EAGER: Developing a Mathematical Framework to Enable Bi-Directional Interactions of Humans with Smart Engineered Systems Using Relational Elements

University Of Southern California, Los Angeles CA

Investigators

Abstract

Buildings consume a staggering 38 percent of our nation's total energy use. Existing automation approaches to address this problem focus either on buildings, allowing them to better sense and respond to the behavior of their occupants, or focus on occupants, seeking to educate them on making more efficient energy choices. In contrast, this EArly-concept Grant for Exploratory Research (EAGER) project considers ways to enhance the interaction between buildings and occupants. The research team hypothesizes that user-building interactions will be most effective when building users establish a relationship of trust with building automation. By developing mathematical models and theory that amplify user capabilities through relational features, users are empowered to improve individual performance as well as building performance, while also improving societal well-being. To do so, the work draws on theories from the behavioral sciences to mathematically model when and how a building should interact with a user and how these interactions should be framed. The results will change the way we perceive and experience today's built environments, leading what could become the creation of unprecedented built environments that are attentive and have an identity. The project will enhance infrastructure for research and education by making the models and data available via a free research license, incorporating research findings into the engineering curriculum, disseminating research findings via publications, and national and international presentations. The modeling framework for user-virtual human agent interactions is the key contribution to smart engineered systems modeling and design and occurs at the intersection of engineering, the behavioral sciences and computational modeling. If successful, the mathematical framework will be used to design smart buildings that have two-way interactions with people. The research objectives contribute to the ultimate goal of enabling cyber-physical systems to interact and collaborate with humans. This project integrates experimental data into the mathematical models, testing the inclusion of relational elements embedded in the personification of a building. The models will predict which response is the most suitable for a building-user interaction. This model will also be informed and constrained by existing theoretical work on persuasion. The model will account for various contextual, temporal and personal factors as well as the changes in user response due to continuous interactions with the building. The multiple-step modeling methodology incorporates a combination of machine learning techniques, mathematical projections for the classification problem, and statistical models such as Markov model, and autoregressive moving average models. In particular, the contributions are twofold: (1) modeling user-building interactions using virtual human agents personifying buildings; and (2) performing fundamental research on how theories of human interpersonal trust and influence can inform the design of automation. The research will contribute to the fundamental understanding of human-machine teamwork, including elucidating theories of why and how people build connections with automated systems and advance our general understanding of how automation exhibiting relational features can facilitate behavior change in the population served by those systems.