Student Achievement in Temperature and Heat Following Artificial Intelligence–Integrated Problem-Based Learning

Abstract

Understanding temperature and heat remains challenging for junior secondary students, motivating instructional approaches that combine problem solving with digital support. This study examined changes in science achievement following artificial intelligence (AI)-integrated problem-based learning (PBL) on temperature and heat concepts. A quantitative pre-experimental one-group pretest–posttest design involved 30 seventh-grade students at an Islamic junior secondary school (MTs) in Telaga Biru District, Gorontalo Regency, Indonesia. Instruction followed five PBL phases and used ChatGPT and Meta AI to support learning-resource searches and comparison of students’ answers. Achievement was assessed using a 25-item multiple-choice test. Paired scores were examined using the Wilcoxon signed-rank test, individual normalized gains, and standardized effect-size estimates. Mean scores increased from 43.20 (pretest) to 86.67 (posttest), with a statistically significant paired difference (|Z| = 4.789, p < 0.001). The mean individual normalized gain was 0.76 (high), with 19 of 30 students achieving high gains. The Wilcoxon effect size was r = 0.87; complementary Cohen’s d_z was 3.29 (95% CI [2.37, 4.21]). These findings indicate substantial within-group achievement improvement, but the absence of a comparison group precludes causal attribution to the instructional intervention or its AI component