Analisis Respon ChatGPT terhadap Penggunaan Campur Kode pada Penulisan Prompt

Abstract

The integration of artificial intelligence, particularly ChatGPT, has become an increasingly common phenomenon among university students in supporting various academic activities, such as searching for information, generating ideas, and completing assignments. On the other hand, in multilingual societies such as Indonesia, code-mixing is a natural, dynamic, and unavoidable linguistic practice in everyday communication, including interactions with artificial intelligence-based technologies. This study aims to analyze how ChatGPT responds to user prompts written using code-mixing, namely a combination of Indonesian with English or local languages. The study employed a qualitative descriptive method using sociolinguistic and human–computer interaction approaches. The research data were obtained by testing a series of simulated prompt texts containing various forms of code-mixing and analyzing the responses generated by ChatGPT. The findings indicate that ChatGPT is capable of understanding instructions expressed in mixed languages, adapting its response style to the communication context, and maintaining semantic accuracy in most cases. However, the study also reveals that the intensity and variation of code-mixing in user prompts significantly influence the accuracy of interpretation, the consistency of language style, and the formal structure of the generated output. Therefore, although ChatGPT demonstrates strong linguistic adaptability, the effectiveness of its responses remains influenced by the characteristics of code-mixing employed in user prompts.