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

TL;DR

PersonaDrift introduces a synthetic benchmark for evaluating machine learning methods in detecting subtle, progressive communication changes in dementia patients over time, emphasizing personalized and temporal modeling.

Contribution

This work presents a novel synthetic benchmark, PersonaDrift, for assessing anomaly detection methods in longitudinal language data of dementia patients, incorporating caregiver-informed behavioral variations.

Findings

Simple statistical models detect flattened sentiment in low-variability users.

Temporal models and personalization improve semantic drift detection.

Personalized classifiers outperform generalized models across tasks.

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.