The Role of Prior User Interactions in Shaping ChatGPT's Response Patterns

This master thesis investigates the extent to which previous interactions with OpenAI's ChatGPT influence its future responses, focusing on usage patterns and the languages used. The study employs a mixed-method approach, combining qualitative and quantitative methods to examine how different types of user accounts affect ChatGPT's responses. In the qualitative phase, semi-structured interviews were conducted with Russian-speaking ChatGPT users and international students from the EC2U Master's Program "European Languages, Cultures and Societies in Contact". Respondents shared their experiences with ChatGPT, including their usage purposes, the languages they use, and general insights on its performance. They also provided examples of tasks they pose to the model and how they typically formulate their questions. This phase provides nuanced background information on usage patterns. The quantitative phase involves controlled experiments where participants follow specific instructions to ask standardized questions to ChatGPT. The responses were analyzed for variations and differences, aiming to identify correlations between previous usage patterns and the model's responses or to refute this assumption. The final analysis synthesizes the findings from both phases, detailing the experiment's results and discussing their implications for future research. This study aims to enhance the understanding of how ChatGPT interacts with different user accounts and the extent to which it retains and utilizes user-specific information.