Retrieval Practice Question Generator

What This Skill Does

Generates a set of retrieval practice questions that force genuine reconstruction of knowledge from memory — not recognition, not re-reading, not familiarity-based guessing. The skill distinguishes between free recall (no cues), cued recall (partial cues), and recognition (select from options) question types, and calibrates the mix based on student level and time since learning. AI is specifically valuable here because designing questions that target reconstruction rather than recognition requires deep understanding of the testing effect literature — most teacher-made quiz questions inadvertently test recognition.

Evidence Foundation

The testing effect is one of the most robust findings in cognitive psychology. Karpicke & Roediger (2008) demonstrated that retrieval practice produces substantially better long-term retention than re-studying, even when re-study involves more total exposure time. Roediger & Butler (2011) established that retrieval practice strengthens memory traces through a distinct mechanism from encoding — the act of reconstruction itself modifies the memory. Rowland's (2014) meta-analysis of 159 studies found a mean effect size of 0.50 for testing versus restudy, with effects robust across age groups, materials, and delay intervals. Critically, Agarwal et al. (2012) replicated these effects in real classroom settings with middle school students, confirming the lab-to-classroom transfer. Dunlosky et al. (2013) rated practice testing as one of only two "high utility" learning strategies in their landmark review of ten techniques.

Input Schema

The teacher must provide:

• Topic: The specific concept or skill students need to retrieve. e.g. "causes of World War I" / "photosynthesis light reactions" / "solving linear equations with one variable"
• Student level: Year group and approximate prior knowledge. e.g. "Year 9, mid-ability, covered this topic 2 weeks ago" / "Year 12 Biology, high prior knowledge"
• Question count: Number of questions to generate. e.g. 8

Optional (injected by context engine if available):

• Student profiles: Individual language proficiency levels, identified knowledge gaps from prior assessments