Combining ChatGPT and knowledge graph for explainable machine learning-driven design: a case study

Xin Hu,Ang Liu,Yun Dai

DOI: https://doi.org/10.1080/09544828.2024.2355758

2024-05-22

Journal of Engineering Design

Abstract:Machine learning has been widely used in design activities, enabling more informed decision-making. However, high-performance machine learning models, often referred to as 'black-box', result in a lack of explainability regarding predictions. The absence of explainability erodes the trust between designers and these models and hinders human-machine collaboration for desirable design decisions. Explainable AI focuses on creating explanations that are accessible and comprehensible to stakeholders, thereby improving explainability. A recent advancement in the field of explainable AI involves leveraging domain-specific knowledge via knowledge graph. Additionally, the advent of large language models like ChatGPT, acclaimed for their ability to output domain knowledge, perform complex language processing, and support seamless end-user interaction, has the potential to expand the horizons of explainable AI. Inspired by these developments, we propose the novel hybrid method that synergizes ChatGPT and knowledge graph to augment post-hoc explainability in design context. The outcome is the generation of more contextual and meaningful explanations, with the added possibility of further interaction to uncover deeper insights. The effectiveness of the proposed method is illustrated through a case study on customer segmentation.

engineering, multidisciplinary

What problem does this paper attempt to address?

This paper primarily addresses the issue of interpretability in machine learning-driven design, particularly focusing on how to improve the post-hoc explainability of "black-box" models to enhance designers' trust in these models' outputs and promote human-machine collaboration for better design decisions.

To address the aforementioned issue, the authors propose an innovative approach that combines ChatGPT and Knowledge Graph (KG) to enhance the interpretability of machine learning-driven design. Specifically, this approach includes the following key steps:

1. **Knowledge Acquisition**: First, accumulate data, information, and knowledge related to the machine learning model and design tasks. This includes extracting local explanations from the training dataset using feature-based Explainable AI (XAI) methods and utilizing ChatGPT to obtain structured domain-specific knowledge.

2. **Knowledge Graph Construction**: Construct a knowledge graph to encode domain knowledge. This process involves defining nodes, attributes, edges, and relationships to structurally represent the previously acquired knowledge.

3. **Knowledge Graph Deployment**: Deploy the knowledge graph in practical applications to generate more context-relevant and meaningful explanations. By combining ChatGPT with the knowledge graph, a more interactive and supportive explanation process can be provided to designers, thereby better understanding the predictions of the machine learning model.

In summary, this paper aims to improve the interpretability of machine learning models in engineering design by combining the technical means of ChatGPT and knowledge graphs, thereby enhancing designers' trust in model outputs and promoting better design decisions.