Improving Tenth-Grade Students’ Anecdote Writing Skills through a Generative AI-Assisted Think Pair Share Strategy

Abstract

Writing anecdote texts requires students to combine text structure, humour, and social criticism, yet many tenth-grade students produce incomplete texts with weak satirical force. This classroom action research aimed to describe the implementation of a Think Pair Share (TPS) strategy in which generative artificial intelligence (AI) served as a second layer of feedback after peer feedback, and to examine changes in students’ anecdote writing skills. Following the Kemmis and McTaggart model, two cycles were conducted with 43 students of class X E at SMAN 20 Bandung, Indonesia, in August 2026. Writing was assessed with a weighted five-aspect analytic rubric scored independently by two raters (quadratic weighted κ = 0.67). Success was defined as a class mean of at least 75 and classical mastery of at least 75%, and gains were also expressed as normalised gain, categorised as low (< 0.3), medium (0.3–0.7), or high (≥ 0.7). The class mean rose from 60 (pre-cycle) to 67.5 (Cycle I) and 79 (Cycle II), and classical mastery from 27.9% to 41.9% and 76.7%, meeting both criteria (normalised gain = 0.48, medium; Wilcoxon p < .001), while student engagement rose from 65% to 85%. Gains were largest in spelling and language use and smallest in content and criticism, and five students declined because they wrote single-paragraph, purely comic texts. Students’ writing skills therefore improved after the implementation of the combined strategy; however, because TPS and AI were introduced together without a comparison group, the improvement cannot be attributed to AI alone.