Comparative Analysis of Artificial Intelligence Models' Responses to Hadith History Questions: The Case of ChatGPT and Gemini


Ayhan M., Kılıç Z.

DINBILIMLERI AKADEMIK ARASTIRMA DERGISI-JOURNAL OF ACADEMIC RESEARCH IN RELIGIOUS SCIENCES, cilt.24, sa.3, ss.137-159, 2024 (ESCI, TRDizin) identifier

Özet

In this study, the responses of two artificial intelligence models, ChatGPT and Gemini, to eight questions related to the history of hadith were analyzed in detail using comparative analysis and multiple case study methods, which are qualitative research techniques. These questions were designed to test the knowledge, analysis, and synthesis levels of the AI models, addressing key concepts and historical developments in the science of hadith. The questions posed to the AI models are as follows: 1. Compare the oral and written transmission of hadith during the time of the Prophet Muhammad. 2. Explain the function of the Sunnah in relation to the Qur'an. 3. Provide information on hadith transmission before the period of codification. 4. Compare the definition of the Companions according to hadith scholars and juristic theorists, considering their respective purposes. 5. Provide information about the Companions based on the number of hadiths they transmitted. 6. Comment on the justice of the Companions. 7. Explain the fundamental differences between the Ahl al-Hadith and Ahl al-Ra'y. 8. Explain the use of isnad in the second century AH. The responses of both AI models to these questions were evaluated and scored by five hadith experts (two professors, two associate professors, and one doctor from three different universities). The experts reviewed the models' answers based on criteria such as accuracy, consistency, level of detail, contextual appropriateness, and fluency of expression, assigning scores to each response. This scoring process was designed to measure the competence of both models in providing information-based answers to topics related to hadith history. Consequently, the contributions and limitations of the AI models in the context of hadith history were evaluated, and the findings offer an analysis of how the models' capacities to generate knowledge, analysis, and synthesis may differ at an academic level. Additionally, both expert evaluations and the content, accuracy, level of detail, language used, and approach to the topics in each model's responses were examined in depth. The analysis reveals that ChatGPT approached the questions more systematically and in greater depth, whereas Gemini provided a more superficial and general perspective. For instance, when explaining the function of the Sunnah in relation to the Qur'an, ChatGPT addressed the general role of the Sunnah and its details comprehensively, examining its explanatory function in both general and specific contexts. When comparing the differences between Ahl al-Hadith and Ahl al-Ra'y, ChatGPT discussed the methodological differences of these two schools in detail. On the other hand, Gemini's responses were shorter and typically limited to basic information. For example, its response to the comparison of oral and written transmission of hadith during the time of the Prophet Muhammad was more superficial, avoiding details. Moreover, in discussing the justice of the Companions, Gemini offered a general explanation, while ChatGPT provided a more in-depth analysis and examined the various interpretations of justice. Although both AI models were successful in providing a general overview of hadith history, they exhibited shortcomings in certain concepts and terminology. In particular, both models fell short in addressing more technical topics such as the isnad system and the reliability of hadith, areas that require a more academic approach. Furthermore, instead of providing more in-depth information about the Companions and hadith transmission, Gemini simply listed names and presented a superficial treatment of the subject. In conclusion, this study highlights the potential of AI in fields such as hadith science, but it also demonstrates the limitations of these technologies. While ChatGPT provided a more detailed and academic approach, Gemini's responses remained simpler and more general. Both models contribute to the transmission of knowledge in Islamic studies, but they display deficiencies in terminology, depth, and use of sources. It is expected that AI models will provide important contributions to the integration of traditional sciences and the development of more comprehensive information systems. This analysis emphasizes the need for more sophisticated systems that offer in-depth knowledge in the field of hadith science. In addition, the response times and lengths of the answers provided by the AI models to the posed questions were also tracked and presented in tables and graphs. When considering all the questions asked, it was observed that Gemini responded faster and in less time compared to ChatGPT. However, when examining the content of the responses, it was noted that ChatGPT provided more detailed answers regarding the topics.