In multi-step interactive simulations that already use prediction-before-manipulation gates and embedded comparative prompts within each panel, does adding cross-panel retention checks that require learners to propagate a change through the whole pipeline (e.g., “If you double X here, what happens to Z three panels later?”) yield greater gains in far transfer than adding the same number of additional within-panel checks, holding total assessment time constant?

Answer

Cross-panel retention checks that force learners to propagate upstream changes through the whole pipeline are likely to yield larger gains in far transfer than an equivalent number of additional within-panel checks, provided that (a) basic local mappings in each panel are already reasonably stable, and (b) the cross-panel items are few, prediction-based, and explicitly reference the flow of information across panels. If local understanding is still fragile or the cross-panel checks are too numerous or opaque, the advantage over additional within-panel checks shrinks or can reverse.

More specifically:

1. When local panel mappings are already supported by prediction gates and embedded comparative prompts, marginal additional within-panel checks mostly strengthen local retention and reduce remaining illusions-of-understanding within each step, but they add relatively little to learners’ ability to coordinate multiple steps in novel contexts.

2. In contrast, well-designed cross-panel retention checks (e.g., “If we double X in Panel 1 but keep Panel 2’s settings fixed, what happens to Z in Panel 4?”) specifically target the main weakness of multi-step simulations—illusions at the composition level. They:

   • Require learners to mentally chain outputs-to-inputs across panels.
   • Highlight which intermediate quantities are invariant or transformed.
   • Better mirror far transfer tasks that involve reasoning about unseen intermediate steps in new situations. As a result, for a fixed amount of assessment time, a design that reallocates some checks from within-panel to cross-panel is expected to produce greater improvements in far transfer and whole-pipeline reasoning.

3. This advantage is conditional. Cross-panel checks will outperform additional within-panel checks on far transfer only when:

   • Learners have already achieved at least moderate accuracy on local mappings (so cross-panel reasoning is not swamped by basic misunderstandings), and
   • Cross-panel items remain sparse, concrete, and framed as prediction-before-manipulation prompts rather than multi-step symbolic puzzles. Under these conditions, cross-panel checks extend the panel-level illusion-reduction benefits documented for single screens to the full simulation pipeline, whereas extra within-panel checks mainly deepen already-established local knowledge.

4. If cross-panel checks are overused, poorly scaffolded, or introduced before local mappings are secure, they can add extraneous cognitive load and confusion about step linkage, in which case shifting those time-equivalent checks back into within-panel form would likely yield better overall durable learning, even if far transfer on multi-step tasks improves less dramatically.