they’re just optimizing for the wrong metric. If the metric was “user satisfaction” they could target it day and night and nobody would be complaining.
Any metric can be abused if you try to optimize it at all costs. You can easily increase average user satisfaction by reducing the usage by less satisfied users, it doesn't mean you improved things.
The issue is that something like this difficult to operationalize. User Engagement conflated with user satisfaction, with poor results. Maybe you could force your users to fill out Likert scales all the time, I bet they'd love that.
