If an interactive visual explanation enforces prediction-before-manipulation and delayed, out-of-context retention checks, does automatically surfacing prior interaction traces (e.g., ghosting a learner’s earlier parameter choices and predicted outcomes as they revisit the visual) strengthen durable conceptual understanding more than using the same screen space and time for additional new manipulation opportunities, and for which learners does replaying their own traces mainly reinforce unproductive patterns or illusion-of-understanding?

Answer

Surfacing prior interaction traces can slightly improve durable conceptual understanding over using the same space/time for extra manipulations, but mainly when traces are (a) sparse, aligned with key contrasts, and (b) tied to new prediction or explanation prompts. Replayed traces help learners who already engage in roughly hypothesis-driven manipulation; they mainly reinforce illusion-of-understanding for outcome-chasing, low-regulation novices unless tightly coupled to correction prompts.

A. When ghosted traces beat additional new manipulations

• Traces are: • Summarized (a few prior settings/outcomes, not full playbacks). • Anchored to diagnostic contrasts (where past predictions were wrong or shifted). • Framed with prediction or self-explanation prompts ("Last time you predicted Y at this setting; what do you predict now, and why?").
• Learners revisit the visual after delay, so traces act as retrieval cues plus comparison targets across time, supporting durable learning more than more in-session tinkering.
• Extra manipulation slots would otherwise encourage unsystematic sweeping rather than new, well-structured contrasts.

B. When more new manipulations are better

• Local understanding is still weak and interaction history is noisy (random sweeping, many non-diagnostic states).
• Learners have not yet seen enough clean contrasting cases; replaying messy traces just re-exposes confusion.
• In these cases, using space/time for a few additional, well-constrained new manipulations with strong prediction gates yields more learning than resurfacing old, unstructured choices.

C. Learners who benefit most from trace replay

• Intermediate learners with some schema and moderate self-regulation, whose earlier traces include semi-systematic tests.
• High-ability, exploratory learners whose prior edge-case probing becomes explicit material for comparison and abstraction.

D. Learners for whom trace replay risks reinforcing illusions

• Novices with low prior knowledge and low self-regulation, whose traces show sweeping, outcome-matching, or copy-what-worked behavior.
• Overconfident learners who treat ghosted “successful” runs as proof of mastery unless prompts force them to revisit and explain.

Design implication: Use trace replay as diagnostic, prompted snapshots rather than full replays; gate it on evidence of minimally productive traces, and keep prediction-before-manipulation and delayed, out-of-context checks as the primary drivers of durable learning.