Routine Computing: A Systematic Review of Sensing Daily Life Dimensions Towards Human-Centered Goals

Paper Title

Routine Computing: A Systematic Review of Sensing Daily Life Dimensions Towards Human-Centered Goals

Publication Info

• Topic area: Computational modeling and sensing of human routines in daily life.
• Keywords: Routine computing, human activity recognition, context-aware systems, temporal granularity, behavior modeling, cognitive factors, adaptive systems, privacy, health monitoring, population-level insights.

Background and Problem

• Problem / challenge: Existing computational systems struggle to model and interpret the complexity of human routines, particularly in bridging low-level activity recognition with high-level intent, managing personalization versus generalization, addressing privacy concerns, and overcoming data limitations.
• Significance: Understanding human routines has practical implications for health monitoring, aging care, habit formation, and population-level insights, making it critical for advancing human-centered technologies.
• Motivation and related work: Previous research in Human Activity Recognition (HAR) and context-aware computing has focused on isolated activities or single-category behaviors, neglecting the temporal, contextual, and psychological dimensions that structure daily routines. This paper addresses these gaps by introducing and systematically reviewing the emerging field of routine computing.

Solution

• Proposed approach: A systematic review and taxonomy of routine computing, synthesizing 203 studies to identify key dimensions, goals, challenges, and future directions in the field.
• Novelty:
  1. First systematic review of routine computing, consolidating 203 studies across sensing, modeling, and application domains.
  2. Conceptualization of routine computing as distinct from activity recognition and context-aware computing.
  3. Development of a taxonomy along temporal, behavioral, cognitive, and adaptive dimensions.
  4. Identification of critical challenges and opportunities, including interpretability, personalization, ethical design, and data enrichment.
• Procedure and key techniques:
  1. Literature search using ACM DL and IEEE Xplore, yielding 2,631 records, screened down to 203 studies.
  2. Analysis of sensing strategies, modeled activities, cognitive factors, and adaptation mechanisms.
  3. Classification of studies by temporal granularity, behavioral contexts, cognitive dimensions, and variability detection.
  4. Identification of application goals and challenges in routine computing.

Results

• Concrete findings:
  • Temporal granularity: Studies span seconds-to-minutes (54 papers), days-to-weeks (69 papers), and months-to-years (80 papers).
  • Behavioral contexts: Routines involve solo actions (66 papers), mobility (71 papers), and social interactions (33 papers).
  • Cognitive factors: Key dimensions include prospective memory (4 papers), motivation (9 papers), and self-reflection (40 papers).
  • Variability and adaptation: Systems address routine deviations, provide interventions, and adapt models over time (83 papers).
• Advantage over baselines: Routine computing extends beyond activity recognition by integrating temporal, contextual, and cognitive dimensions, enabling richer and more meaningful insights into human behavior.
• Experiments / evaluation: Studies were evaluated based on sensing modalities, data granularity, user-study characteristics, and application goals, with applications in health monitoring, habit formation, adaptive assistance, and population-level insights.
• Limitations and future work:
  • Limited interpretability of high-level intent from low-level activities.
  • Challenges in balancing personalization and generalization in longitudinal modeling.
  • Privacy concerns with continuous monitoring.
  • Data scarcity and lack of diverse, large-scale datasets.
  • Future work includes integrating richer contextual cues, enhancing privacy-preserving mechanisms, and developing sustainable data collection practices.

Summary

This paper provides the first systematic review of routine computing, synthesizing 203 studies to define the field's conceptual dimensions, application goals, and challenges. It introduces a taxonomy encompassing temporal granularity, behavioral contexts, cognitive factors, and adaptation mechanisms, highlighting the potential of routine computing to support health, aging care, habit formation, and population-level insights. Persistent challenges include interpretability, personalization, privacy, and data limitations, with proposed solutions focusing on richer contextual integration, ethical design, and sustainable data practices. This work establishes routine computing as a critical domain for advancing human-centered technologies in everyday life.