Are ZPD / scaffolding (§2.2) implemented, or framing only?

[sh-taheri]

Hello,
Thanks a lot for the awesome repo!

The paper grounds the tutoring design in Vygotsky's Zone of Proximal Development and Wood et al.'s scaffolding theory (section 2.2, stage 3), where C_mem is said to steer explanation depth and tone: scaffolded derivations for beginners, concise summaries for proficient learners.

However, I couldn't find a corresponding mechanism in the code. Neither theory is named anywhere in the repo, and the only depth-adaptation path I found is deeptutor/multi_user/learner_profile.py, which serializes six learner-declared fields (age, grade_level, reading_level, explanation_style, ...) into the prompt with the instruction "Use it only to adapt explanation difficulty and presentation", self-reported, not inferred from performance, and with no corresponding logic in the chat agent prompts.

Meanwhile the real adaptive-difficulty engine in deeptutor/learning/ is explicitly built on a different tradition: policy.py gates advancement at 0.9 mastery with the comment "mirrors Alpha School's '90% before you advance'," which is Bloom-style
mastery learning rather than ZPD.

The two also appear disconnected: mastery scores don't seem to feed the writer that chooses explanation depth. Am I missing something, or is the ZPD framing describing intent rather than shipped behavior?

Either way it'd be useful to know, since it affects how to read section 2.2.

Thank you :)