PELATIHAN VERIFIKASI INFORMASI DAN PENCEGAHAN HALUSINASI ARTIFICIAL INTELLIGENCE (AI) BAGI MAHASISWA

Abstract

The emergence of generative artificial intelligence (AI) technologies such as ChatGPT, Claude, Gemini, and Copilot has transformed the learning landscape in higher education in a very short time. Informatics students are the group with the highest adoption rates, utilizing AI not only for academic writing but also for code generation and error detection. However, this technology has a structural limitation in the form of AI hallucinations, which is the tendency of models to generate information, data, citations, references, and even programming library names that appear convincing but lack factual basis. This community service activity aims to improve students' AI literacy, information verification skills, and academic ethics awareness. The activity, carried out as a two-day intensive training followed by four weeks of mentoring, was attended by 42 fourth- and sixth-semester Informatics students from the AAS Indonesia Institute of Business Technology. The implementation method included seven stages: needs analysis, pre-test, delivery of five material modules, demonstration of AI hallucinations on scientific references and program code, guided verification practice, focus group discussions, and a post-test with evaluation and reflection. The evaluation instruments included a 25-item knowledge test, a verification skills rubric, and a satisfaction questionnaire. The results of the activity showed an increase in the average knowledge score from 49.16 in the pre-test to 83.61 in the post-test, with an N-Gain of 0.68, in the moderate category. A paired-sample t-test showed a highly significant difference with a significance value less than 0.001 and a very large effect size. The ability to identify fabricated references increased from 26.2% to 90.5%, while the proportion of participants who received AI output without tracing decreased from 40.5% to 2.4%. The participant satisfaction rate reached 91.3%. The activity concluded that the training based on demonstrations of real-world errors combined with hands-on verification practices effectively developed students' critical thinking without rejecting the productive use of AI.

Keywords: AI hallucinations; artificial intelligence literacy; information verification; academic integrity; informatics students