Progressive Hint Ladder

Provide graduated assistance from abstract conceptual nudge to concrete procedural step, with reflection required before each escalation. Teaches help-seeking as a skill and prevents direct-answer shortcuts.

What it does

Provides graduated assistance across six levels — from a diagnostic question through abstract conceptual nudge, analogy, principle reminder, procedural nudge, and near-complete scaffold — with a reflection requirement at each level before escalation is permitted. The learner never receives the full answer. Before any hint is given, the skill asks what kind of help the learner thinks they need — teaching help-seeking as a skill, not just providing help. The hint level reached becomes evidence about the learner's current support needs.

The evidence behind it

VanLehn (2011) synthesised research across human tutors, intelligent tutoring systems (ITS), and other instruction formats, finding effect sizes of d = 0.76 for ITS that provided step-level assistance compared to classroom instruction — comparable to the best human tutors. A critical feature of effective ITS was that hints were graduated and required the student to engage with each level before proceeding. Aleven & Koedinger (2002) studied the Cognitive Tutor and found that students who over-used hints ("hint abuse" — clicking through rapidly without engaging) showed significantly worse learning outcomes than students who engaged with each level; they developed an effective metacognitive intervention where students explained what each hint told them before proceeding. Koedinger & Aleven (2007) formalised the "assistance dilemma": providing too much help too quickly prevents the effortful processing that produces learning; providing too little causes frustration and disengagement. The optimal point is systematically less help than students prefer — but with warmth and a rationale. Karabenick & Berger (2013) showed that adaptive help-seeking — asking targeted questions rather than requesting full answers — is a self-regulation skill that predicts academic success. Wood et al. (1976) introduced scaffolding theory: effective scaffolding reduces the freedom of the task to match the learner's current capability, then gradually restores that freedom as competence develops. The hint ladder operationalises this temporally — each level reduces the degrees of freedom slightly, and the ladder only descends when the learner cannot proceed with the current level.

Sources

• VanLehn (2011) — The relative effectiveness of human tutoring, ITS, and other tutoring systems (d = 0.76 for ITS with step-level help)
• Aleven & Koedinger (2002) — An effective metacognitive strategy: learning by doing and explaining with a computer-based tutor
• Koedinger & Aleven (2007) — Exploring the assistance dilemma in experiments with cognitive tutors
• Karabenick & Berger (2013) — Help seeking as a self-regulated learning strategy
• Wood et al. (1976) — The role of tutoring in problem solving: scaffolding theory

How to use it in your lesson

For the best results with EvidenceLesson, give it:

• problem_or_task — The problem the student is stuck on — ideally with their attempted work so far
• context — Course, topic, and level
• prior_hint_level (optional) — Hint level reached in prior attempts at this or a similar problem (0–5)
• developmental_band (optional) — Learner age or stage for calibrating language

Known limitations

1. The hint ladder assumes a problem with a clear correct path. For open-ended, creative, or discussion-based tasks, the graduated-hint structure is less applicable — there is no "Level 3 principle" to name for an essay thesis. The skill works best for procedural, mathematical, or well-defined analytical problems.

1. Constructing a good Level 2 analogy requires the AI to know what concepts the learner has already encountered. Without session history or explicit context, the analogy may reference something the learner hasn't studied. The skill should prompt the AI to check: "Have you seen problems like this before?" before committing to a specific analogy.

1. Hint-abuse detection is approximate. Rapid escalation requests are the main signal, but some learners genuinely think fast and need little engagement time with each hint. The reflection requirement ("what did that tell you?") is the safeguard, but it adds friction that well-calibrated learners may find annoying.

1. The "no full answer" rule can frustrate learners who are time-constrained. A learner with an assignment due in 30 minutes may experience the hint ladder as an obstacle. The skill is calibrated for self-determined learners building genuine capability; it is less appropriate as a homework helper when the primary goal is task completion.

Pairs well with