Artificial Intelligence in Historical Research: Potential and Limits of Effectiveness

Abstract

The article thoroughly analyzes the strengths and weaknesses of applying artificial intelligence technologies in historical research using the example of topic modeling methods. The use of popular machine learning algorithms such as Latent Dirichlet Allocation (LDA) for analyzing large arrays of textual data is examined in detail. The key advantages of topic modeling are discussed, including the abilities to process large volumes of text, identify hidden thematic structures, and track topic dynamics over time. At the same time, significant limitations of this approach are considered, such as the assumption of topic stability in most models, poor interpretability of results, their instability and strong dependence on parameter settings. This requires a critical attitude towards the results obtained and their careful verification based on the subject knowledge of the researcher. Specific limitations analyzed include the static nature of topics in many models, weak interpretability of results, their instability, and high dependence on parameter settings. Based on the analysis, it is concluded that it is essential to balance artificial intelligence methods with traditional qualitative approaches in the humanities. The article provides concrete recommendations on the application of topic modeling in historical research to maximize the benefits while minimizing the risks.