A framework for the use of agent based modeling to simulate inter- and intraindividual variation in human behaviors
Citation:
Price, P., N. Brandon, K. Dionisio, R. Tornero-Velez, AND K. Isaacs. A framework for the use of agent based modeling to simulate inter- and intraindividual variation in human behaviors. 2016 Annual International Society of Exposure Science Meeting, Utrecht, NETHERLANDS, October 09 - 13, 2016.
Impact/Purpose:
Presents preliminary modeling results for Agent Based Modeling research. Efforts will provide the basis for a key module of the HEM software under the LC-HEM project.
Description:
Simulation of human behavior in exposure modeling is a complex task. Traditionally, inter-individual variation in human activity has been modeled by drawing from a pool of single day time-activity diaries such as the US EPA Consolidated Human Activity Database (CHAD). Here, an agent-based model (ABM) is used to simulate population distributions of longitudinal patterns of four macro activities (sleeping, eating, working, and commuting) in populations of adults over a period of one year. In this ABM, an individual is modeled as an agent whose movement through time and space is determined by a set of decision rules. The rules are based on the agent having time-varying “needs” that are satisfied by performing actions. Needs are modeled as increasing over time, and taking an action reduces the need. Need-satisfying actions include sleeping (meeting the need for rest), eating (meeting the need for food), and commuting/working (meeting the need for income). Every time an action is completed, the model determines the next action the agent will take based on the magnitude of each of the agent’s needs at that point in time. Different activities advertise their ability to satisfy various needs of the agent (such as food to eat or sleeping in a bed or on a couch). The model then chooses the activity that satisfies the greatest of the agent’s needs. When multiple actions could address a need, the model will choose the most effective of the actions (bed over the couch). In addition, multiple activities can be linked to a single decision (e.g., commuting must precede and follow working). An agent’s needs and the rate at which the needs increase over time are varied across agents and are correlated with the agents’ fixed personal attributes (e.g., age, gender, etc.) and household physical characteristics (distance between residence and work). Model parameters such as individuals’ rates of need increases are informed using data from CHAD. The advantage of ABM is that, unlike CHAD, it can provide information on human activity over periods of time longer than one day. We will present predictions for a population of adults for the four activities and compare the model outputs to the CHAD data. In future work we propose to extend this “need-based” framework to model usage of consumer products. For example, each agent is assigned personal hygiene and home cleanliness needs which drive their use of personal care products and household cleaning supplies.
