Emotional Drivers of Financial Decision-Making: Unveiling the Link Between Emotions and Stock Market Behavior (Part 3)

This study is the third part of a research project involving analyzing written documents to assess emotional charges. Those documents have been written following a three-day trading experience. The first part focused on the textual analysis of these documents using several measurement scales. The second part consisted of a comparative analysis of the results using three Artificial Intelligences (ChatGPT, Gemini, and DeepSeek), working through several queries designed to clarify the emotional fields under analysis as well as the general context that led to the documents being written. Our results demonstrated the ability of Artificial Intelligence to identify a general emotional context but revealed difficulties in approaching some emotional nuances. In this third study, we refined the queries addressed to the three Artificial Intelligences, particularly about the emotions to be taken into consideration, making a distinction between positive anticipation and negative anticipation, as well as a distinction between a good surprise and a bad surprise. The emotional typology used prevented us from considering how the two aspects could differ. Finally, the Artificial Intelligences were informed that a reward would be granted to the best-performing portfolio at the end of the experiment. Our results show that DeepSeek gives great importance to this parameter and generates many positive emotions, which neither ChatGPT nor Gemini do. Our results once again demonstrate a consistent ability of Artificial Intelligence to clarify the prevailing general emotional context, but difficulties in identifying consistently and uniformly the emotional nuances associated with the experiment. Concluding these three articles, it seems clear that Artificial Intelligences could be used to begin the process of understanding qualitative data, but it must be complemented by a human dimension in order to capture the emotional nuances that Artificial Intelligences analysis cannot.