Gamifying Sustainability: Harnessing AI and IoT for Human-Centric Energy Efficiency

Abstract

Addressing climate change requires urgent and sustained improvements in building energy efficiency. While the proliferation of Internet of Things (IoT) technologies has enabled granular real-time energy monitoring, presenting raw data through static dashboards often fails to foster lasting behavioural change. This disconnect, known as the intention-action gap, highlights that without active user engagement, the energy-saving potential of IoT systems remains largely unrealized. Current solutions frequently rely on a rigid "one-size-fits-all" approach, leading to reward fatigue and user disengagement over time. To bridge this gap, this dissertation, developed in collaboration with Fraunhofer Portugal AICOS, proposes GreenShift: a privacy-preserving edge-computing architecture built as a custom Home Assistant component. GreenShift transforms passive energy monitoring into an engaging user-centric experience by integrating real-time eco-feedback, meaningful gamification and Artificial Intelligence (AI)-driven personalization. To maintain motivation, the system dynamically adapts the difficulty of energy-saving tasks to the user's progress. Concurrently, a Reinforcement Learning from Human Feedback agent tailors proactive interventions to individual contexts. By actively modelling a user's Fatigue Index, the AI intelligently throttles notifications, preventing digital burnout while maximizing the relevance of actionable alerts. The architecture's efficacy was empirically validated through deployments in four distinct real-world settings (two laboratories and two smart homes). Results demonstrate that integrating dynamic gamification with adaptive AI significantly outperforms traditional static eco-feedback, which proved highly volatile. Active engagement with the gamified mechanics yielded substantial instantaneous power reductions, eliminating up to nearly 70% of the active baseload in localized setups, and successfully stabilized weather-normalized energy consumption. The study further revealed that gamification strategies must strictly align with the user's physical agency: localized settings thrived on specific device tasks, whereas shared spaces responded best to normative social comparisons. Furthermore, the AI agent successfully stabilized its decision-making policy over time, autonomously blocking approximately 60% of generated notifications to preserve user attention. Ultimately, this research validates that when energy conservation is framed as a personalized, context-aware and intrinsically rewarding experience, occupants successfully embed sustainable habits into their daily routines.