@masterthesis{Zuendorf, type = {Bachelor Thesis}, author = {Z{\"u}ndorf, Lennard Peter}, title = {Building an Interpretable Natural Language AI Tool based on Transformer Models and Approaches of Explainable AI}, school = {Hochschule f{\"u}r Technik und Wirtschaft Berlin}, pages = {48}, abstract = {The rapid advancement of Artificial Intelligence, particularly in the context of transformer-based Large Language Models, has ushered in a new era in consumer and business software. However, a significant aspect often overlooked is the integration of Explainable AI, specifically interpretable machine learning approaches. This thesis explores the relevance and applicability of existing XAI methods to contemporary transformer based models, widely used in applications such as ChatGPT. Through the development of an explainable application prototype, this study assesses the feasibility of integrating XAI techniques into practical, industry level applications. To this end, open-source technologies were employed to create a usable chatbot application prototype that provides decision-making explanations. Despite the successful development of a functional prototype, challenges remain in achieving userfriendly interpretability, particularly with complex models. Thus, this study demonstrates the feasibility of integrating XAI into chat applications but also highlights the vital areas for future research in enhancing post hoc interpretability methods. The prototype reveals limitations in current post hoc interpretability methods and underscores the need for improved approaches, especially considering the increasing size and complexity of LLMs.}, language = {en} }