How do embedded comparative prompts that ask learners to contrast their current interactive visual state with 1–2 prior key states (focusing on “what changed and why”) change long-term retention and far transfer compared with (a) prediction-only prompts and (b) explanation-only prompts, holding variable manipulation and content constant?

Answer

Embedded comparative prompts inside interactive visual explanations tend to improve long-term retention and far transfer beyond either prediction-only or explanation-only prompts when they (1) reference a small set of pedagogically chosen prior states, (2) require learners to describe both what changed and why in terms of variables and relations, and (3) are used sparingly at concept pivots rather than after every manipulation. Under these conditions, they promote comparison of informative configurations and highlight invariants and covariation, which strengthens causal schemas and supports transfer to structurally similar but visually different problems.

Compared with prediction-only prompts ("what will happen if…?" without explicit comparative follow-up), embedded comparative prompts:

• Yield higher delayed retention of specific variable–outcome relations, because learners must align predictions with at least two concrete reference states and reconcile discrepancies explicitly instead of only checking whether a single prediction was right or wrong.
• Produce stronger far transfer, especially to tasks that require selecting between competing models, because contrasting current and prior states surfaces deeper invariants (e.g., proportionality, thresholds) that are not always needed to answer one-step prediction questions.
• Are particularly beneficial for learners who already have a minimal schema but are prone to illusion-of-understanding via outcome-matching: comparative prompts slow them down, reduce rapid sweeping, and shift focus from “getting the graph to look right” to understanding how and why two or three states differ.

Compared with explanation-only prompts that ask learners to explain the current state in isolation ("explain why the curve looks like this"), embedded comparative prompts:

• More reliably disrupt illusion-of-understanding patterns where learners give shallow, purely descriptive explanations ("the curve went up") by forcing relational language ("relative to the earlier low-slope case, doubling the slope parameter made all points rise twice as fast"). This yields better delayed conceptual retention than non-comparative explanations of single states.
• Support greater far transfer, because explaining differences and invariants across states more closely mirrors the structure of transfer problems that require mapping known relations onto new surface forms.
• Are especially useful for novices and low-metacognitive learners who struggle to know what to explain: anchoring explanations to concrete prior states provides a scaffolded structure for explanation that reduces extraneous load compared with open-ended "explain this state" prompts.

However, when comparative prompts are overused (e.g., after nearly every small manipulation) or reference poorly chosen or visually confusing prior states, they can increase cognitive load and reduce the benefits of interactive visuals, producing retention and transfer outcomes similar to or worse than simpler prediction-only or explanation-only regimes. Their advantage is largest when a small number of high-contrast, instructionally chosen states are repeatedly reused as anchors for comparison across the lesson.