Developing Thai Language Communication Competency with Intelligent Prompt through Generative AI

This study aimed to analyze the quality of a Thai language communication competency assessment developed using generative AI. A total of 151 students in Grade 5 participated in the study. The results found that The design of the intelligent prompt consists of the following elements: 1. Command goal 2. Situation, event, or condition of the question 3. Suitability for the test taker group 4. Test format 5. Dimension of measurement 6. Learning content and 7. Files necessary for AI to learn. Model fit the items generated by ChatGPT 4o, considering the Chi-square Likelihood values (χ² = 56.47, df= 5, p = .05), AIC and BIC are suitable as a multidimensional model. The reliability of the AI-generated test is within the acceptable standard, with a separate reliability value of 0.94, an EAP/PV value of 0.87, an alpha coefficient of 0.90, and a standard error of measurement of only 0.25, indicating a high level of test validity. AI can imitate humans so well that experts misidentify 61.11% of responses as human, showing the difficulty of distinguishing between Humans and AI.