AI was found lacking in diagnosing student misconceptions because models do not understand lack of logic or sense making¶

Claim · All claims
Evidence · 1 study · design r2 · q2
Currency · newest study 2024 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? Cohort teams found models currently lack the quality needed to diagnose student misconceptions or generate explanations matched to a student's frame of mind. → misconceptions

Evidence¶

misconceptions¶

Noakes, S., Shell, A., Murillo, A., Van Nostrand, P., Ruiz, P., Cornick, S., and Karim, S. (2024, October). An Ethical and Equitable Vision of AI in Education: Learning Across 28 Exploratory Projects. Digital Promise. https://doi.org/10.51388/20.500.12265/232

q2 · i? · design · r2

Technical findings across the cohort, reported in Table 5 under diagnosing student misconceptions. The report states models "do not understand lack of logic or sense making," and consequently AI was not successful generating explanations relevant to a specific student's frame of mind.

"This is an area where models are currently lacking in quality as it does not understand lack of logic or sense making. In turn, this also means AI is not successful with generating explanations that are relevant to the specific frame of mind for a student."

Discussion¶

Related Claims¶

• AI pilots found models not yet capable of high-quality math output generation, including multiple choice problems, coherent word problems, and math visuals — related
• Using AI to evaluate AI outputs was often challenging because AI act as yes people and humans struggled to agree on evaluation rubrics — related
• Instruction that confronts science misconceptions (conceptual-conflict interventions, refutation text) produces conceptual change more often than standard exposition — related
• AI is currently not best practice for competency-based micro-credential assessment; human assessors remain indispensable — a broader claim this one bears on
• Accurate LLM solution diagnosis does not reliably produce pedagogically actionable feedback, with lowest rubric scores on error identification and actionability — related
• LLMs expressed unjustified confidence in recommendations based on visual inputs, asserting incorrect pin-connection diagnoses with high stated certainty — related
• Students report negative experiences with ChatGPT including inaccurate outputs, lack of emotional connection, and risk of overreliance, which constrain competence — related