> I suppose that these are most likely school-provided and school-administered devices, so they should arguably allow for some sort of time-based kiosk mode where the student is restricted in what they can do on the device.

In our case, schools purchase and bring their own devices past elementary school. But even with technical measures, there are a nearly infinite number of ways to screw around.

> they seem to have the potential get closest to the "Bloom's two sigma" result of fully individualized instruction.

They don't, though. In my experience, there's 3 reasons why a student will devote effort to improvement in a classroom. In order of their efficacy and difficulty to instill:

1. Pressure from grades/the gradebook. In my experience, this is only weakly effective. Even in my environment where families are really achievement focused. There is too much of a delay; even if things are updated in Khan or the gradebook nigh-immediately, the measure doesn't become consequential for a long time.

2. Social pressures in the classroom: desire to not look foolish; relationship with an adviser; desire to please the teacher; effects of appropriate praise; desire to do fun things that other students are doing.

3. True interest and independent engagement in the subject.

You could alternatively view this also as a scale of how quick and effective feedback is. By #3, the student starts to measure themselves.

Khan or DI will have a hard time taking a student to #3. Khan's a super-strong, super immediate version of #1; DI is a very weak version of #2 (but possibly the easiest to implement).