AI Assisted Domain Modeling Explainability and Traceability

Domain Models are abstract representations of selected elements in a domain that is created in a collaborative process between domain and modeler experts. The participants share domain knowledge to conceptualize and reason about the elements that will create the domain models. Through this exchange, a comprehensive and
accurate representation of the domain is achieved, ensuring that the model captures the relevant aspects and relationships in the domain. Research in Artificial Intelligence (AI) has explored various methods to assist in the creation of domain models from text using Natural Language Processing (NLP) and Machine Learning (ML). Recent advancements with Large Language Models (LLMs) have shown that it is possible to create domain models using prompting techniques; however, the generated domain models contain errors and remain constrained by the performance of the LLM used. Despite the impressive capabilities of LLMs to create domain models, it is evident that it does not address the needs of domain and modelers experts that participate in the creation of domain models. Every AI technique has its advantages and limitations that must be integrated with human feedback in a collaboration process. Therefore, we propose an approach that incorporates human-AI collaboration supported by AI assistants that follows a dialogue approach to understand the users needs and purpose to suggest
relevant models. Our proposal combines symbolic and subsymbolic AI techniques with explainability and traceability of the decisions that assist to create domain models that are relevant for the users.