Home · Ideas that failed
Ideas that failed
Expansion also means recording what does not work. This shelf holds named exclusions — ideas we will not promote as claims — and any claim that attacks killed outright.
Plain language. Counts of attacks, not scores. We keep these visible so the library does not quietly forget rejected ideas.
Named exclusions
From the project’s exclusion list. These are not living claims; they are rejects that still matter for policy talk.
- Learning styles / VARK matchingMatching teaching to a preferred “style” (visual, auditory, read/write, kinesthetic) is not a sound teaching principle. Preference is not evidence that matching raises learning.
- Wrong-domain proxies for GenAI collaborationStudies that measure chatbot satisfaction or short preference surveys are not evidence that AI collaboration improves later unaided learning.
- Vendor white papers without closed-book outcomesMarketing reports that never test learning without the tool do not clear the bar for a claim here.
- Untimed preference surveys treated as comprehensionLiking a medium or feeling that you understood is not the same as understanding on a test.
- HE-only chatbot satisfaction used as K–12 policyUniversity students saying they like a chatbot is not a reason to set policy for schools.
- Hattie-style aggregated d-lists as causal mechanismGiant ranked lists of “what works” average across unlike studies; they are not a mechanism and not a kill-test.
- “Digital natives” as evidenceThe idea that today’s children learn differently because they grew up with screens is not established evidence for teaching choices.
- Brain Gym / perceptual-motor as reading interventionMovement programmes sold as reading treatments do not clear a fair academic kill-test.
- Working-memory training sold for far transferTraining working-memory tasks does not reliably raise school subjects or general thinking far from the trained tasks.
- Viral phone-ban anecdotes without DiD/RCTA vivid story about a ban is not the same as a difference-in-differences or randomised evaluation of learning outcomes.
- Thin growth-mindset PD packs without dose/fidelityShort mindset tip-sheets without a real lesson dose or fidelity check are not the interventions that trials tested.
- Generic “21st century skills” frames without measurable closed-book knowledgeSlogans about future skills that never measure later unaided knowledge do not become claims here.