How does granularity of variable control in an interactive visual explanation (e.g., coarse preset states vs medium-step sliders vs continuous fine-grained sliders) alter productive struggle and long-term conceptual retention, compared with each other and with matched worked examples that show only the preset states?

Answer

Coarser, structured control generally supports more productive struggle and better long-term retention than very fine, continuous control, and all interactive conditions tend to outperform matched preset-only worked examples when prediction/explanation prompts are present.

1. Coarse preset states

• Learners click among a few informative configurations.
• Struggle is focused on comparing states rather than exploring a space.
• With brief prediction–explanation prompts at each preset, this condition typically yields high retention and reduced illusions, often matching or slightly exceeding worked examples that show the same states passively.
• Risk: if presets are too few or too similar, interaction adds little over a worked example.

2. Medium-step sliders (discrete, limited range)

• Sliders move in meaningful steps (e.g., 3–7 positions) around key relations.
• This often gives the best balance: enough room for hypothesis testing, but not enough for aimless sweeping.
• With one-variable-at-a-time constraints and sparse prediction–feedback–explanation cycles, medium granularity tends to produce stronger durable learning and far transfer than both coarse presets and worked examples, because learners can discover patterns across multiple nearby states.
• Risk: if step size is arbitrary or unlabeled, learners may still sweep without interpreting steps.

3. Continuous fine-grained sliders

• Many possible states, easy rapid movement.
• Strongly prone to rapid sweeping and outcome-chasing (illusion-of-understanding patterns identified in c11, c36, c39) unless heavily constrained.
• Without strict scaffolds (snapping to key states, gated predictions, one-variable-at-a-time), this condition often yields higher immediate performance but weaker delayed retention than medium-step sliders and may fall back to or below a well-designed worked example.

4. Comparison with matched worked examples (preset states only)

• Static worked examples that present the same preset states, with text-based explanations, provide low-load exposure and usually reliable but modest retention and transfer.
• Coarse presets with prompts typically match or slightly beat these on retention and near transfer.
• Medium-step sliders with prompts typically exceed them on far transfer and conceptual flexibility.
• Poorly scaffolded continuous sliders can underperform good worked examples on delayed retention, despite higher engagement.

Design implications:

• Default to medium granularity, discrete steps, one-variable-at-a-time.
• Use coarse presets early or for novices; treat continuous sliders as an advanced/unlocked mode with added scaffolds (snapping, highlighted key states, prediction gates).
• Always pair higher granularity with sparse prediction–explanation prompts to convert extra freedom into productive struggle rather than noise.