Drop by Drop: Facilitating the Energy-saving Behavior of Smart House/Office Inhabitants using AI and Gamification

Climate change is among the most important and impactful problems of this era. UN recognizes the severity of this problem by allocating three sustainable development goals on Climate Action, Responsible Consumption and Production, and Affordable and Clean Energy. To combat climate change, international treaties are signed, policymakers issue strict regulations, and technology innovators develop new methods, yet, we are failing to meet the goals. The common actor of this dynamic system is the human. In this multidisciplinary PhD proposal, we aim at developing methods to inspire human behavior change towards a more sustainable life, supported by (beyond)-state-of-the-art technology grounded on psychology and educational theory and practice. More specifically, we propose the use of Internet-of-Things (IoT), Artificial Intelligence (AI), gamification, and psychological intervention designs to facilitate behavior change.

To facilitate behavior change, first, we need to understand the behavior patterns, and psychological determinants of the inhabitants of a house or an office. Smart-houses and offices that are equipped with an array of IoT sensors (e.g., motion, temperature, power usage) assist in this pursuit by providing vital information regarding the behavior patterns of the inhabitants and how a living space is being used. We will use the IoT sensor data and train inhabitant behavior models using machine learning algorithms of varying complexity; (e.g., Bayesian network, random forest, LSTM deep learning) (Amayri et al., 2019; Trivedi & Badarla, 2020). These behavior models will include the following:
- Occupancy(room, time): Based on historical usage data, this model predicts the occupancy of a room at a given time.
- Heating(Occupancy(room, time), temperature): Based on the occupancy model and temperature measurement, this model predicts optimal heating settings.
- Power(Occupancy(room, time), appliance): Based on the occupancy model and electrical appliance, this model predicts the optimal appliance power settings.
- Activity(Occupancy(room, time), Sensor(x)): Based on the occupancy model and all available sensors, this model classifies the inhabitant activity at a given time.

These models will be used for predicting the inhabitants’ behavior patterns, which will create value in two ways: a) energy-saving automation, b) supporting our proposed pedagogical model to foster learning and behavior change. We will combine this with qualitative inquiry with the inhabitants to map the corresponding behavioral determinants.

Our research questions are as follows:
Q1) How can we build and implement IoT and AI models to capture the household/office inhabitant behavior?
Q2) How can we define (and detect) energy-saving behavior?
Q3) Which psychological determinants does an intervention need to target to produce behavior change?
Q4) How can we design a gamified pedagogical model that facilitates a long-lasting behavior change toward sustainable living?
Q5) What are the psychological parameters of an effective intervention design that supports the proposed pedagogical model?