A Consumer-Centric Framework for Measuring Product Obsolescence Using User-Generated Content and Large Language Models: Evidence From IoT Devices
本研究利用在线消费者评论和大型语言模型（ChatGPT-4o）识别产品过时因素，通过用户生成内容时间序列定义过时指数，量化各因素影响，为工程管理者提供基于实时客户洞察的过时缓解策略，尤其适用于物联网设备领域。
Identifying product obsolescence factors is essential for guiding sustainable design and extending product longevity. Unlike prior studies, this research leverages online consumer reviews to explore product obsolescence factors. First, ChatGPT-4o, an advanced pre-trained large language model (LLM), is utilized to identify these factors. User-generated content (UGC) time series-based product obsolescence indexes are then defined to quantify each factor's impact, offering a UGC-based complement to earlier methods that depended on expert judgment, supplier input, or survey data. By leveraging real-time customer insights, this approach aligns with Industry 4.0 principles, offering a UGC-based method that can support engineering managers to proactively address product obsolescence. It integrates factors' relative importance, determined through frequency–analytic hierarchy process (Freq-AHP), with their severity impact on consumers, assessed using the Robustly optimized Bidirectional Encoder Representations from Transformers approach (RoBERTa). This is further supported by a robustness check, where small perturbations were applied to sentiment intensities and all indices recalculated, confirming the aggregated obsolescence index remained stable across all product categories. This study focuses on consumer IoT devices, an area underexplored in existing literature, analyzing 47,695 online consumer reviews across nine product categories and selecting 4,771 online obsolescence-related reviews for detailed analysis. Findings reveal nineteen key factors and demonstrate a fundamental shift in obsolescence, indicating that product obsolescence of consumer IoT devices is increasingly driven by adaptability, interoperability, and digital resilience rather than physical durability. These insights demonstrate the potential of the proposed approach to inform product obsolescence mitigation strategies and guide more resilient, user-centered design in IoT ecosystems.