AI and Sustainable Building Automation Systems (BAS): Envisioned Roles and Emerging Challenges

Paper Title

AI and Sustainable Building Automation Systems (BAS): Envisioned Roles and Emerging Challenges

Publication Info

• Topic area: Integration of AI into Building Automation Systems for energy efficiency and occupant comfort.
• Keywords: AI, Building Automation Systems, HVAC, sustainability, energy efficiency, occupant comfort, fairness, governance, transparency, socio-technical systems.

Background and Problem

• Problem / challenge: While AI holds promise for optimizing energy efficiency and occupant comfort in Building Automation Systems (BAS), its integration is hindered by fragile infrastructures, resource constraints, and organizational politics. Existing inequities in infrastructure and comfort further complicate its deployment.
• Significance: Buildings account for a significant share of global energy use and emissions, making sustainable BAS critical for achieving energy efficiency and carbon neutrality goals. AI could address inefficiencies but must navigate socio-technical complexities.
• Motivation and related work: Prior research in Sustainable HCI (SHCI) and Human-Building Interaction (HBI) has explored energy efficiency, governance, and occupant agency in building systems. However, gaps remain in addressing fairness, transparency, and the socio-technical dynamics of AI in BAS. This study builds on these fields to examine AI’s potential roles and challenges in institutional settings.

Solution

• Proposed approach: A qualitative study of stakeholder perspectives on AI in BAS, focusing on HVAC and occupant comfort at the University of Toronto, a campus with advanced BAS infrastructure and sustainability goals.
• Novelty:
  1. Identification of promises and limitations of AI for forecasting and occupancy modeling in BAS, based on empirical stakeholder insights.
  2. Conceptualization of fairness in AI as a situated practice of infrastructural governance rather than a universal property of algorithms.
  3. Proposal of communication as a practical instantiation of occupant agency, enabling transparency and contestability in automated systems.
• Procedure and key techniques:
  • Conducted 23 semi-structured interviews with energy professionals, AI researchers, and student representatives.
  • Used visual probes to illustrate AI applications and elicit reflections on benefits, concerns, and future roles.
  • Analyzed transcripts using a bottom-up coding approach to identify themes such as envisioned applications, infrastructural inequities, and governance challenges.

Results

• Concrete findings:
  • Participants envisioned AI improving forecasting, fault detection, occupancy modeling, and occupant interaction.
  • Data accuracy issues, model limitations, AI’s carbon footprint, and institutional politics were identified as key constraints.
  • Fairness was framed as a contested value, shaped by infrastructural inequities and stakeholder priorities.
  • Privacy, transparency, and accountability were deemed essential for AI’s legitimacy.
• Advantage over baselines:
  • AI was seen as potentially more adaptive and efficient than static rule-based systems, capable of integrating diverse datasets for predictive analytics and real-time adjustments.
• Experiments / evaluation:
  • The study used qualitative methods, including interviews and visual probes, to explore stakeholder perceptions and expectations before AI deployment.
  • Participants represented diverse roles, including energy professionals (15), AI researchers (4), and student representatives (4).
• Limitations and future work:
  • Perceptions were speculative as AI features in the campus BAS were not yet activated.
  • Governance actors and software vendors were not included in the study.
  • Future work should examine post-deployment experiences and align findings with institutional sustainability metrics.

Summary

This study explores the integration of AI into Building Automation Systems (BAS) at the University of Toronto, focusing on energy efficiency and occupant comfort. Stakeholders envisioned AI as a tool for forecasting, fault detection, and occupant interaction but highlighted challenges such as data accuracy, model limitations, and infrastructural inequities. Fairness was framed as a situated practice, and communication emerged as a key pathway for occupant agency. The findings emphasize the need for transparent, participatory, and equitable AI governance in BAS, offering design recommendations for communicative and fairness-oriented systems.