In interactive visual explanations that already use prediction-before-manipulation and a visible manipulation budget, does adding a short, system-generated “trace summary” (e.g., a simple schematic or bullet list of the learner’s own key variable changes, predictions, and outcomes) before delayed, out-of-context retention checks further reduce illusion-of-understanding and improve far transfer compared with no summary, and for which learner profiles does this reflective layer meaningfully add to productive struggle versus just reloading surface memories?

Answer

Adding a brief, system-generated trace summary before delayed, out-of-context checks will probably yield small extra gains in reducing illusion-of-understanding and improving far transfer, mainly for learners with some schema and regulation; for fragile novices and very high-knowledge learners it often adds little or can backfire if it becomes a cue sheet instead of a prompt for abstraction.

Best bets:

• Supported novices / intermediates: net positive if the summary is compact, structural (variables, directions, a few key states), and followed by prediction/explanation items that do not mirror the original steps.
• Fragile novices: risk of overload or rote review; better with minimal or highly guided summaries, or none.
• High prior-knowledge / strong self-regulation: small added value; effects similar to a light self-explanation prompt.

Mechanism sketch: the trace summary externalizes the learner’s own interaction trace, helping them see patterns (e.g., OVAT vs sweeping, key contrasts) and re-encode structure without re-entering the visual. This can convert some surface success into a more organized mental model, but only if summaries stay schematic and are tied to generative retrieval, not simple recognition.

Net: treat trace summaries as a low-cost, modest booster layered onto prediction + manipulation budgets, with tight constraints on length and format and selective use by learner profile.