Development and Validation of an AI-Based Model for Identifying Senior Secondary School Students’ Errors in Chemistry Calculations in Lokoja Metropolis, Kogi State, Nigeria

This study evaluated the development and validation of an AI-based model for identifying senior secondary school students’ errors in chemistry calculations in Lokoja Metropolis, Kogi State, Nigeria. The study adopted a Research and Development (R&D) design. The study population comprised all 24,152 SS1 and SS2 students, from which a sample of 269 students was selected. The instruments used for data collection were the researcher-developed Chemistry AI Test (CALT) and Identification of Students’ Errors in Chemistry Calculations (ISECC) questionnaire. The instruments were validated by two chemistry teachers and one Measurement and Evaluation expert. Reliability was established through pilot testing, and Kuder–Richardson Formula 20 and Cronbach’s alpha yielded coefficients of 0.79 and 0.81, respectively. Data were analysed using descriptive statistics, while students’ working steps were analysed using the ChemError-AI model to classify errors into mathematical/arithmetic, conceptual, formula, unit, interpretation, and substitution categories. Findings showed that the ChemError-AI model correctly classified 58 out of 64 unseen test cases, representing 90.6% accuracy. The findings indicate that the model can provide preliminary information about the types of errors students make in chemistry calculations. It was concluded that students experience difficulties at different stages of chemistry problem solving and that these difficulties extend beyond mathematical operations. It was recommended, among other measures, that teachers should receive practical training on how to interpret AI-generated classifications and combine AI-supported information with their professional judgement before making instructional decisions.