XAIport: A Service Framework for the Early Adoption of XAI in AI Model Development

Zerui Wang,Yan Liu,Abishek Arumugam Thiruselvi,Abdelwahab Hamou-Lhadj

DOI: https://doi.org/10.1145/3639476.3639759

2024-03-25

Abstract:In this study, we propose the early adoption of Explainable AI (XAI) with a focus on three properties: Quality of explanation, the explanation summaries should be consistent across multiple XAI methods; Architectural Compatibility, for effective integration in XAI, the architecture styles of both the XAI methods and the models to be explained must be compatible with the framework; Configurable operations, XAI explanations are operable, akin to machine learning operations. Thus, an explanation for AI models should be reproducible and tractable to be trustworthy. We present XAIport, a framework of XAI microservices encapsulated into Open APIs to deliver early explanations as observation for learning model quality assurance. XAIport enables configurable XAI operations along with machine learning development. We quantify the operational costs of incorporating XAI with three cloud computer vision services on Microsoft Azure Cognitive Services, Google Cloud Vertex AI, and Amazon Rekognition. Our findings show comparable operational costs between XAI and traditional machine learning, with XAIport significantly improving both cloud AI model performance and explanation stability.

Computer Science

What problem does this paper attempt to address?

The problem addressed in this paper is how to introduce Explainable Artificial Intelligence (XAI) in the early stages of AI model development to improve model quality and assurance. The researchers proposed a framework called XAIport, which consists of a series of microservices packaged as open APIs, aiming to provide early insights as quality assurance observability for the models. They focused on three key attributes of XAI: explanation quality (consistency across multiple XAI methods), architectural compatibility, and configurable operations. Through XAIport, configurable XAI operations can be integrated into the machine learning development process and the operational cost of adopting XAI on cloud platforms can be quantified.

The paper conducted experiments on three cloud computer vision services: Microsoft Azure Cognitive Services, Google Cloud Vertex AI, and Amazon Rekognition, to analyze the implementation effect of XAIport. The results showed that XAIport not only significantly improves the performance of cloud AI models but also enhances the stability of explanations. Moreover, compared to traditional machine learning, the operational cost remains comparable.

In addition, the paper discussed the cost analysis of XAI operations computation and deployment, demonstrating the time consumption differences among different cloud services and emphasizing the need for optimizing evaluation metrics. The paper concludes by stating that XAIport provides a reliable approach for quality assurance of open AI models and future work will involve developing a software development toolkit based on the XAIport framework to support automated deployment, dynamic A/B testing of open pre-trained models, and validation of new XAI methods.