User Perception Changes Following AI Agent Service Reorganization: A Time-Series Impact Analysis Based on Sentiment and Topic

With the rapid expansion of AI agent services, this study examines how service changes in SK Telecom's “Adot Phone” affect user perception. To investigate this, we analyzed Google Play Store review data using a combination of NLP techniques, including KoBERT-based sentiment analysis, Latent Dirichlet Allocation (LDA) topic modeling, and Interrupted Time Series Analysis to assess the impact of the service change. The results show that the service modification led to a notable drop in positive sentiment and a sharp increase in negative sentiment, largely due to core functionality issues and concerns about AI feature usability. The analysis also found that user discussions became increasingly focused on a smaller set of critical problems after the update. This study highlights the priority of service stability over technological innovation and offers a comprehensive framework for dynamically analyzing changes in user experience.