For interactive visual explanations deployed repeatedly over a course, can we define a small set of course-level progression rules that automatically adjust variable granularity (coarse presets → medium-step sliders → constrained multi-variable control) based solely on simple interaction-trace features (e.g., success on manipulable retention checks, OVAT rate, repeat errors), and does such rule-based progression match the durable learning and far-transfer benefits of hand-tuned, topic-specific progression by expert designers?

Answer

A small course-level rule set that upgrades variable granularity based on simple trace features will likely recover a large fraction, but not all, of the durable-learning and far-transfer benefit of expert hand-tuned progressions. It should work well for many mid-complexity topics and typical learners, but will underperform expert designs on edge-case topics, very fragile novices, and very advanced learners unless instructors override or refine the rules.

A plausible minimal progression policy:

• Start: coarse contrasting presets only.
• Step 1 unlock (medium sliders, 1 variable): triggered when learner shows repeated success on manipulable retention checks at preset states and a modest OVAT rate with low repeat errors.
• Step 2 unlock (constrained 2-variable control): triggered when learner maintains success on mixed single-variable checks and avoids new systematic errors after Step 1.
• Global backoff: if repeat errors spike or checks fail, temporarily step down a level.

Expectation vs expert topic-specific progression:

• Typical domains with similar structural demands (basic functions, kinematics, simple probability): rule-based progression probably comes close to expert progression on delayed, out-of-context retention and yields slightly weaker but still meaningful far transfer.
• Hard, idiosyncratic, or high-dimensional topics with specific misconceptions: experts can time unlocks around those traps; generic rules cannot, so far-transfer gains likely lag.
• Fragile novices and highly exploratory experts: generic thresholds will misclassify some learners; experts can detect and adjust pacing or keep some learners constrained longer.

So: course-level, trace-based progression is worth pursuing as a strong default that captures much of the benefit of fine-grained topic tuning, but key units and edge learner profiles will still need human or more sophisticated adaptive oversight.