SPEAKER：TAN Chee-Wee（陈致玮），Professor，The Hong Kong Polytechnic University

TIME/DATE：2025.12.1   13:30

TENCENT：810 916 450

PW：291680

ABSTRACT：

With advances in Artificial Intelligence (AI), conventional static rule-based Anomaly Detection (AD) could be augmented by incorporating real-time feedback to dynamically adjust AD rules. When infusing such AI-augmented AD modules into incumbent systems, challenges occur in both design and implementation. On one hand, the design of AI-augment AD modules must incorporate dynamic AD rule updating mechanisms to respond to changes in operational environments and bolster AD accuracy. On the other hand, implementing AI-augmented AD module demands a careful consideration on the frequency of AD rule refinement given the constraint of system capabilities, leading to a tradeoff between accuracy and efficiency. Yet, despite the prevailing challenges, there is a dearth of research on theoretical and practical principles to guide the design and implementation of AI-augmented AD modules. To bridge this knowledge gap, we collaborate with a leading digital freight platform, which is currently employing a static rule-based AD module, to develop an AI-augmented AD system based on the dual-process theory. For the AI-augmented AD system, an AI module is deployed to predict anomalies after the rule module. We implemented three rounds of interventions to iteratively improve the AI-augmented system. Specifically, during the intervention process, we also explored two other system designs to combine the AI and rule modules: (i) a parallel combination that integrates these two modules through logical disjunction (labeled as ‘Rule while AI’) and (ii) a sequential combination that positions the AI module before the rule module (labeled as ‘AI-then-Rule’). We empirically validated the effectiveness of our Rule-then-AI system by comparing its performance to that of the two alternative systems.

GUEST BIO：

Chee-Wee Tan is a Professor at the Department of Management and Marketing in the Hong Kong Polytechnic University (PolyU). Before joining PolyU, Chee-Wee is a Professor w/ Special Responsibilities in Research Excellence at Copenhagen Business School (CBS). Chee-Wee received his PhD in Management Information Systems from the University of British Columbia. His research interests focus on design and innovation issues related to digital platforms. His work has been published in leading peer-reviewed journals such as MIS Quarterly (MISQ), Journal of Operations Management (JOM), Information Systems Research (ISR), Journal of Management Information Systems (JMIS), and the Journal of the Association for Information Systems (JAIS), among others. Chee-Wee is holding or has held Honorary and Guest Professorship positions at Chongqing University (CQU), Lingnan University (LNU), Monash University Malaysia (MUM), the University of New South Wales (UNSW), the University of Nottingham Ningbo China (UNNC), the University of Science and Technology of China (USTC), the Weizenbaum Institute for the Networked Society, and Zhejiang University (ZJU). Chee-Wee has served or is currently serving on the editorial boards for ACM Distributed Ledger Technologies: Research and Practice (DLT), DSS, EJIS, Industrial Management & Data Systems (IMDS), IEEE Transactions on Engineering Management (IEEE-TEM), Information & Management (I&M), Information Systems Journal (ISJ), Internet Research (IntR), JAIS, Journal of Computer Information Systems (JCIS), Journal of Management Analytics (JMA), JMIS, and MISQ. Finally, Chee-Wee is the Vice President of Publications for the Association for Information Systems.