Skip to main content
NCBI home page
EPA Author Manuscripts logoLink to EPA Author Manuscripts
. Author manuscript; available in PMC: 2024 Apr 29.
Published in final edited form as: Environ Int. 2023 Jan 16;172:107753. doi: 10.1016/j.envint.2023.107753

Simulating patterns of life: More representative time-activity patterns that account for context

Robert Hubal a, Elaine A Cohen Hubal b,*
PMCID: PMC11057331  NIHMSID: NIHMS1985352  PMID: 36682205

Abstract

Background:

Complex contributions of environment to health are intimately connected to human behavior. Modeling of human behaviors and their influences helps inform important policy decisions related to critical environmental and public health challenges. A typical approach to human behavior modeling involves generating daily schedules based on time-activity patterns of individual humans, simulating ‘agents’ with these schedules, and interpreting patterns of life that emerge from the simulation to inform a research question. Current behavior modeling, however, rarely incorporates the context that surrounds individuals’ truly broad scope of activities and influences on those activities.

Objectives:

We describe in detail a range of elements involved in generating time-activity patterns and connect work in the social science field of behavior modeling with applications in exposure science and environmental health. We propose a framework for behavior modeling that takes a systems approach and considers the broad scope of activities and influences required to simulate more representative patterns of life and thus improve modeling that underlies understanding of environmental contributions to health and associated decisions to promote and protect public health.

Methods:

We describe an agent-based modeling approach reliant on generating a population’s schedules, filtering the schedules, simulating behavior using the schedules, analyzing the emergent patterns, and interrogating results that leverages general empirical information in a systems context to inform fit-for-purpose action.

Discussion:

We propose a centralized and standardized program to codify behavior information and generate population schedules that researchers can select from to simulate human behavior and holistically characterize human-environment interactions for a variety of public health applications.

Keywords: Behavior modeling, Time-activity patterns, Patterns of life, Exposure science

1. Introduction

Many of today’s challenging problems occur at the nexus of human interactions within the context of social, built, and natural environments. Research is focused on understanding holistic impacts of the full range of environmental exposures across the life course on human health (Price et al., 2022). To elucidate biological outcomes corresponding to measures of environmental contribution requires approaches for characterizing complex exposures that vary with time and space (Chatzidiakou et al., 2020; Fuentes et al., 2022). Public and environmental health decisions to intervene on these interactions require an understanding of individuals’ behaviors and influences on those behaviors. For instance, to prepare for, and respond to, health impacts associated with changes in climate and extreme weather, information is critical on likely locations and behaviors of at-risk populations (Holm et al., 2021; Liu et al., 2022; Wang et al., 2021), areas prone to floods and fire that are commonly accessed (Miranda et al., 2021), and potential for damage to infrastructure and utilities (NGA, 2018; Nogal et al., 2016)). To assess human-health risk, profiles of behavior in the context of an individual’s physical and social environment are required to characterize the nature of contact between human receptors and chemical or biological stressors (Beckx et al., 2013; Breen et al., 2020; Karrer et al., 2019; McPartland et al., 2022; Zeise et al., 2013). Analyses for these complex problems are often conducted at the level of national or subnational jurisdiction (Hague et al., 2021; Molander and Cohen, 2012). Ideally, guidance is provided for actions to be implemented as well at the community or local level, with the imperative that impacts of these actions be anticipated and protective for a range of vulnerable individuals and groups (Cuervo et al., 2017; CDC, 2015; Schroeder and Bouldin, 2019; U.S. EPA, 2019).

1.1. Methods to empirically study time-activity patterns

Profiles of behavior are commonly identified through time-activity information that is in turn traditionally collected through methods such as surveys and time use diaries, more recently using more direct measures including GPS tracking, wearable cameras, and other sensing devices. For instance, instruments administered across the population such as the American Time Use Survey, the Canadian Time Use Survey, the Japanese Survey on Time Use and Leisure Activities, and the Harmonized European Time Use Surveys all capture details on how individuals spend time working, doing chores, taking care of family, volunteering, socializing, and traveling. Similarly, time use diaries, through which individuals track daily activities, support study of behavior and behavior change, as with electricity usage (Suomalainen et al., 2019), social media usage (Mireku, 2021), and the coronavirus (Sullivan et al., 2021). These diary entries may be augmented and validated by sensed data (Cui et al., 2021; Gershuny et al., 2019; Kelly et al., 2015; Kiza-kevich et al., 2006; Thompson and Oelker, 2013; Ye et al., 2009).

1.2. Why simulation is needed to augment empirical study

In most cases, empirical information on time-activity patterns that can be directly applied for the problem at hand is limited and understanding individuals’ behaviors involves simulating their schedules for emergent “patterns of life” (Folsom-Kovarik et al., 2013). Patterns of life describes the rhythm of individuals’ daily activities, and how activities are influenced by contextual factors. Consider an individual employed in an office job. The individual has a typical daily regimen: Rise early, dress and prepare for work, eat, commute, work in an office setting, interact with co-workers, return home, spend time with family, sleep. The schedule, however, can easily change; the individual occasionally takes care of kids, shops for food, cleans, exercises, and relaxes. Another individual’s schedule may overlap with the first individual’s but differ because the second individual works in the service sector, or works from home, or also takes care of parents, or is in school and so spends part of each day studying. In essence, an individual is embedded in a social and environmental system that acts on the individual as the individual acts on the system.

A dominant approach to simulating schedules for emergent patterns of life uses agent-based models (Collins et al., 2015; Hammond, 2015; Steinbacher et al., 2021; Tracy et al., 2018). Each individual is represented as an agent with a personal schedule; in combination with contextual factors, the behavior modeling involves simulating the interactions of very many agents. The intent is to identify emergent patterns from numerous runs of agents’ activities over some time period (e. g., a day or week), including their interactions with each other (Jiang et al., 2020; Züfle et al., 2021) and their environment (Abraham et al., 2017; Rand and Stummer, 2021; Zhu et al., 2013), based on imposed rules. For instance, many iterations of community-level agent-based models, covering diverse activities among individuals, yield usable patterns of culturally-aware agents that exhibit behaviors that reflect locally meaningful customs or that change based on local preferences or demands (Bergier and Faucher, 2018; Tracy et al., 2018; Xu, 2012). The sum of runs of these models, that is, combined patterns of life, is used to answer questions, depending on the modeler’s intention, such as how individuals move in aggregate, who and where individuals meet, what they breathe or touch, what activities they undertake, how their health improves or deteriorates, or how the situation escalates or resolves.

1.3. Why current simulation approaches need improvement

The importance of individualized patterns of life is in showing how they contribute, say, to exposure to chemicals or biological stressors. For each part of an individual’s schedule, based on data derived from time-activity patterns, it is possible to weave in potential agent contacts and from that demonstrate possible exposure impact. But, currently, these schedules and their underlying contextual rules and data (when exactly on this day the individual wakes up; whether the individual eats lunch, and if so what the individual chooses to eat; whether the individual commutes this day or works from home; what time the individual stops work to attend to kids or parents) are developed case by case. As researchers embrace the complexity of human-environment interactions across the life course to elucidate relationships to health (e.g., Vineis and Barouki, 2022; Martin-Sanchez et al., 2020), and as policy makers work toward decisions that are health-protective for diverse individuals and populations, improved behavior-driven exposure models are required (Chesney and Duderstadt, 2022; Min et al., 2019), down to the individual level.

In this paper we suggest that there is a need for improved and more efficient behavior modeling, that explicitly considers system context. In line with a few other researchers who have recognized a need to take context and scale into account in their modeling, albeit for focused applications such as personal exposure (e.g., Chapizanis et al., 2021) or travel (e.g., Hafezi et al., 2021; Kitamura et al., 1997; Pendyala et al., 2013; Rezaee et al., 2019), we propose improving how patterns of life may be simulated through a process of codifying rules based on contextual factors, generating a population’s schedules using time-activity data, filtering the schedules, simulating agents, and interrogating results to evaluate decisions. The goal is to extend current behavior-driven exposure models by improving how context involving scale and across time and space influences behavior. Over time, application of the proposed approach would drive development of the most important data required to inform highest priority actions.

2. Human behavior and agent-based modeling

2.1. Using agent-based modeling to simulate patterns of life

A benefit of agent-based models is that their computational structure allows for much complexity, and therefore greater representativeness, particularly as it relates to capturing how individual contextual factors influence time-activity patterns. Current agent-based models of behavior do have important limitations. To simulate influences at the general community level, behavior models that describe many individuals’ activities and the contexts surrounding them commonly employ simplified agent-based approaches, particularly where many individuals (agents) are simulated. Agent-based models of large groups of individuals tend to simplify treatment of individuals, occasionally accounting for differences (e.g., in individuals’ ages, their jobs, their socioeconomic statuses, their close connections, their motivations) or with some individualized time-activity patterns possibly incorporated (Fehr et al., 2021), but mainly giving agents only basic needs to minimize computation. Whether simplified or not, public health and policy decisions may be based on such large group models.

Additional limitations relate to the resolution of agent behavior that is modeled. For instance, there are accurate portrayals of how pedestrians move and how crowds shift (Braun et al., 2003; Chen et al., 2017; Fridman and Kaminka, 2010; Kim et al., 2009; Pelechano et al., 2005), as well as team dynamics (Kozlowski and Chao, 2018), but aspects such as where the individual agents come from and where they are going have not typically mattered. That is, agents traditionally lack underlying goals, meaning that the rules agents are set up with are generally focused on particular behaviors (but see Fehr et al., 2021). Hence the crowds are not necessarily moving purposefully based on facets such as culture, neither during normal routine nor in response to chaotic or surprise events (Hubal et al., 2015; Shendarkar et al., 2008; Zheng et al., 2009). Meanwhile, commercial activity, weather, special events, even tendencies of denizens toward walking in streets versus on sidewalks can influence movement by affecting what paths are available to individuals, where individuals are tending to go, and how many individuals are out and about. Just as pedestrian activity is culture-dependent, so is vehicular activity. Aside from obvious cross-cultural differences such as traffic density (e.g., cars per person) or rights of way (e.g., drive on the left vs. right) there are culture-specific influences such as types of vehicles on the road (to include size of cars and trucks, presence of motorbikes, and bicycles, rickshaws, and other nonmotorized vehicles; Chao et al., 2015) and their pollutant contributions (Frey et al., 2020), general hurriedness, respect for the law, and time-of-day or day-of-week factors (rush hour characteristics, Sabbath day restrictions) (Hook and Diaz, 2003; Zaidel, 1992). Current agent-based models of vehicular traffic, when used in environmental studies, do not take all of these cultural influences into account.

Similar limitations are seen in the application of agent-based models to simulate behavior of smaller groups of agents (Brandon et al., 2020). There is work, for instance, to simulate exposure to household contaminants and airborne particles (Kvasnicka et al., 2020, Kvasnicka et al. (2022) but the daily schedules used are simplified. A game-based simulation explored localized aspects of social, built, and natural environments related to congregate living (Ottmann, 2020) but did not involve agent-based modeling. Agent-based models have been used to represent behaviors based on restricted sociocultural patterns (Scott et al., 2019) but narrowed to those that are relevant to evolutionary processes. When researchers have investigated culture-specific characteristics, they have not normally considered a myriad of alternative factors that can influence behavior. An obvious example is from machine learning, outside agent-based modeling, where inherent bias in the selection of data from which patterns are learned leads to algorithmic unfairness (Mehrabi et al., 2022), with the system not adapting (Zhang et al., 2021) to take into account exogenous factors. As another example, the use of agent-based modeling has benefit to criminology (Groff et al., 2019) but involves challenges to the validity of findings depending how many factors are considered, the empirical value of those factors (see also Keyes et al., 2017), and how relationships that are uncovered are applied to real-world situations. Further, agent-based models directed at simulating patterns of life within multi-generational households or congregate living, developed to inform effective decisions, need to consider individual factors and their effects on daily activity. An example from one of the authors’ past work in serious gaming (Hubal et al., 2015) involves religious Islamic cultures such as those prevalent in Asia and the Middle East and North Africa, in which men are instructed to pray five times per day, and may choose to do so individually from work or home or in a group at a mosque.

2.2. Data to simulate patterns of life

One solution to limitations of agent-based models is by data acquisition from publicly-available sources and by use of findings from socioeconomic and sociological studies. In such work, daily schedules based on patterns of activity at the individual level are modeled largely by reference to publicly-available data. For instance, synthesized populations use U.S. Census and geographic data to model households and individuals and account for schools, workplaces, and other central locations (Wheaton et al., 2009). There are numerous public sources of data to inform behavior models that cover demographics, housing characteristics, labor, education, health, and activity. As examples of global primary sources:

As needed, other government agencies make useful specialized data available: In the U.S., the Energy Information Administration collects household data as part of its Residential Energy Consumption Survey (https://www.eia.gov/consumption/residential); the Centers for Disease Control and Prevention collects health statistics in its National Health and Nutrition Examination Survey (https://www.cdc.gov/nchs/nhanes/index.htm); the Department of Health & Human Services conducts the National Survey on Drug Use and Health (https://www.samhsa.gov/data/data-we-collect/nsduh-national-survey-drug-use-and-health).

Particularly valuable data come from time-use studies that focus on how individuals allocate their time during a typical day (Bauman et al., 2019; Vrotsou et al., 2009). Data from government sources may supplement time-use data; the U.S. Environmental Protection Agency houses the Consolidated Human Activity Database (epa.gov/fera/consolidated-human-activity-database-chad) for use in human exposure and health studies and predictive models. There are also data from the many tools created to capture activity data or in-the-moment experience (ecological momentary assessment: Doherty et al., 2020; experience sampling method: van Berkel et al., 2018) as well as diary information (e.g., Kizakevich et al., 2018; Mondol et al., 2016; Nishiyama et al., 2020; Wang et al., 2021; Martin-Sanchez et al., 2020), developed for reasons as varied as monitoring post-deployed warfighters suffering from PTSD and depression to tracking behaviors to ward off coronavirus. Similar data consider travel and commuting (Cui et al., 2021; Kitamura et al., 1997; Yang, 2020). Individual time-activity patterns may also be extracted from GPS data (Ye et al., 2009). Even informal data, such as when individuals typically wake up (Lazovick, 2015), or what percentage of individuals skip lunch (Gervis, 2018), can be used to inform the time-activity patterns that are generated, by sampling from distributions of wake-up times or setting a rule for some likelihood of missing a meal. There are evolving methods to manage limitations with these various data, such as characterizing contributors of variability to important parameters (Isaacs et al., 2013), learning meaningful patterns in data to develop more representative parametric values (Lund et al., 2020a), and devising methods to impute for data lapse or loss (Meseck et al., 2016).

But publicly available data–and most details captured in time-use research, diary tools, and other behavior modeling work–may lack critical content. Some time-use data are collected along with context such as location and exposure through experience sampling (e.g., Lal et al., 2020) or momentary assessment (e.g., MacKerron and Mourato, 2013), but most data do not come with contextual details or social determinants such as the job status or actual job of the individual, the individual’s emotional state when engaging in an activity, social pressures in how to behave, or stressors that could influence a given day’s activity. Meanwhile, most activity-capture applications focus on specific information key to intervening appropriately (e.g., responses to a depression index, or recent diet and exercise) and not on everyday behaviors. To accommodate for these deficits, our recommended approach considers the broad range of scale and temporospatial context that may inform the generation of a population’s worth of activity schedules that can in turn be simulated, and emergent patterns analyzed. Whether it is a single entity that generates population schedules, or multiple entities generating their own schedules, our framework still describes the process in coming up with research or policy decisions, making obvious how or whether or not outcomes can be compared (e.g., Does one researcher’s or policymaker’s result make sense if applied to the other’s data?).

2.3. Influence of systems-based context on behavior

Patterns of life are a social construct, meaning that the interpretation of any patterns that result from simulation using agent-based modeling relies on recognizing the contextual factors that drive the result. For instance, different behaviors are, as just described, exhibited at different scales. (In this observation we follow the levels of influence promoted, for example, to support a health disparities research framework; Alvidrez et al., 2019. The concept is similar as well to organization levels such as individual-community-population; Yeakley and Cale, 1991.) Scale can mean focusing on individuals’ behaviors alone, in their households (Hu et al., 2023), in small groups (work or school environments, social settings), and in the larger community (Lippe et al., 2019). In addition, all of these activities are time and space dependent. Societal determinants such as agents’ health status, family status, work status, religion, and many other factors can affect schedules (deFur et al., 2007). Personal preferences, social influences, temporal considerations, and environmental or spatial influences all also play roles (Lund et al., 2020b). There are important considerations to the interpretation of emergent patterns of life, such as possible interactions among scale levels (e.g., some community-level properties may not be reducible to the properties of individuals) or that the data underlying the modeling may be incomplete (Silverman and Bryden, 2018). We thus propose that it is important to carefully describe contextual factors.

2.4. Scale

At their most basic, models of behavior are models of individuals’ behavior. This is not to say the models themselves are basic, but that behaviors occur at the individual level. It is at this level that both everyday and non-ordinary actions are of interest. As a matter of course individuals eat, sleep, groom themselves, drive, walk, converse, work, and play. They also variously clean, cook, garden, take care of children, parents, and pets, and travel. Every such action has potential implication for public health decisions. For instance, that individuals throughout their day come into contact with consumer goods at home, or polluted air on the streets, or pandemic viruses at the local restaurant, influences individualized exposure estimates (Lu, 2021; Schweizer et al.,2007; Yang et al., 2022). Workable models themselves can range in complexity, from the least granular that simplify behavior (e.g., the susceptible-infected-recovered epidemiology model presumes perfect mixing of infectious with susceptible individuals) to those that endorse intricate individual interactions.

Neighborhood, clan, and culture inform individuals’ health and wellbeing (Ward-Caviness et al., 2020), though links between these factors and behavior are only starting to be considered (Van Horne et al., 2022; Zota and VanNoy, 2021). Individuals engage with small groups in their homes, in their yards, with fellow students, with work colleagues, with extended family, on sports teams, as members of charitable boards. Individuals’ interactions, whether with other individuals in person or through social media, affect their subsequent behaviors. All kind of actions are influenced by these engagements and interactions, including individuals’ diets, product use, mass transit use, child rearing practices, time indoors studying, time outdoors playing in greenspace, participation in organizational or promotional events, attendance at mass gatherings. In turn, effective public and environmental health decisions need to take individuals’ actions affected by affiliated groups into account.

2.5. Temporospatial factors

Geographic location and time course may influence behaviors or the effects of behaviors. Weather, for instance, can determine if an activity is held outdoors (sunny, warm) or indoors (rainy, cold). Hilly versus flat terrain can affect gait or breathing as well as, say, an inclination for an individual at this simulated moment to traverse the distance on foot or bicycle, or using public transport. A given metropolitan area might provide cleaner drinking water, more efficient sanitation, or faster response to a mass casualty than another.

Models may take into account as well agent-centered factors that are a function of time and space. Such factors include personal, social, built, and natural environmental dimensions. For instance, descriptors can be associated with models of individuals at different time points or locations, referring to aspects of those individuals such as ‘parent’, ‘coworker’, ‘golfer’, or ‘Buddhist’. Rules may subsequently be used to define how different individuals behave (e.g., where they go and what they do) in different situations. Constraints in individual agents’ schedules, in turn, reflect different personally-relevant activities (childcare, work, play, praying), their timing (before or after work, in the evening, sporadic throughout the day), their location type (residence, workplace, field of play, religious site), and means of attaining those activities (travel to a location, or specific actions relevant to accomplishing the activities). Other agent-related factors are those behaviors that typically include group effect such as sharing all manner of objects with family members, traveling together with friends, conducting meetings with coworkers, and relying on housing and community resources. Constraints in individual agents’ schedules reflect simulated relationships (e.g., family relationships that meet realistic age and number checks, or workplace relationships that mirror real-world distributions of all those features of individuals of interest, that is, their ages, genders, religious affiliations, hobbies). Models of small group activity can be influenced by rules that define individuals in these contexts as ‘member’, ‘participant’, or ‘leader’ and affect how, for instance, a daily routine is simulated.

Temporospatial elements included in modeling might yield insight into environmental health trends (NAS, 2021). The exposome concept, as an example, is the “integrated function of exposure on our body including what we eat and do, our experiences, and where we live and work” (Vermeulen et al., 2020). As efforts to measure the cumulative influence of environment on health over the life course advance (Chung et al., 2021; Miller and Jones, 2014), behavior modeling will enhance how data on the social, built, and natural environments are used to glean insight.

3. Conceptual framework for simulating patterns of life

To advance efficient simulation of agent schedules for assessing emergent patterns of life for science-based analyses and decisions, we propose a common platform to codify relevant and emerging data and rules that address systems-based context, and to generate schedules of time-activity patterns that can be accessed by researchers and policy makers to support agent-based behavior modeling. This platform would enable leveraging of community knowledge generated by researchers and policy makers to inform new analyses and, in an iterative fashion, to enrich the community knowledge.

3.1. Improved behavior modeling

A simulation of agents for emergent patterns of life, and interpretation of findings, requires a systems approach that takes into account relevant scale and temporospatial context required to support a given application, decision, or action. The daily schedule used for modeling of a given agent is influenced by these factors; two otherwise similar agents may have different agendas due to differences in their age, employment, living status, marital status, religion, city of residence, or any number of other considerations. In terms of variability and exposure, inter- and intra-individual exposure profiles vary by context. To make population inferences about important drivers of exposure, especially where inter- and intra-individual variability is high, requires a number of iterations of agent-based simulation, each iteration relying on a realistic population of individuals (and their daily schedules, based on time-activity patterns) to serve as those agents. Only by generating representative schedules for a large number of agents is it feasible to take into account the variety and interdependency of contexts they involve and identify emergent patterns of life.

Such an approach produces agents by generating each agent’s daily schedule, possibly over many days, taking into account a spectrum of conditions that trigger rules codified from the literature, then integrates non-agent content including geographical, social, built, and natural environmental elements (Brandon et al., 2020). Further, individuals’ schedules can be allowed to contain branches with probabilistic selection based on context, so that specific activities can have a probability of occurring at specific times. Thus, with some probability an individual may feel ill and miss the day’s work, school, or other events. Similarly, an individual may avoid a location that this day happens to have pesticide spraying done, due to conditions such as allergy or pregnancy, potentially throwing off the remaining day’s schedule. To generate schedules, everyday activities (habits and practices such as waking up, dressing, using facilities, cleaning, eating, commuting, etc.) are given time slots. These time slots, and even whether or not they are assigned to a given agent on a given day, are dependent on sampling of demographic factors such as age, gender, employment, parental status, living location, and working location, as well as on a sampling of population-wide distributions of those activities (e.g., how different percentages of individuals rise before 6 am, 7 am, or 8 am; or what segment of the population does or does not each a given meal). At least two additional dimensions are then possible that may influence modeled behavior.

The first involves coding explicit or deriving learned rules that influence modeled agents at the individual level. It is partly based on work focused on modeling a ‘whole’ person to correlate, process, and act on information from multiple dynamic sources. In that work, a set of independent, communicating programs model physiological, psychological, and sociocultural processes for a given agent (Hubal and Heneghan, 2017). Rules may also be derived from findings of fit-for-purpose exposure studies, such as emissions studies investigating store hours (Dons et al., 2012) and pandemic-related media effects on compliance with protective measures (Melki et al., 2022). Though sufficient information may not yet exist in the literature to fully structure and parameterize these complex rule systems, specific applications would drive development of rules and parameters as key uncertainties and the most significant gaps are identified through iterative modeling and data development.

The second involves use of group- or population-level characteristics to influence rules for modeled agents. Defining descriptors for controlling agent behaviors largely involves identifying important characteristics of behavior. For instance, urban conventions regarding pedestrian movement, greetings and conversations on the street, sports activities, religious attendance, family structures, work, and influences of weather may guide modeling. Thus, if in one city a greeting normally occurs on the sidewalk and causes other individuals to move around the conversational partners (unless and until they move to the side) but not step off the sidewalk, whereas in another city a greeting can occur anywhere in the street and not affect pedestrian flow in the same way, then this observation leads to a ‘conversation location’ consideration to integrate into an urban scenario. Similarly, if in one city a preponderance of commercial activity takes place in open-air markets then a sudden rainstorm could cause a general clearing of shoppers, but not so in another city where most shoppers enter covered stores, and this observation could be implemented as a ‘marketplace exposure’. Other concepts of interest are air quality, number and percentage of cars (vs. manor animal-pulled vehicles as well as density on the streets), predominant passing on the right versus left (both pedestrians and vehicles), presence of relatively stationary street pushcarts that cause travelers to skirt around them, and similarity of activities between day and night. These concepts tend to influence individual and crowd behaviors. For instance, characters’ background guide their behaviors in response to a surprise event; certain characters in a crowd would flee to safety due to fear or family commitment, others approach with curiosity or intent to assist. The contextual conditions that can inform these rules are widely diverse, but would be narrowed based on specific needs associated with the given policy decision.

3.2. Simulation process

To simulate patterns of life a four-part process is required (see Fig. 1): Generate a population’s schedules based on time-activity data, filter the schedules, run an agent-based simulation using the schedules, and analyzing emergent patterns so as to interrogate results to inform an action.

Fig. 1.

Fig. 1.

Open in a new tab

Proposed behavior modeling process. At upper left, a centralized resource is responsible for generation of population schedules. At upper right, the population schedules are down-selected (filtered) using contextual parameters to those needed to address a given research question. At lower right, an agent-based model is used to simulate patterns of life of interest to the research question. At lower left, these patterns are evaluated for their impact and value to public health.

3.3. Generate schedules

We propose that consistency and standardization in generation would be ideal for problem-specific interrogation of behavioral models (i.e., to meet research or policy objectives). We imagine a central store of, for example, 300 + million agents representing the U.S. population (or the equivalent in any other country or region), and their daily schedules mapped out over 365 consecutive days (we advocate that, ideally, generation should produce a population’s year’s worth of schedules to account for numerous contextual factors). We suggest that a single entity might lead the generation of this store of agent schedules, producing results that are made openly available for researchers to employ. This store would be mined by individual researchers, as described next. To maintain relevant information this central entity would determine when to regenerate, for instance based on ‘significant’ changes in population data suggesting new time-activity patterns.

Ideally, the data generated would be application-independent and platform-agnostic. This ideal is tenuous, we understand: Those data needed by, say, an epidemiological model to accommodate “wicked” public health decision-making (Silverman et al., 2021) are likely quite different from those needed in modeling non-kinetic military operations (Bharathy et al., 2012). But it is not impossible; both applications involve models of complex interpersonal and physical behaviors based, ultimately, on characteristics of individuals that we advocate.

3.4. Filter schedules

To answer their research questions, researchers will care about distinct subsets of agents. Someone studying the effects of children’s exposure to household chemicals will need to simulate a different group of agents, for a different length of time, in different locations, than someone studying outcomes from healthcare workers’ exposure to a virus, or someone studying the differential impacts of school closures on parents’ income in urban and rural settings. We envision a set of centralized and custom tools to allow researchers to down-select, or filter, from the central store to those schedules of interest. The tools will employ elements of scale and temporospatial context to enable this filtering. If the analysis of simulated data is analogous to that of experimental data, in which a modeled system is connected directly to research questions of interest (e.g., exposure patterns), then a standardized set of tools and approaches to extract and analyze information from a central store could facilitate rigorous science.

3.5. Running agent-based models to simulate behavior

Addressing relevant behaviors is also researcher-specific, and involves running simulation. Simulation makes each agent’s schedule “real”, showing what that agent does at timepoints throughout the day and where those actions occur. Rules are used to define the context that influences daily schedules of agents to drive changes in patterns of behavior. Only through multiple agent-based modeling runs, exploring the solution space (Lee et al., 2015), will someone find how often two people happen to be at the same place at the same time, or how crowds leaving a sports event move differently in response to bad weather on Sundays versus Thursdays, or how a combination of inflation and drought, at least as reflected in a year-long timescale, might affect demand for certain products. The choice of method of simulation would be left to researchers, but their data would be derived from the central store.

3.6. Interrogate emergent patterns and take action

The result of filtering and simulating feeds into the given research or policy question. At this stage the researcher determines measures of interest, and calculation yielding impact of some kind. Also at this stage is the researcher’s ability to evaluate and validate that models produce reasonable results regarding that impact. The impact informs a policy or health decision, itself evaluated and possibly leading to a continuation of the analytic cycle.

4. Discussion

Behavior is dependent on scale, and time and space factors, thus it is not feasible to model behavior without data to simulate those factors. Currently, decision makers generalize from limited data. Better informed, more sustainable decision-making requires more comprehensive data on time-activity patterns to simulate individual agents for emergent patterns of life that address the most important behaviors for any given problem. Behavior models at different levels of abstraction–those that involve rules reflecting context and scale-can help to identify and predict patterns. Further, it is possible to update schedules as new data are identified and become available and rerun agent-based simulations to identify and predict new patterns or trends. This approach enables computational investigations that may yield insights on structural causes (i.e., context) and then facilitates testing of potential for structural changes (interventions) to elicit changes in emergent behavior in the system.

Today, a single central database would be challenged to include sufficient and representative data to account for many of the contextual factors discussed above, for the entire population of interest. For example, how would this central database obtain and incorporate data for every individual and location to properly simulate the frequency with which an individual goes to a market; weather patterns; the preponderance of open-air versus closed markets; the proximity of enclosed areas to flee to when a storm hits; the probability that a given individual flees a storm or stays to shop based on their risk-tolerance; when fleeing, how each individual travels to the new location based on their individual and community level propensity for moving on sidewalks or streets? In addition, to control the immense complexity of these models (and the massive amount of data required to parameterize them) might demand constraining the model to simulate only the behaviors, characteristics, rules, or contexts that a researcher believes to be most important for their research question. Perhaps more practical for the present would be a population of agents parameterized with a few key factors relevant to many questions and where there exists fine-scale national data (e.g., demographic information, occupations, some basic physical attributes) which a researcher would then take and build upon after identifying the additional context most relevant to their research question. This first step would facilitate standardized contextual data from which researchers may down-select to provide a starting point for consistency and completeness ultimately needed for representative modeling at scale. Results on these analyses will further identify key uncertainties and drive development of data to address the most critical and common gaps.

The simulation to identify representative patterns of life is quite complex, we argue, and would be enabled by a systems-based approach that includes important context and associated influences. This approach would serve to both better the generation of schedules and better the means of identifying emergent behavior. To continue to elucidate important contributions of environment to health, we encourage incorporating complexity required to simulate behavior that includes scale and time and space considerations that can support health protective actions decisions.

Acknowledgements

EPA disclaimer: The views expressed in this presentation are those of the authors and do not necessarily represent the views or policies of the U.S. EPA.

Footnotes

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability

No data was used for the research described in the article.

References

  1. Abraham YS, Zhao Z, Anumba CJ, Asadi S, 2017. Exploring agent-based modeling for human-centered energy consumption prediction. Proceedings of the Canadian Society for Civil Engineering Annual Conference and General Meeting 6, 660–669. [Google Scholar]
  2. Alvidrez J, Castille D, Laude-Sharp M, Rosario A, Tabor D, 2019. The National Institute on Minority Health and Health Disparities research framework. Am. J. Public Health 109 (S1), 16–20. 10.2105/AJPH.2018.304883. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Bauman A, Bittman M, Gershuny J, 2019. A short history of time use research: Implications for public health. BMC Public Health 19, 607. 10.1186/s12889-019-6760-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Beckx C, Lefebvre W, Degraeuwe B, Vanhulsel M, Kochan B, Bellemans T, Dhondt S, Int PL, 2013. Assessing the environmental impact associated with different trip purposes. Transp. Res. Part D: Transp. Environ 18, 110–116. 10.1016/j.trd.2012.10.002. [DOI] [Google Scholar]
  5. Bergier JY, Faucher C, 2018. Considering the needs and culture of the local population in contemporary military intervention simulations: An agent-based modeling approach. In: Faucher C (Ed.), Advances in Culturally-Aware Intelligent Systems and in Cross-Cultural Psychological Studies. Springer, Cham, pp. 211–233. 10.1007/978-3-319-67024-9_10. [DOI] [Google Scholar]
  6. Bharathy G, Tolk A, Yilmaz L, 2012. Agent-directed simulation for combat modeling and distributed simulation. In: Tolk A (Ed.), Engineering Principles of Combat Modeling and Distributed Simulation. Wiley, pp. 669–713. 10.1002/9781118180310.ch27. [DOI] [Google Scholar]
  7. Brandon N, Dionisio KL, Isaacs K, Tornero-Velez R, Kapraun D, Setzer RW, Price PS, 2020. Simulating exposure-related behaviors using agent-based models embedded with needs-based artificial intelligence. J. Eposure Sci. Environ. Epidemiol 30, 184–193. 10.1038/s41370-018-0052-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Braun A, Musse SR, de Oliveira LPL, Bodmann BEJ, 2003. Modeling individual behaviors in crowd simulation. Proceedings of the IEEE International Workshop on Program Comprehension, pp 143–148. DOI: 10.1109/CASA.2003.1199317. [DOI] [Google Scholar]
  9. Breen M, Chang SY, Breen M, Xu Y, Isakov V, Arunachalam S, Carraway MS, Devlin R, 2020. Fine-scale modeling of individual exposures to ambient PM2.5, EC, Nox, CO for the Coronary Artery Disease and Environmental Exposure (CADEE) study. Atmos. 11 (1), 1–65. 10.3390/atmos11010065. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. CDC, 2015. Homeowner’s and renter’s guide to mold cleanup after disasters. Centers for Disease Control and Prevention. https://www.cdc.gov/mold/pdfs/Homeowners_and_Renters_Guide.pdf. [Google Scholar]
  11. Chatzidiakou L, Krause A, Han Y, Chen W, Yan L, Popoola OAM, Kellaway M, Wu Y, Liu J, Hu M, Barratt B, Kelly FJ, Zhu T, Jones RL, 2020. Using low-cost sensor technologies and advanced computational methods to improve dose estimations in health panel studies: Results of the AIRLESS project. J. Eposure Sci. Environ. Epidemiol 30, 981–989. 10.1038/s41370-020-0259-6. [DOI] [PubMed] [Google Scholar]
  12. Chao Q, Deng Z, Jin X, 2015. Vehicle-pedestrian interaction for mixed traffic simulation. Comput. Anim. Virtual Worlds 26 (3–4), 405–412. 10.1002/cav.1654. [DOI] [Google Scholar]
  13. Chapizanis D, Karakitsios S, Gotti A, Sarigiannis DA, 2021. Assessing personal exposure using agent based modelling informed by sensors technology. Environ. Res 192, 110141 10.1016/j.envres.2020.110141. [DOI] [PubMed] [Google Scholar]
  14. Chen X, Li H, Miao J, Jiang S, Jiang X, 2017. A multiagent-based model for pedestrian simulation in subway stations. Simul. Model. Pract. Theory 71, 134–148. 10.1016/j.simpat.2016.12.001. [DOI] [Google Scholar]
  15. Chesney ML, Duderstadt K, 2022. Children’s rights, environmental justice, and environmental health policy in the United States. J. Pediatr. Health Care 36 (1), 3–11. 10.1016/j.pedhc.2021.08.006. [DOI] [PubMed] [Google Scholar]
  16. Chung MK, Rappaport SM, Wheelock CE, Nguyen VK, van der Meer TP, Miller GW, Vermeulen R, Patel CJ, 2021. Utilizing a biology-driven approach to map the exposome in health and disease: An essential investment to drive the next generation of environmental discovery. Environ. Health Perspect 129 (8), 85001. 10.1289/EHP8327. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Collins A, Petty M, Vernon-Bido D, Sherfey S, 2015. A call to arms: Standards for agent-based modeling and simulation. J. Artif. Soc. Soc. Simul 18 (3), 12. 10.18564/jasss.2838. [DOI] [Google Scholar]
  18. Cuervo I, Leopold L, Baron S, 2017. Promoting community preparedness and resilience: A Latino immigrant community-driven project following Hurricane Sandy. Am. J. Public Health 107, S161–S164. 10.2105/AJPH.2017.304053. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Cui Y, He Q, Bian L, 2021. Generating a synthetic probabilistic daily activity-location schedule using large-scale, long-term and low-frequency smartphone GPS data with limited activity information. Transport. Res. Part C: Emerging Technol 132, 103408 10.1016/j.trc.2021.103408. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. deFur PL, Evans GW, Cohen Hubal EA, Kyle AD, Morello-Frosch RA, Williams DR, 2007. Vulnerability as a function of individual and group resources in cumulative risk assessment. Environ. Health Perspect 115 (5), 817–824. 10.1289/ehp.9332. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Doherty K, Balaskas A, Doherty G, 2020. The design of ecological momentary assessment technologies. Interact. Comput 32 (1), 257–278. 10.1093/iwcomp/iwaa019. [DOI] [Google Scholar]
  22. Dons E, Beckx C, Arentze T, Wets G, Int Panis L, 2012. Shop opening hours and population exposure to NO2 assessed with an activity-based transportation model. In: Rauch S, Morrison GM (Eds.), Urban Environment: Proceedings of the Urban Environment Symposium, Springer, pp 161–70. DOI: 10.1007/978-94-007-2540-9_15. [DOI] [Google Scholar]
  23. Fehr A, Stoffa JA, Newton J, White J, 2021. Growing people: Generating realistic populations and explainable, goal directed behavior. Proceedings of the Interservice/Industry Training, Simulation, and Education Conference, 21253. [Google Scholar]
  24. Folsom-Kovarik JT, Schatz S, Jones RM, Bartlett K, Wray RE, 2013. Scalable models for patterns of life. Proceedings of the Innovative Applications of Artificial Intelligence Conference, 25, 1582–4. DOI: 10.5555/2891460.2891683. [DOI] [Google Scholar]
  25. Frey HC, Gadre D, Singh S, Kumar P, 2020. Quantification of sources of variability of air pollutant exposure concentrations among selected transportation microenvironments. Transp. Res. Rec 2674 (9), 395–411. 10.1177/0361198120929336. [DOI] [Google Scholar]
  26. Fridman N, Kaminka GA, 2010. Modeling pedestrian crowd behavior based on a cognitive model of social comparison theory. Computational Math. Organizational Theory 16 (4), 348–372. 10.1007/s10588-010-9082-2. [DOI] [Google Scholar]
  27. Fuentes ZC, Schwartz YL, Robuck AR, Walker DI, 2022. Operationalizing the exposome using passive silicone samplers. Curr. Pollut. Rep 8 (1), 1–29. 10.1007/s40726-021-00211-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Gershuny J, Harms T, Doherty A, Thomas E, Milton K, Kelly P, Foster C, 2019. Testing self-report time-use diaries against objective instruments in real time. Sociol. Methodol 50 (1) 10.1177/0081175019884591. [DOI] [Google Scholar]
  29. Gervis Z, 2018. Majority of Americans no longer eat 3 meals a day, study finds. Fox News, https://www.foxnews.com/food-drink/majority-of-americans-no-longer-eat-3-meals-a-day-study-finds. [Google Scholar]
  30. Groff ER, Johnson SD, Thornton A, 2019. State of the art in agent-based modeling of urban crime: An overview. J. Quant. Criminol 35, 155–193. 10.1007/s10940-018-9376-y. [DOI] [Google Scholar]
  31. Hafezi MH, Daisy NS, Millward H, Liu L, 2021. Framework for development of the scheduler for activities, locations, and travel (SALT) model. Transportmetrica A: Transport Sci. 18 (2), 248–280. 10.1080/23249935.2021.1921879. [DOI] [Google Scholar]
  32. Hague C, Orford R, Gaulton T, Thomas E, Hall L, Duarte-Davidson R, 2021. Development of a mechanism for the rapid risk assessment of cross-border chemical health threats. J. Eposure Sci. Environ. Epidemiol 31, 876–886. 10.1038/S41370-021-00344-2. [DOI] [PubMed] [Google Scholar]
  33. Hammond RA, 2015. Considerations and best practices in agent-based modeling to inform policy. In: Wallace R, Geller A, Ogawa VA (Eds), Assessing the Use of Agent-Based Models for Tobacco Regulation (Appendix A). National Academies Press; 2015. DOI: 10.17226/19018. [DOI] [PubMed] [Google Scholar]
  34. Holm SM, Miller MD, Balmes JR, 2021. Health effects of wildfire smoke in children and public health tools: a narrative review. J. Eposure Sci. Environ. Epidemiol 31 (1), 1–20. 10.1038/s41370-020-00267-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Hook W, Diaz O, 2003. Preserving and expanding the role of non-motorised transport. GTZ Transport and Mobility group, Germany. [Google Scholar]
  36. Hu Y, van Wee B, Ettema D, 2023. Intra-household decisions and the impact of the built environment on activity-travel behavior: A review of the literature. J. Transp. Geogr 106, 103485 10.1016/j.jtrangeo.2022.103485. [DOI] [Google Scholar]
  37. Hubal R, Folsom-Kovarik J, Woods A, Jones R, Carbone J, 2015. Patterns of life in the foreground and background: Practical approaches to enhancing simulation-based interaction skills training. Proceedings of the Behavior Representation in Modeling and Simulation Conference 24, 75–83. [Google Scholar]
  38. Hubal R, Heneghan J, 2017. Carolina virtual patient initiative [abstract]. Pharm. Educ 17 (1), 292. [Google Scholar]
  39. Isaacs K, McCurdy T, Glen G, Nysewander M, Errickson A, Forbes S, Graham S, McCurdy L, Smith L, Tulve N, Vallero D, 2013. Statistical properties of longitudinal time-activity data for use in human exposure modeling. J. Eposure Sci. Environ. Epidemiol 23, 328–336. 10.1038/jes.2012.94. [DOI] [PubMed] [Google Scholar]
  40. Jiang N, Burger A, Crooks AT, Kennedy WG, 2020. Integrating social networks into large-scale urban simulations for disaster responses. Proceedings of the ACM SIGSPATIAL International Workshop on GeoSpatial Simulation, 3, 52–55. DOI: 10.1145/3423335.3428168. [DOI] [Google Scholar]
  41. Karrer C, de Boer W, Delmaar C, Cai Y, Crépet A, Hungerbühler K, von Goetz N, 2019. Linking probabilistic exposure and pharmacokinetic modeling to assess the cumulative risk from the bisphenols BPA, BPS, BPF, and BPAF for Europeans. Environ. Sci. Tech 53 (15), 9181–9191. 10.1021/acs.est.9b01749. [DOI] [PubMed] [Google Scholar]
  42. Kelly P, Thomas E, Doherty A, Harms T, Burke Ó, Gershuny J, Foster C, 2015. Developing a method to test the validity of 24 hour time use diaries using wearable cameras: A feasibility pilot. PLoS One 10 (12), e0142198. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Keyes KM, Tracy M, Mooney SJ, Shev A, Cerdá M, 2017. Agent-based models-Bias in the face of discovery. Am. J. Epidemiol 186 (2), 146–148. 10.1093/aje/kwx090. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Kim J, Hill RW, Durlach P, Lane HC, Forbell E, Core M, Marsella SC, Pynadath DV, Hart J, 2009. BiLAT: A game-based environment for practicing negotiation in a cultural context. Int. J. Artif. Intell. Educ 19, 289–308. [Google Scholar]
  45. Kizakevich PN, Eckhoff R, Brown J, Tueller SJ, Weimer B, Bell S, Weeks A, Hourani LL, Spira JL, King LA, 2018. PHIT for Duty, a mobile application for stress reduction, sleep improvement, and alcohol moderation. Mil. Med 183 (S1), 353–363. 10.1093/milmed/usx157. [DOI] [PubMed] [Google Scholar]
  46. Kizakevich PN, McCartney M, Zhang A, Furberg R, Duncan S, Whitmore R, 2006. Indoor location tracking using Bluetooth proximity beacons. Epidemiology 17(6) S39. [Google Scholar]
  47. Kitamura R, Chen C, Pendyala RM, 1997. Generation of synthetic daily activity-travel patterns. Transp. Res. Rec 1607 (1), 154–162. 10.3141/1607-21. [DOI] [Google Scholar]
  48. Kozlowski SWJ, Chao GT, 2018. Unpacking team process dynamics and emergent phenomena: Challenges, conceptual advances, and innovative methods. Am. Psychol 73 (4), 576–592. 10.1037/amp0000245. [DOI] [PubMed] [Google Scholar]
  49. Kvasnicka J, Cohen Hubal EA, Ladan J, Zhang X, Diamond ML, 2020. Transient multimedia model for investigating the influence of indoor human activities on exposure to SVOCs. Environ. Sci. Tech 54 (17), 10772–10782. 10.1021/acs.est.0c03268. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Kvasnicka J, Cohen Hubal EA, Siegel J, Scott J, Diamond ML, 2022. Modeling clothing as a vector for transporting airborne particles and pathogens across indoor microenvironments. Environ. Sci. Tech 10.1021/acs.est.1c0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Lal RM, Das K, Fan Y, Barkjohn KK, Botchwey N, Ramaswami A, Russell AG, 2020.Connecting air quality with emotional well-being and neighborhood infrastructure in a US city. Environ. Health Insights 14. 10.1177/1178630220915488. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Lazovick M, 2015. Wake me up: What time do Americans start their day? Edison Research. https://www.edisonresearch.com/wake-me-up-series-2. [Google Scholar]
  53. Lee JS, Filatova T, Ligmann-Zielinska A, Hassani-Mahmooei B, Stonedahl F, Lorscheid I, Voinov A, Polhill JG, Sun Z, Parker DC, 2015. The complexities of agent-based modeling output analysis. Journal of Artificial Societies and Social. SIMULATION 18 (4). 10.18564/jasss.2897. [DOI] [Google Scholar]
  54. Lippe M, Bithell M, Gotts N, Natalini D, Barbrook-Johnson P, Giupponi C, Hallier M, Hofstede GJ, Le Page C, Matthews RB, Schlüter M, Smith P, Teglio A, Thellmann K, 2019. Using agent-based modelling to simulate social-ecological systems across scales. GeoInformatica 23, 269–298. 10.1007/s10707-018-00337-8. [DOI] [Google Scholar]
  55. Liu J, Varghese BM, Hansen A, Zhang Y, Driscoll T, Morgan G, Dear K, Gourley M, Capon A, Bi P, 2022. Heat exposure and cardiovascular health outcomes: a systematic review and meta-analysis. Lancet Planet Health. 6 (6), e484–e495. 10.1016/S2542-5196(22)00117-6. [DOI] [PubMed] [Google Scholar]
  56. Lu Y, 2021. Beyond air pollution at home: Assessment of personal exposure to PM2.5 using activity-based travel demand model and low-cost air sensor network data. Environ. Res 201:111549. 10.1016/j.envres.2021.111549. [DOI] [PubMed] [Google Scholar]
  57. Lund AM, Gouripeddi R, Facelli JC, 2020a. Generation and classification of activity sequences for spatiotemporal modeling of human populations. Online J. Public Health Inform 12 (1), e9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. Lund AM, Gouripeddi R, Facelli JC, 2020b. STHAM: An agent based model for simulating human exposure across high resolution spatiotemporal domains. J. Eposure Sci. Environ. Epidemiol 30, 459–468. 10.1038/s41370-020-0216-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  59. MacKerron G, Mourato S, 2013. Happiness is greater in natural environments. Glob. Environ. Chang 23 (5), 992–1000. 10.1016/j.gloenvcha.2013.03.010. [DOI] [Google Scholar]
  60. Martin-Sanchez F, Bellazzi R, Casella V, Dixon W, Lopez-Campos G, Peek N, 2020. Progress in characterizing the human exposome: A key step for precision medicine. Yearb. Med. Inform 29 (1), 115–120. 10.1055/s-0040-1701975. [DOI] [PMC free article] [PubMed] [Google Scholar]
  61. McPartland J, Shaffer RM, Fox MA, Nachman KE, Burke TA, Denison RA, 2022. Charting a path forward: Assessing the science of chemical risk evaluations under the Toxic Substances Control Act in the context of recent National Academies recommendations. Environ. Health Perspect 130 (2), 25003. 10.1289/EHP9649. [DOI] [PMC free article] [PubMed] [Google Scholar]
  62. Mehrabi N, Morstatter F, Saxena N, Lerman K, Galstyan A, 2022. A survey on bias and fairness in machine learning. ACM Comput. Surv 54 (6), 1–35. 10.1145/3457607. [DOI] [Google Scholar]
  63. Melki J, Tamim H, Hadid D, Farhat S, Makki M, Ghandour L, Hitti E, 2022. Media exposure and health behavior during pandemics: The mediating effect of perceived knowledge and fear on compliance with COVID-19 prevention measures. Health Commun. 37 (5), 586–596. 10.1080/10410236.2020.1858564. [DOI] [PubMed] [Google Scholar]
  64. Meseck K, Jankowska MM, Schipperijn J, Natarajan L, Godbole S, Carlson J, Takemoto M, Crist K, Kerr J, 2016. Is missing geographic positioning system data in accelerometry studies a problem, and is imputation the solution?. Geospatial. Health 11 (2). 10.4081/gh.2016.403. [DOI] [PMC free article] [PubMed] [Google Scholar]
  65. Miller GW, Jones DP, 2014. The nature of nurture: Refining the definition of the exposome. Toxicol. Sci 137 (1), 1–2. 10.1093/toxsci/kft251. [DOI] [PMC free article] [PubMed] [Google Scholar]
  66. Min E, Gruen D, Banerjee D, Echeverria T, Freelander L, Schmeltz M, Saganić E, Piazza M, Galaviz VE, Yost M, Seto EYW, 2019. The Washington state environmental health disparities map: Development of a community-responsive cumulative impacts assessment tool. Int. J. Environ. Res. Public Health 16 (22), 4470. 10.3390/ijerph16224470. [DOI] [PMC free article] [PubMed] [Google Scholar]
  67. Miranda ML, Callender R, Canales JM, Craft E, Ensor KB, Grossman M, Hopkins L, Johnston J, Shah U, Tootoo J, 2021. The Texas flood registry: A flexible tool for environmental and public health practitioners and researchers. J. Eposure Sci. Environ. Epidemiol 31823–31831. 10.1038/s41370-021-00347-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
  68. Mireku MO, 2021. Waking activities and sleep: Analysis of United Kingdom adolescents’ daily time-use diaries. J. Adolesc. Health 68 (2), 385–393. 10.1016/j.jadohealth.2020.05.050. [DOI] [PubMed] [Google Scholar]
  69. Molander L, Cohen A, 2012. EU and US regulatory approaches to information on chemicals in products: Implications for consumers. Eur. J. Risk Regulation 3 (4), 521–533. 10.1017/S1867299X00002440. [DOI] [Google Scholar]
  70. Mondol AS, Ra HK, Salekin A, Yoon HJ, Kubovy M, Son SH, Stankovic JA, 2016. LifeMaps: An automated diary system based on the structure of lives. Proceedings of the ACM International Conference on Embedded Networked Sensor Systems, 14, 348–349. DOI: 10.1145/2994551.2996700. [DOI] [Google Scholar]
  71. NAS, 2021. Leveraging advances in remote geospatial technologies to inform precision environmental health decisions: Proceedings of a workshop-in brief. National Academies of Sciences, Engineering, and Medicine. Washington, DC: The National Academies Press. DOI: 10.17226/26265. [DOI] [PubMed] [Google Scholar]
  72. NGA, 2018. GEOINT services: Mobile awareness GEOINT environment (MAGE). National Geospatial Intelligence Agency. http://ngageoint.github.io/MAGE/. [Google Scholar]
  73. Nishiyama Y, Yonezawa T, Sezaki K, 2020. SelfGuard: Semi-automated activity tracking for enhancing self-protection against the COVID-19 pandemic. Proceedings of the ACM International Conference on Embedded Networked Sensor Systems, 18, 780–1. DOI: 10.1145/3384419.3430592. [DOI] [Google Scholar]
  74. Nogal M, O’Connor A, Caulfield B, Brazil W, 2016. A multidisciplinary approach for risk analysis of infrastructure networks in response to extreme weather. Transp. Res. Procedia 14, 78–85. 10.1016/j.trpro.2016.05.043. [DOI] [Google Scholar]
  75. Ottmann DA, 2020. Co+living design: Participatory design simulation gamification for life-enhancing built environments in age-appropriate, inclusive and multi-generational co-living buildings without limitation. Pixo: J. Arch., City Contemporaneity 4 (14), 16–29. 10.15210/PIXO.V4I14.19883. [DOI] [Google Scholar]
  76. Pelechano N, O’Brien K, Silverman B, Badler N, 2005. Crowd simulation incorporating agent psychological models, roles and communication. Proceedings of the International Workshop on Crowd Simulation, pp. 21–30, 2005. [Google Scholar]
  77. Pendyala RM, Bhat CR, Goulias KG, Paleti R, Konduri K, Sidharthan R, Christian KP, 2013. SimAGENT population synthesis. Southern California Association of Governments. [Google Scholar]
  78. Price EJ, Vitale CM, Miller GW, David A, Barouki R, Audouze K, Walker DI, Antignac JP, Coumoul X, Bessonneau V, Klánová J, 2022. Merging the exposome into an integrated framework for “omics”. sciences. iScience 25 (3), 103976. 10.1016/j.isci.2022.103976. [DOI] [PMC free article] [PubMed] [Google Scholar]
  79. Rand W, Stummer C, 2021. Agent-based modeling of new product market diffusion: an overview of strengths and criticisms. Ann. Oper. Res 305, 425–447. 10.1007/s10479-021-03944-1. [DOI] [Google Scholar]
  80. Rezaee H, Kalantari N, Babaei M, 2019. New methodology for synthesizing population in metropolitans. J. Traffic Transport. Eng 7, 181–189. 10.17265/2328-2142/2019.04.005. [DOI] [Google Scholar]
  81. Schroeder J, Bouldin ED, 2019. Inclusive public health preparedness program to promote resilience in rural Appalachia (2016–2018). Am. J. Public Health 109 (S4), S283–S285. 10.2105/AJPH.2019.305086. [DOI] [PMC free article] [PubMed] [Google Scholar]
  82. Scott R, MacPherson B, Gras R, 2019. EcoSim, an enhanced artificial ecosystem: Addressing deeper behavioral, ecological, and evolutionary questions. Cognitive Arch. 94, 223–278. 10.1007/978-3-319-97550-4_14. [DOI] [Google Scholar]
  83. Shendarkar A, Vasudevan K, Lee S, Son YJ, 2008. Crowd simulation for emergency response using BDI agents based on immersive virtual reality. Simul. Model. Pract. Theory 16 (9), 1415–1429. 10.1016/j.simpat.2008.07.004. [DOI] [Google Scholar]
  84. Silverman E, Bryden J, 2018. Modelling for the social sciences. In: Methodological Investigations in Agent-Based Modelling. Springer, Cham, pp. 85–106. 10.1007/978-3-319-72408-9_5. [DOI] [Google Scholar]
  85. Silverman E, Gostoli U, Picascia S, Almagor J, McCann M, Shaw R, Angione C, 2021.Situating agent-based modelling in population health research. Emerg. Themes Epidemiol 18, 10. 10.1186/s12982-021-00102-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  86. Schweizer C, Edwards RD, Bayer-Oglesby L, Gauderman WJ, Ilacqua V, Jantunen MJ, Lai HK, Nieuwenhuijsen M, Künzli N, 2007. Indoor time-microenvironment-activity patterns in seven regions of Europe. J. Eposure Sci. Environ. Epidemiol 17, 170–181. 10.1038/sj.jes.7500490. [DOI] [PubMed] [Google Scholar]
  87. Steinbacher M, Raddant M, Karimi F, Camacho Cuena E, Alfarano S, Iori G, Lux T, 2021. Advances in the agent-based modeling of economic and social behavior. SN Business 7 Economics, 1, 99. DOI: 10.1007/s43546-021-00103-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  88. Sullivan O, Gershuny J, Sevilla A, Foliano F, Vega-Rapun M, Lamote de Grignon J, Harms T, Walthéry P, 2021. Using time-use diaries to track changing behavior across successive stages of COVID-19 social restrictions. Proceedings of the National Academy of Sciences, 118(35), e2101724118. DOI: 10.1073/pnas.2101724118. [DOI] [PMC free article] [PubMed] [Google Scholar]
  89. Suomalainen K, Eyers D, Ford R, Stephenson J, Anderson B, Jack M, 2019. Detailed comparison of energy-related time-use diaries and monitored residential electricity demand. Energ. Build 183, 418–427. 10.1016/j.enbuild.2018.11.002. [DOI] [Google Scholar]
  90. Thompson ME, Oelker A, 2013. Use of participant-generated photographs versus time use diaries as a method of qualitative data collection. Int. J. Qualitative Methods 12. 10.1177/160940691301200133. [DOI] [Google Scholar]
  91. Tracy M, Cerdá M, Keyes KM, 2018. Agent-based modeling in public health: Current applications and future directions. Annu. Rev. Public Health 39, 77–94. 10.1146/annurev-publhealth-040617-014317. [DOI] [PMC free article] [PubMed] [Google Scholar]
  92. U.S. EPA, 2019. Wildfire smoke: A guide for public health officials. U.S. Environmental Protection Agency, EPA-452/R-19-901; 2019. https://oehha.ca.gov/media/downloads/public-information/report/wildfire-smoke-guide-revised-2019.pdf. [Google Scholar]
  93. van Berkel N, Ferreira D, Kostakos V, 2018. The experience sampling method on mobile devices. ACM Comput. Surv 50 (6), 1–40. 10.1145/3123988. [DOI] [Google Scholar]
  94. Van Horne YO, Alcala CS, Peltier RE, Quintana PJE, Seto E, Gonzales M, Johnston JE, Montoya LD, Quirós-Alcalá L, Beamer PI, 2022. An applied environmental justice framework for exposure science. J. Eposure Sci. Environ. Epidemiol 10.1038/s41370-022-00422-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
  95. Vermeulen R, Schymanski EL, Barabási AL, Miller GW, 2020. The exposome and health: Where chemistry meets biology. Science 367 (6476), 392–396. 10.1126/science.aay3164. [DOI] [PMC free article] [PubMed] [Google Scholar]
  96. Vineis P, Barouki R, 2022. The exposome as the science of social-to-biological transitions. Environ. Int 165, 107312 10.1016/j.envint.2022.107312. [DOI] [PubMed] [Google Scholar]
  97. Vrotsou K, Ellegård K, Cooper M, 2009. Exploring time diaries using semi-automated activity pattern extraction. Electronic Int. J. Time Use Res 6 (1), 1–25. 10.13085/eIJTUR.6.1.1-25. [DOI] [Google Scholar]
  98. Wang S, Wu CYH, Richardson MB, Zaitchik BF, Gohlke JM, 2021a. Characterization of heat index experienced by individuals residing in urban and rural settings. J. Eposure Sci. Environ. Epidemiol 31, 641–653. 10.1038/s41370-021-00303-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  99. Wang Y, Li B, Gouripeddi R, Facelli JC, 2021b. Human activity pattern implications for modeling SARS-CoV-2 transmission. Comput. Methods Programs Biomed 199, 105896 10.1016/j.cmpb.2020.105896. [DOI] [PMC free article] [PubMed] [Google Scholar]
  100. Ward-Caviness CK, Pu S, Martin CL, Galea S, Uddin M, Wildman DE, Koenen K, Aiello AE, 2020. Epigenetic predictors of all-cause mortality are associated with objective measures of neighborhood disadvantage in an urban population. Clin. Epigenetics 12 (1), 44. 10.1186/s13148-020-00830-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  101. Wheaton WD, Cajka JC, Chasteen BM, Wagener DK, Cooley PC, Ganapathi L, Roberts DJ, Allpress JL, 2009. Synthesized population databases: A US geospatial database for agent-based models. RTI Press; 10, 905. [DOI] [PMC free article] [PubMed] [Google Scholar]
  102. Xu JW. Hierarchical graphical models for social and behavioral analysis and forecasting. In Abraham A, Hassanien AE (eds), Computational Social Networks (pp 25–49). Springer; 2012. DOI: 10.1007/978-1-4471-4048-1_2. [DOI] [Google Scholar]
  103. Yang J, Wu W, Liu M, Bi J, 2022. Leveraging individual-level data to advance air pollution health risk management. Frontiers of Engineering Management, 9, 337–342. DOI: 10.1007/s42524-022-0189-1. [DOI] [Google Scholar]
  104. Yang T, 2020. Understanding commuting patterns and changes: Counterfactual analysis in a planning support framework. Environ. Planning B: Urban Anal. City Sci 47 (8), 1440–1455. 10.1177/2399808320924433. [DOI] [Google Scholar]
  105. Ye Y, Zheng Y, Chen Y, Feng J, Xie X, 2009. Mining individual life pattern based on location history. International Conference on Mobile Data Management: Systems, Services and Middleware 10, 1–10. 10.1109/MDM.2009.11. [DOI] [Google Scholar]
  106. Yeakley JA, Cale WG, 1991. Organizational levels analysis: A key to understanding processes in natural systems. J. Theor. Biol 149, 203–216. 10.1016/S0022-5193(05)80277-2. [DOI] [PubMed] [Google Scholar]
  107. Zaidel DM, 1992. A modeling perspective on the culture of driving. Accid. Anal. Prev 24 (6), 585–597. 10.1016/0001-4575(92)90011-7. [DOI] [PubMed] [Google Scholar]
  108. Zeise L, Bois FY, Chiu WA, Hattis D, Rusyn I, Guyton KZ, 2013. Addressing human variability in next-generation human health risk assessments of environmental chemicals. Environ. Health Perspect 121 (1), 23–31. 10.1289/ehp.1205687. [DOI] [PMC free article] [PubMed] [Google Scholar]
  109. Zhang W, Valencia A, Chang NB, 2021. Synergistic integration between machine learning and agent-based modeling: A multidisciplinary review. IEEE Trans. Neural Networks Learn. Syst 10.1109/TNNLS.2021.3106777. [DOI] [PubMed] [Google Scholar]
  110. Zheng X, Zhong T, Liu M, 2009. Modeling crowd evacuation of a building based on seven methodological approaches. Build. Environ 44 (3), 437–445. 10.1016/J.BUILDENV.2008.04.002. [DOI] [Google Scholar]
  111. Zhu W, Nedovic-Budic Z, Olshansky RB, Marti J, Gao Y, Park Y, McAuley E, Chodzko-Zajko W, 2013. Agent-based modeling of physical activity behavior and environmental correlations: An introduction and illustration. J. Phys. Act. Health 10 (3), 309–322. 10.1123/jpah.10.3.309. [DOI] [PubMed] [Google Scholar]
  112. Zota AR, VanNoy BN, 2021. Integrating intersectionality into the exposome paradigm: A novel approach to racial inequities in uterine fibroids. Am. J. Public Health 111 (1), 104–109. 10.2105/AJPH.2020.305979. [DOI] [PMC free article] [PubMed] [Google Scholar]
  113. Züfle A, Wenk C, Pfoser D, Crooks A, Kim JS, Kavak H, Manzoor U, Jin H, 2021. Urban life: A model of people and places. Comput. Math. Organ. Theory 10.1007/s10588-021-09348-7. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Data Availability Statement

No data was used for the research described in the article.

RESOURCES