NetLogo banner

Home
Download
Help
Forum
Resources
Extensions
FAQ
NetLogo Publications
Contact Us
Donate

Models:
Library
Community
Modeling Commons

Beginners Interactive NetLogo Dictionary (BIND)
NetLogo Dictionary

User Manuals:
Web
Printable
Chinese
Czech
Farsi / Persian
Japanese
Spanish

  Donate

NetLogo User Community Models

(back to the NetLogo User Community Models)

[screen shot]

Download
If clicking does not initiate a download, try right clicking or control clicking and choosing "Save" or "Download".(The run link is disabled because this model uses extensions.)

WHAT IS IT?

MIMICS purports to simulate a social group that uses two contrasting abstract concepts to communicate about a given situation. Instead of viewing concepts as the result of an inductive process that extracts properties and their statistical structure from the environment, MIMICS views abstract concepts as subjective points of view, and sees communication as an attempt at inferring other peoples' subjective points of view. Furthermore, MIMICS assumes that abstract concepts require explicit teaching to be learned.

HOW IT WORKS

In MIMICS, agents use concepts to interact with other agents. Concepts may be used in learning interactions or in communication interactions. In learning interactions, agents request other agents to teach them the conceptual content of the concept currently in their minds. In communication interactions, agents look for other agents to provide confirmatory evidence of their own conceptual state (this is called "agreement"). Every time another agent teaches them the content of a concept, and every time agreement is reached, the interaction is deemed successful. The more successful interactions, the better a concept becomes differentiated from the alternative concept. The more differentiated concepts are, the more agents believe they need no further learning, and tend to use concepts in communication.

HOW TO USE IT

The only free parameters in MIMICS are the number of agents (N) and the number of potential properties (P). Play around with these parameters. Another setting that can be set is the stopping criterion. A simulation run may be stopped by waiting until the global distribution of properties stabilizes (as described in the manuscript that accompanies this simulator), or by setting the maximum number of steps in the simulation run. Note that setting the number of steps to zero, means that the simulation will run indefinetely and will need to be manually stopped.

General steps to perform a simulation run:

1.- Set the values for "N" and "P" in the corresponding green sliders.
2.- Select the stopping condition: "Global stability" or "Nr. of steps".
3.- If in 2.- you selected "Nr. of steps", set the maximum nr. of steps to end the run in the "max-step" green slider. Note that setting it to 0 will run the simulation forever.
4.- Press the "Setup MIMICS" blue button.
5.- Press the "Run MIMICS" blue button. Pressing it again will pause the simulation and pressing it again will resume the run (i.e., it is a "toggle" button).

THINGS TO NOTICE

There are many interesting things to notice. Before diving into the more substantive issues, it is important to understand what the different windows are showing.

The "property frequency distribution for concepts" window is built by asking all agents, at each simulation step, to inform their conceptual content for each of the concepts C1 and C2. These lists are accumulated across agents to produce a frequency distribution akin to what would be found in a conceptual properties norming study.

The "success probability" window, found immediately below, shows the probability that a property will lead to a successful interaction (as defined above and in the accompanying manuscript) when used in the context of a given concept. Because agents use success probabilities to decide to which concept to assign a property, the two distributions (frequency and success probability) can be mapped onto each other.

Other interesting windows show the probabilities of true and illusory agreement (discussed in detail in the accompanying manuscript). In a nutshell, true agreement (p.a1) is the probability that an agent will correctly infer another agent's point of view (concept, state of mind, etc.) given evidence provided by that other agent (i.e., a communicated property). Similarly, illusory agreement (p.a2) is the probability that an agent will incorrectly infer another agent's point of view given a communicated property.

Finally, it is also necessary to understand the variability windows. The "intra-agent variability" windows show the percentage of agents that change their conceptual content at any given moment in a simulation run. Successful and unsuccessful interactions may both lead properties in or out of a concept. These windows show whether agents' conceptual content (their lists of properties) becomes stabilized or fixed, or instead remain unstable. The "inter-agent variability" window dynamically computes the value of (1 - p.a1). Because p.a1 is an index of homogeneity of conceptual content in the agent group (as discussed in the accompanying manuscript), 1 - p.a1 is an index of inter-agent variability. If this index were to reach zero, it would mean that all agents developed the same set of properties to describe a concept. If it were to reach a value of one, it would mean that all agents developed different sets of properties to describe a concept.

THINGS TO TRY

By using the parameter settings for N and P discussed in the manuscript, inspect the probability distributions generate by the agents history of interactions (notice they are not uniform distributions). Also inspect p.a1 and p.a2, which show that agents are more likely than not to make correct inferences about other agents' points of view. Note that this happens though agents never have access to other agents real state of mind, and thus cannot use this information to calibrate their inferential process. Finally, inspect also the variability windows, and notice that inter-agent and intra-agent variability occur at a wide variety of settings, and that the system may stabilize at a global level (as shown by the success probability distribution), and still be unstable at the individual agent level.

To test all the above behaviors, you may set MIMICS to stop on its own when it has reached global stability. Additionally, check the number of steps needed for reaching stability, and set MIMICS to continue running well beyond that number (e.g., leave it running to infinity by setting the number of steps to zero). Then, verify that variability remains even in the very long run, suggesting that it is an inherent feature of the model.

CREDITS AND REFERENCES

Chaigneau, S.E., Canessa, E. & Barra, C. (submitted). Using concepts to communicate may account for conceptual variability: an agent-based model.

For a copy, please contact sergio.chaigneau@uai.cl

(back to the NetLogo User Community Models)