AI tutoring models systematically fail to detect student misconceptions when flawed reasoning coincidentally produces the correct answer, with 71% of failures concentrated in two predictable question types.
John Catt Educational, Wood- bridge, Suffolk (2018)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CY 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Catching The Correct Answer Trap: Characterising AI Tutor Blind Spots When Analysing Student Reasoning
AI tutoring models systematically fail to detect student misconceptions when flawed reasoning coincidentally produces the correct answer, with 71% of failures concentrated in two predictable question types.