personalization algorithms means (1) a means and reason to collect and retain data on users (that can be used for targeted advertisement) & (2) a means to execute targeted advertisement (via insertion into "recommended" content).
It's likely the only way these services are possible at the "free" price point. But it's disingenuous to suggest that they're the only model by which personalization could be realized.