COMPARATIVE ANALYSIS OF ETHICAL INCIDENT APPROACHES IN GENERATIVE ARTIFICIAL INTELLIGENCE APPLICATIONS UTILIZING LARGE LANGUAGE MODELS (LLM)

In this study, the response performances of generative artificial intelligence applications used in many different business areas to ethical situations were observed. It is important to carefully examine the responses produced by generative artificial intelligence applications with Large Language Model (LLM) in all business areas. In the study, the answers to sample ethical cases in the context of LLM were examined through 5 large generative artificial intelligence applications with LLM structure. The reasons, explanations, justification elements and interpretations given by Deepseek, ChatGpt 4o, QwenChat 2.5 Max, Gemini 2.0 Flash and Copilot applications were requested in their responses to ethical cases. According to the comparison results, the agreement and disagreement between the applications were also examined and the approaches of LLMs to ethical issues were revealed through the answers they gave. The reason for examining ethical cases in this study is that there is no absolute one-way answer to ethical cases. Ethical situation evaluations that vary from person to person are also a challenging problem area in terms of LLM applications. 13 sample ethical cases were explained and questions were asked to these 5 generative artificial intelligence applications without any prior preparation stage. In the answers received, generative artificial intelligence applications were asked to base them on and comment on them. As a result of the findings obtained, evaluations were made according to common points, differences, and general trends. These findings show that LLMs have made progress in addressing ethical issues. It has been observed that applications should continue to develop in producing consistent and fair solutions to ethical dilemmas. This situation emphasizes once again the importance of human control in the ethical decision-making processes of LLMs and the importance of the integration of ethical rules.