TrashCat Interventions

Intervention Categories & Selection Logic

The system uses adaptive intervention selection based on error patterns:

1. Cross-operation error detected (e.g., answered 12 for 3×4, which is 3+4+3+3) → Specialist Tools (deterministic)
2. Repeated error in same session (same fact wrong twice) → Production/High-Effort Tools (random selection)
3. First error → Scaffolded/Low-Effort Tools (random selection)
4. If student fails an intervention → Reuse same intervention type (avoid intervention hopping)

Scaffolded/Low-Effort Tools

For simple slips—brief corrections with minimal cognitive load

CueFading.FRQ

How it works: Flash the correct answer (2-3 seconds), hide it, ask student to recall from memory.

Learning science: Implements cued recall transitioning to free recall. The brief exposure primes working memory, then immediate retrieval demand strengthens the memory trace through the testing effect. Minimizes production cost—students don't generate the answer, just recognize and retrieve it.

TurnWheels.AnswerOnly

How it works: Display the correct answer using scrollable digit wheels/spinners.

Learning science: Passive exposure with interactive element. Student manipulates the interface to reveal the answer, creating light motoric encoding without production demands. The scrolling action adds mild engagement beyond pure visual presentation.

Retry.MCQ

How it works: Show the exact same MCQ question again.

Learning science: Immediate retrieval practice while memory trace is hot. The first error eliminates wrong answer from consideration (negative priming), making second attempt easier. Tests whether error was due to momentary lapse vs. genuine knowledge gap.

DragDrop.AnswerOnly

How it works: Show the problem with factors visible (e.g., "7 × 8 = ___"), student drags only the answer digits into the blank. Provides draggables for answer choices only (correct + distractors).

Learning science: Production effect lite—student generates the answer through assembly rather than pure recall, but with the problem structure visible as support. Lower cognitive load than full equation reconstruction since factors remain displayed. Digit manipulation creates motoric memory traces and forces attention to answer structure (place value, digit order) without computational or relational assembly demands.

Production/High-Effort Tools

For deeper encoding when student is stuck—requires generative effort

DragDrop.FITB

How it works: Show equation with ALL parts as blanks (e.g., "_ × _ = ___"), student drags digits to reconstruct the entire equation. Provides draggables for all equation parts (factors and answer) plus distractors.

Learning science: Deep generative learning—student must reconstruct the complete relational structure, not just retrieve the answer. Requires holding the problem in working memory while assembling all three components. The empty equation forces attention to how all three numbers relate through the operation. High production demand—physically manipulating all parts creates strong motoric encoding and forces active processing of the mathematical relationship.

CueFading.ListenMCQFRQ

How it works: Play equation as audio, show MCQ, then ask free response.

Learning science: Multi-modal encoding (auditory + visual) creates multiple retrieval paths. Audio narration supports phonological loop in working memory, helpful for auditory learners. The MCQ→FRQ progression is scaffolded production, moving from recognition to generation.

TimedRepetition.Recall.FRQ

How it works: Show answer, let student copy it, cover it, ask for recall.

Learning science: Classic copy-cover-recall method from reading fluency research. Copying creates dual encoding (visual perception + motor production). Immediate covered recall is retrieval practice with minimal retention interval, strengthening the fresh memory trace.

AnswerFirst.MCQ.FRQ

How it works: Display full equation with answer, then quiz with MCQ followed by free response.

Learning science: Pre-exposure + retrieval practice. Seeing the complete fact first establishes the memory trace, then the MCQ→FRQ sequence provides graduated retrieval difficulty. The full equation display emphasizes the relationship between all three numbers.

Specialist Tools

For diagnosed misconceptions requiring targeted remediation

TurnWheels.All

How it works: Show entire equation using interactive digit wheels for all values.

Learning science: Reserved for cross-operation errors (e.g., adding when should multiply). Interactive manipulation of all equation components forces attention to the operator and structural relationships. The wheels allow exploration of how changing each element affects others, supporting relational understanding rather than rote memorization. Addresses systematic misconceptions, not random errors.

Key Design Principles

1. Interventions are untimed - removes retrieval pressure, allows encoding-focused practice
2. No XP during interventions - architectural penalty for errors (time cost) without punitive feel
3. Parallel with demotion - interventions provide support, but fact still demotes in learning stages (maintains data integrity)
4. Adaptive intensity - light corrections for slips, heavy production for struggles, diagnostic tools for misconceptions
5. No intervention hopping - if student fails an intervention, they get the same type again (consistency aids learning)