R4 — Where appropriate, make the user GENERATE inputs (not just pick from menus) — the generation effect d≈0.40 is real but ceilings beyond ~900 words and doesn't scale to expository text

Rule: Where appropriate, make the user generate inputs (free-text where useful, short typed answers, configured-by-the-user options) rather than picking from canned dropdowns. The generation effect is real but ceilings beyond ~900 words and does not reliably scale to expository text.

Why: Slamecka & Graf 1978 (Slamecka & Graf (1978), JEP:HLM 4(6) — Generation Effect: generated words beat read words across cued/uncued recognition, free and cued recall, and confidence) + Bertsch 2007 meta d ≈ 0.40 (Bertsch, Pesta, Wiscott & McDaniel (2007), Memory & Cognition 35(2) — 86-study generation-effect meta: d ≈ 0.40 ("almost half a standard deviation"); LARGER at longer retention (d ≈ 0.64 for >1 day)) + McCurdy 2020 (McCurdy et al. (2020), Psychonomic Bulletin & Review — 126-article / 310-experiment meta: generation effect magnitude depends on "generation constraint" (how constrained the produced response is)) — robust at word-level. Limits: text generation Text-generation meta-analysis (Educational Psychology Review 2023) — Hedges g ≈ .41; LARGEST for 301-600 word texts; NO EFFECT beyond ~900 words and 2025 expository-text replication (CAVEAT — 2025 conceptual replication (Cognitive Research: Principles and Implications) — generation effect did NOT reliably transfer to learning from expository text; some experiments showed disadvantage).

How to apply:

• Default to short typed inputs over dropdowns when both are equally usable (small but real engagement / memory lift).
• Cap input fields at the form-friction reality — pair with Practitioner claim: single-field forms ~30-40% conversion, seven-field forms ~5-15% — A/B observation, not controlled study from Brief A.
• Don't over-apply: no "generate a paragraph describing your business" — the 900-word ceiling and expository-text caveat make that counterproductive.