Flowchart starting with the Underlying Theories of Decision Making

The DYMATICA computational modeling structure simulates interactions among governmental organizations, groups, and societies within selected country(ies) of interest (COI). The data used to condition the model can originate from a large spectrum of sources including previous studies, subject matter expert (SME) guidance, reports, surveys, observations, and public media.

The DYMATICA structure and process is based on a specific combination of well-established psychological, social, and economic theories of decision making, as well as established techniques in knowledge elicitation, statistics, system dynamics modeling, uncertainty quantification, and sensitivity analysis.

Focus of DYMATICA

While DYMATICA cannot point-predict the timing of specific state/non-state behaviors, it can indicate the range of likely behaviors across time and how specific courses of action (COA) can modify or strengthen the overall shape of behaviors.

For example, the figure below illustrates how a potential intervention can change the shape of an adversary’s behaviors across time so that it is more favorable to US and ally interests.

Model9.png

The Scientific Rigor Behind DYMATICA

DYMATICA uses a scientific and systems engineering approach to the modeling and analysis of geopolitical assessments.

Various validation techniques are used to determine and improve model accuracy, the efficacy of data inputs on model response, and intervention points with the greatest effect(s) on system results.

Model11.png

This process involves the identification and quantification of uncertainties associated with data/information pertaining to model parameters, such as SME opinion, and open source data.

  • Statistical methods are used to characterize data uncertainties. The formal analysis of the simulation results characterizes model confidence and robustness for what-if queries regarding various policies and actions.
  • Model confidence management procedures typically incorporate uncertainty quantification (UQ) and sensitivity analysis (SA), as well as other validation techniques. This involves ongoing, collaborative assessment to ensure that the final product provides useful information for the desired application. Uncertainty quantification is also used to learn how uncertainty in inputs ultimately propagates through the model to affect results. By simultaneously performing UQ for model parameters and potential interventions, the framework is able to determine the portfolio of interventions within a range of probabilities of success despite uncertainty. The risk associated with the intervention is also quantified.

Major Theories Underlying DYMATICA

In order to address why a country or group of interest is doing what they are doing or how they would respond to a US or allied COA or policy, theories on how humans actually make decisions should be incorporated into any assessment.

DYMATICA is a theory-based computational model that incorporates empirically based, well established theories on how individuals make decisions across multiple domains of interest.

Model6.png

The DYMATICA Modeling Process

The DYMATICA structure consists of a modeling framework, model simulators, and an analysis approach. The current structure allows for assessment of models across different domains (i.e., different countries, groups, individuals, and scenarios of interest).

Each simulated behavior is a function of psychological characteristics along with environmental and group dynamic factors. This enables the assessment of group behaviors as the groups react to other’s perceptions and world conditions.

For example, the figure below shows a simplified conceptual representation of a hypothetical DYMATICA structure that involves the modeling of two interacting groups and several leaders. Exogenous inputs to the model (e.g., global economic factors and general population support) influence the dynamic interactions within and between the entities.

Model10.png

The figure below shows a real-world example of a country of interest and organizations that would be modeled at a systems level. Here exogenous, rest-of-the-world variables along with inter- and intra-country/organization dynamic variables influence the modeled assessment outputs. Each node circle represents the decision making of a specific modeled entity.

Model12.png

When decisions associated with specific nodes significantly perturb the system in question these nodes are represented within a cognitive structure. This structure represents how decisions are actually made.

The modeled entities process and respond to the system-level information based upon empirically supported theories of decision-making. For example, these entities incorporate psychological and sociological elements within their structure, such as environmental cues, perceptions, motivations, intentions, and behaviors (see Figure below).

The modeled system-level information and behavioral outputs of one entity will serve as cues to other entities, which can, in turn, process that information to produce additional behaviors across time. Thus, across time steps, the modeled entities can respond to actions of other modeled entities as well as to co-occurring environmental cues. Modeled human behavior is determined by local perceptions of world conditions, contained in a feedback process that link behaviors of others, conditions, and events.

Model13.png

The Computational Framework of DYMATICA

To develop a system that can analyze selected individuals, organizations, groups, and countries, DYMATICA uses a hybrid cognitive agent based, system dynamics modeling approach.

  • A system dynamics framework represents interactions, and incorporates both endogenous and exogenous system components
  • Cognitive models represent decision making of individuals within societies, which are embedded within a system dynamics framework
  • Uses decision theories, data, and SME input to construct/parameterize equations using robust statistical regression methods
  • Theories and structure are expressed using differential equations
  • Uses real observations or SME input to construct/parameterize equations using robust statistical regression methods

Agent-based modeling, cognitive modeling, and system dynamics modeling

The general mathematical structure that is instantiated within DYMATICA is shown below. Looking at figure from left to right shows how DYMATICA models how humans generally perceive their environment, process that information, and ultimately act upon this information.

The inputs of the cognitive structure are represented as cues that are perceived from the environment. One collection of cues might form a specific perception of a current situation, whereas another collection of cues might form another perception.

Each environmental cue will have a different weight that provides evidence that a specific perception is active. For some perceptions to be active, multiple cues are needed to be present in order to meet a set threshold. However, for other perceptions, only a few or even one cue might need to be present. It depends on the saliency of each cue and its evidence associated with that perception.

For more information

Back to DYMATICA landing page