Smart Water Conservation: A Behaviourally-Grounded Recommender System for Demand Management Programs

Md Shamsur Rahim, Khoi Nguyen, Rodney A. Stewart, Damien Giurco, Michael BlumensteinWater 2025 · DOI

TL;DR

This paper introduces a novel framework for building recommender systems (RSs) that promote water conservation by integrating behavioral science, user experience (UX) design, and adaptive digital feedback. A prototype RS was developed and evaluated, showing strong user acceptance and highlighting the potential for personalized recommendations and incentives to drive sustainable behavior.

What Problem Does This Paper Address?

Water utilities face increasing pressure to manage demand due to climate change, population growth, and rising costs. While digital solutions offer promise, traditional water conservation campaigns often fall short due to a lack of personalization, low interactivity, and limited long-term impact. Existing recommender systems, while good at personalization, often lack a theoretical grounding in behavioral science, leading to recommendations that may not effectively motivate sustained behavior change. This paper aims to bridge this gap by proposing a generalized framework that combines behavioral theory with RS design principles to create more effective digital interventions for water demand management.

Methods

The core of this research is a novel framework that integrates four key components:

  1. Behavioral Objectives and Theory: This component draws on established behavioral theories like the Theory of Planned Behavior (TPB), the Transtheoretical Model (TTM), and Practice Theory to understand the determinants of water-use behavior. It uses a matrix of change objectives to link behavioral goals with psychological drivers.
  2. Recommender System Functions: This component focuses on the technical aspects of the RS. The prototype uses a knowledge-based approach, matching household characteristics (e.g., size, appliance type, consumption history) with conservation actions. This approach is favored for its transparency and scalability over opaque machine learning models.
  3. User Engagement and Experience Design: This component incorporates principles from the Technology Acceptance Model (TAM), Unified Theory of Acceptance and Use of Technology (UTAUT2), and User Experience (UX) design. The goal is to create an intuitive, easy-to-use interface that maximizes perceived usefulness and minimizes friction, fostering sustained engagement.
  4. Evaluation and Deployment Planning: This component outlines a structured approach for testing and rolling out the system, emphasizing iterative refinement based on user feedback.

The framework also incorporates the Intervention Mapping (IM) taxonomy to systematically select behavior change techniques like feedback, nudging, and goal-setting.

Datasets

Not discussed in the paper. The evaluation was conducted using an online survey of 300 participants, and the prototype system was designed to integrate with household profiling modules and potentially digital metering platforms.

Key Insights and Ideas

  • Behavioral Theory is Crucial for RS Design: Simply providing personalized recommendations isn't enough. Integrating insights from behavioral science (like TPB and TTM) is essential to ensure recommendations align with users' motivations, readiness, and existing routines.
  • Transparency Builds Trust: For public-facing utilities, a knowledge-based RS that explains why a recommendation is made is more trustworthy than a black-box algorithm.
  • Personalization and Usability are Key: While personalization is highly valued, the survey also revealed that ease of use and interactivity are equally, if not more, important for user engagement in water conservation contexts.
  • Incentives Motivate Action: Participants indicated that incentives, such as reward points, would be a significant motivator for adopting water-saving behaviors.
  • Scalability Matters: The knowledge-based approach of the prototype is highlighted for its scalability, making it suitable for widespread adoption by water utilities.

Key Results

The prototype RS was evaluated through an online survey of 300 participants. Key findings include:

  • High Acceptance: 82% of respondents agreed that personalized recommendations would help conserve water.
  • Motivation for Incentives: 76% of respondents indicated that incentives would motivate them to adopt water-saving behaviors.
  • Perceived Usefulness: Participants generally agreed that the prototype would support decision-making, reduce effort, and foster a positive attitude towards sustainable behavior.
  • Valued Features: Personalization was identified as the most important component, followed closely by ease of use, interactivity, progress tracking, and incentives.

Limitations and Future Work

The authors acknowledge several limitations:

  • Prototype Evaluation: The evaluation was based on user perceptions of a prototype using screenshots, not real-time interaction with a live system. This means the results reflect anticipated usefulness rather than measured behavioral outcomes.
  • Self-Reported Data: The study relied on self-reported attitudes, which may not perfectly predict long-term usage or impact.
  • Potential Sampling Bias: The survey recruitment method (university networks, sustainability groups) might have over-represented digitally literate and environmentally aware participants.

Future work should involve:

  • Real-World Pilot Trials: Deploying the RS in live settings to measure actual water use reductions, engagement patterns, and feature effectiveness.
  • Longitudinal Studies: Assessing long-term engagement and system trust.
  • Integration with Utility Systems: Exploring integration with utility billing, customer support, and rebate programs.
  • Broader Applications: Adapting the framework for other sustainability domains like energy conservation, waste reduction, and climate adaptation.

Summary updated

Abstract

Water utilities are increasingly turning to digital solutions to promote conservation behaviours among households; however, traditional campaigns often suffer from limited personalisation, low interactivity, and modest long-term impact. Though computer-tailored and recommender systems (RSs) may offer personalisation, these systems lack a generalised framework that integrates behavioural theory with system design. This study addresses this research gap by introducing a novel framework that unites behavioural science, user experience (UX) design, and adaptive digital feedback to foster water-conscious practices at the residential level. The model draws on established behavioural theories, including the Theory of Planned Behaviour, the Transtheoretical Model, and Intervention Mapping, to ensure that tailored recommendations align with users’ psychological drivers, behavioural readiness, and daily routines. An industry-first prototype RS was developed and evaluated through an online survey (N = 300), assessing user perceptions of relevance, motivation, ease of use, and likelihood of action. The results reveal strong support for personalised suggestions, with 82% of respondents agreeing that personalised recommendations would help conserve water, and 76% indicating incentives would motivate adoption. This evidence indicates early acceptance and high potential impact. This study also addresses a critical research gap: no generic model previously existed to guide the integration of RSs with behaviour change interventions in water demand management. Broader implications are also discussed for applying the model to other sustainability domains such as energy use, waste reduction, and climate adaptation.
BibTeX
@inproceedings{rahim2025smart,
  author = {Md Shamsur Rahim and Khoi Nguyen and Rodney A. Stewart and Damien Giurco and Michael Blumenstein},
  title = {{Smart Water Conservation: A Behaviourally-Grounded Recommender System for Demand Management Programs}},
  booktitle = {Water},
  year = {2025},
  doi = {10.3390/w17192798},
}
Figure