Home // AIHealth 2024, The First International Conference on AI-Health // View article

Human-AI Collaboration Cycle in the Development Stage of an AI-enabled System

Authors:
Pi-Yang Weng
Rua-Huan Tsaih
Hsin-Lu Chang

Keywords: engagement; domain expert; XAI; comprehensibility; trust

Abstract:
Explainable Artificial Intelligence (XAI) has garnered attention in the AI system development in recent years, especially in the high-stakes decision scenarios, such as medical and healthcare domains. In this paper, we present a framework named Human-AI Collaboration Cycle. The framework emphasizes the collaboration between domain experts and AI system in the development stage of an AI-enabled system through an introduction of XAI. We propose that the introduction of XAI can enhance domain experts’ engagement in the stages of model evaluation and validation, then further review and engage in the data preprocessing, which in turn, improves their comprehensibility and trust toward the system. To validate our framework, we will conduct a field experiment in a hospital, in which nurses, as domain experts, and AI engineers will work together to develop an AI-enabled fall detection system with model explainability. We will evaluate the role of Local Interpretable Model-agnostic Explanations (LIME), one of the noted XAI tools, in the proposed Human-AI Collaboration Cycle.

Pages: 12 to 16

Copyright: Copyright (c) IARIA, 2024

Publication date: March 10, 2024

Published in: conference

ISBN: 978-1-68558-136-7

Location: Athens, Greece

Dates: from March 10, 2024 to March 14, 2024