Hybrid conceptual mapping of relationships involving system causal loop diagram linked with ABM. Output of the problem conceptualization phase of the modelling process prior to developing a computational hybrid model in AnyLogic. Includes Nate Osgood's O PARTIES extension of Ross Hammond's PARTE
To start the modelling process we need to be clear about the purpose of the model. To provide focus we must ask a precise focus question ("a carefully placed stake in the ground" as Barry RIchmond described it)
We need to think about the key outcomes and interventions we are interested in, given the purpose of the model. For each view there may be different time frames of interest (strategic, tactical, operational) different outcomes valued and different costs of interest to those involved and those affected. Decide on the the key time frame of interest. To keep some focus we need to set priorities about outcomes and interventions. Keep things simple.
Most people start with mapping processes, or flows of entities (often patients) through the resources provided by a service. These processes have cost, quality and outcome measures including efficiency, effectiveness, patient experience and joy in work..
To look more widely at the system in which the process is embedded, we map influences using the causal loop diagrams of system dynamics. These include nodes and links and feedback loops
Here's a nice example of a causal loop diagram combining two chicken stories, in a precise way that even a computer can understand. (from Tom Fiddaman)
Remember in the systems approach we question the boundary of analysis to make sure we include a wider range of causal influences, including our own past decisions and behaviours (the endogenous view)
An agent is a thing or object that acts, and is often a person. An agent also has properties. Acts or methods can change the agents own properties or the properties of others. Many actions are also conditional on the properties of an agent. So properties provide conditional actions according to rules.
Properties of interest can include discrete states. These states can be thought of as agents changing colour as dots on a map The internal interactions among properties and methods can be logically represented as a statechart of states and conditional transitions.(There can be or many statecharts within an agent, each representing a different concern)
Often decisions by individuals according to rules or conditions are the actions of interest. These rules or conditions may include internal or external properties or actions, like information about the environment or messages from other agents
We can also have continuous variables inside an agent, like a level of concern. Just like the systems level these continuous stocks can be involved in feedback loops inside an agent, like emotion regulation.
agents can interact with other agents or the environment.
key interactions in health care involve the patient, the formal carer and the informal carer
agents can be connected in networks via group relationships or place (spatial locations)
In some cases we are interested in agent movement or mobility
Nodes and links can be classified into groups networks relationship types
Rather than just two levels, there can be agents within agents within agents e.g. networks of networks, systems of systems, with nested time and spatial scales e.g. for strategic, tactical and operational decisions. For example such a structure may be required to link the following questions:
How can EDs have better days, hospitals have better months or district health services have better years?
link with design, agile or embedded modelling
Now construct a hybrid concept map of your own problem situation or design project