To what extent do trace-based indicators of productive struggle identified in single-concept interactive visuals (such as increasing OVAT tests, longer dwell times, and revisits to informative contrasting cases) reliably generalize as predictors of durable learning and transfer when learners work in mixed environments that alternate between interactive visuals, static worked examples, and human tutoring on the same concepts?

Answer

Trace-based indicators of productive struggle from single-concept interactive visuals partly generalize to mixed environments, but their predictive power weakens and becomes more conditional on how the non-interactive components are sequenced and aligned. They remain useful predictors of durable learning and transfer within and immediately around interactive episodes, yet they must be combined with coarse indicators from worked-example study and tutoring sessions to reliably forecast long-term outcomes across the whole mixed pathway.

In practice:

• Direction of prediction generalizes; effect size shrinks. Patterns like increasing one-variable-at-a-time (OVAT) tests, longer dwell on informative states, and revisits to contrasting cases still predict better delayed retention and transfer than outcome-chasing traces, but they typically explain a smaller share of variance once learning time in worked examples and quality of tutoring are accounted for.
• Predictive power is highest when components are instructionally aligned. When worked examples and tutoring explicitly reference the same variables, contrasting cases, and representations as the interactive visuals, trace-based productive-struggle indicators retain moderate, incremental predictive value above quiz scores and tutor ratings. When alignment is weak (different notations, different cases, or tutors focusing on procedures), these indicators predict performance mainly inside or near the interactive tasks, not far-transfer tasks framed in other formats.
• Non-interactive episodes can amplify or damp trace signals. Learners with strong productive-struggle traces who then receive well-timed, clarifying worked examples or tutoring tend to show very robust durable learning and transfer. Conversely, heavy procedural tutoring that encourages answer imitation can wash out the benefit of earlier productive exploration, making trace indicators less predictive of final performance.

So, trace-based productive-struggle indicators from interactive visuals should be treated as local but meaningful signals in mixed environments: they are still worth tracking and using for adaptive support, but they need to be interpreted alongside simple measures of how learners engage with worked examples and tutors, rather than being used as stand-alone predictors of durable learning and transfer.