Adaptive learning - Am I the Only Skeptic?

I've been working in edtech for over a decade and I have heard vendors talk about personalize learning paths, individualized learning, and adaptive learning for the entire period.

I have services that support millions of K12 students. So far:

- I AM a believer in student paced. Students should be able to proceed through digital content at their own pace.  I am a believer in letting students repeat lessons and exercises when they want to.

- I am NOT a believer in any of the adaptive learning systems that I've seen.

Why?

Here's an anecdote but it's from a major player. They visited us a month ago and were pitching their adaptive learning platform.  At the heart of the pitch, they had a slide up with a student entering an antonym exercise and various paths coming out. It sounded good. He spoke well. He was the senior product director.

I asked: "So what would be an example of an antonym question?" His example was was a standard multiple choice question such as, 'Which of these is the best antonym for "hot"?'   A. Warm, B. Cold C. Cooking  D. Ice.

I asked, and if they get it wrong, where does that take them, what would be an easier question that would scaffold them into that question.  No answer other than, well, this might not be the best example. I was polite and didn't really insist on hearing a good example.

I just watched a video about MetaCog. It belongs to Victory and it provides a much better link between recommendation engines (Knewton, Dreambox, Area9, Declara) and content. Rather than use the simple data of right/wrong and time on task, it gathers much more data by gathering data of how they do things online.

It assumes that interactive learning activities and assessments can be instrumented to get more info on how the student behaved. Then, Metacog's platform aggregates, analyzes, and recommends along with visualization tools for the teacher to understand.

Here's again the weakpoint, what content actually allows such data to be gathered?

I do see the mechanisms behind many adaptive elearning platforms and they all seem to make simplifying assumptions about content such as:

a. Vocabulary can be sequenced, easy to hard. All of it.
b. Grammar can be sequences,    easy to hard. All of it.

More later....