AI CHATBOTS DRASTICALLY OUTPERFORM SECONDARY STUDENTS ON THE TEST OF ECONOMIC LITERACY

Abstract

This paper examines the performance of generative artificial intelligence models (OpenAI ChatGPT, Google Gemini, Microsoft Copilot) on the Test of Economic Literacy and compares their results with those of U.S. and Slovenian secondary school students. The study aims to assess whether AI systems demonstrate comparable or even superior levels of economic knowledge and to explore potential language-related differences in test performance. Descriptive statistics and item-level analysis are used to identify patterns and inconsistencies across languages when comparing AI test-takers and secondary students. Results show that AI models achieved near-perfect accuracy, while students performed significantly lower on average. The analysis also highlights differences in the length and explanatory depth of responses among the AI models. Findings contribute to current debates on assessment validity, AI use in economic education, and the future of multiple-choice testing in digitally augmented learning environments. The paper underscores the need for responsible integration of AI into educational frameworks, safeguarding the credibility of learning and the support for the sustainable development of economic knowledge.