Development of a causal inference framework for human-building interaction research

We develop a causal inference framework that enables the development of probabilistic causal models that explain occupant needs, behaviours, and their corresponding decision-making processes.

Status: Current

Research themes: Health and comfort Energy and GHG emissions

Research areas: Human-Building Interaction; Applied AI and ML Technologies for buildings

Primary Objective

The primary objective of this project is to develop a causal inference framework by applying modern causal inference methodologies. The second objective is to demonstrate the efficacy of the proposed framework by developing causal occupant models (i.e., discovery of causal knowledge) and advanced occupant-centric building operational solutions using the framework.

Approach

It has been challenging to discover new causal knowledge because of the difficulty in (i) conducting large-scale controlled experiments and (ii) untangling “causality” from “correlation” in observational datasets with conventional statistical and machine learning methods. To enable the discovery of potential causal knowledge from observational data, the framework will be equipped with state-of-the-art probabilistic causal inference techniques. The framework will allow (i) encoding prior knowledge from previous studies in formal probabilistic ways and (ii) combining available data from heterogeneous sources.

Findings

In the early stage of this project, we plan to identify and suggest a set of appropriate causal inference techniques for human-building interaction research based on a literature review and computational experiments. Once we develop a prototype of the causal inference framework, we will develop causal models explaining occupant behaviours in buildings with the prototype framework and present how such causal models (knowledge) can transform the way we develop solutions for occupants, comparing them with models developed with conventional methods.