By 2016, adaptive learning had fallen into Gartner’s ‘trough of disillusionment’. The Gates-funded ALMAP evaluation found no firm evidence of better grades or completion. Computer-adaptive testing is well validated; adaptive courseware is not.
Evidence reviewed
- SRI/Gates ALMAP evaluation: no firm gains in grades or completion.
- Enyedy (2014) Personalized Instruction: New Interest, Old Rhetoric, Limited Results (NEPC).
- Murray & Pérez (2015), adaptive vs traditional learning.
- OECD (2015) Students, Computers and Learning.
- Industry reports (EdSurge “Decoding Adaptive”, sponsored by Pearson) = marketing.
- “There is no magical silver bullet here.” (Rahim Rajan, Gates Foundation)
Implications for builders
- Separate adaptive testing (psychometrics, item response theory) from adaptive courseware in your claims.
- Ask for outcome data against controls, not engagement metrics.
- State exactly what adapts (item selection, feedback, pacing) and the evidence for each.
- Build local evaluation into pilots from day one.
Apply in my context (Claude)
- RS Insight is on the strong side of this divide if it is a diagnostic test with validated items. Keep its claims there. “Adaptive learning paths” is where the evidence is thin.
Connections
Solutionism - define the problem before the technology · Does ChatGPT enhance learning - performance is not learning · DELTA Module 1 Paper 2 - task-by-task technique (validity, reliability)