PersonaDrift: A Benchmark for Temporal Anomaly Detection in Language-Based Dementia Monitoring

2025-11-20

AI Summary

This study addresses the automatic detection of progressive linguistic abnormalities—specifically affective flattening and semantic drift—in dementia patients’ spoken interactions. We introduce PersonaDrift, the first caregiver-informed, scalable longitudinal synthetic benchmark simulating 60-day dialogues between patients and a digital reminder system. Our methodological contribution is a personalized behavioral baseline modeling framework enabling simulation and evaluation of these two fine-grained temporal anomalies. We integrate unsupervised statistical process control (CUSUM/EWMA), one-class SVM, and GRU-BERT contextual embeddings, and train supervised classifiers under both generic and personalized settings. Experiments demonstrate that affective flattening is detectable in low-variability users using simple statistical models, whereas semantic drift requires joint temporal modeling and personalized baselines. Personalized classifiers significantly outperform generic ones on both tasks, validating the necessity of individualized behavioral modeling for early, sensitive detection of dementia-related language decline.

Technology Category

• Cognitive Modeling & Cognitive Systems: Affective Computing
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• User Modeling, Personalization and Recommendation: Studies of user behavior, including longitudinal effects of personalized systems
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Personalized, context-aware and across-device search

Abstract

People living with dementia (PLwD) often show gradual shifts in how they communicate, becoming less expressive, more repetitive, or drifting off-topic in subtle ways. While caregivers may notice these changes informally, most computational tools are not designed to track such behavioral drift over time. This paper introduces PersonaDrift, a synthetic benchmark designed to evaluate machine learning and statistical methods for detecting progressive changes in daily communication, focusing on user responses to a digital reminder system. PersonaDrift simulates 60-day interaction logs for synthetic users modeled after real PLwD, based on interviews with caregivers. These caregiver-informed personas vary in tone, modality, and communication habits, enabling realistic diversity in behavior. The benchmark focuses on two forms of longitudinal change that caregivers highlighted as particularly salient: flattened sentiment (reduced emotional tone and verbosity) and off-topic replies (semantic drift). These changes are injected progressively at different rates to emulate naturalistic cognitive trajectories, and the framework is designed to be extensible to additional behaviors in future use cases. To explore this novel application space, we evaluate several anomaly detection approaches, unsupervised statistical methods (CUSUM, EWMA, One-Class SVM), sequence models using contextual embeddings (GRU + BERT), and supervised classifiers in both generalized and personalized settings. Preliminary results show that flattened sentiment can often be detected with simple statistical models in users with low baseline variability, while detecting semantic drift requires temporal modeling and personalized baselines. Across both tasks, personalized classifiers consistently outperform generalized ones, highlighting the importance of individual behavioral context.

Problem

• Detecting gradual communication changes in dementia patients over time
• Identifying progressive sentiment flattening and semantic drift patterns
• Evaluating anomaly detection methods for longitudinal behavioral monitoring

Innovation

• Synthetic benchmark simulates dementia communication changes
• Evaluates statistical and machine learning anomaly detection methods
• Personalized classifiers outperform generalized approaches for detection