Chapter+3&4


 * Chapter 3**

2. The following source was used to complete this exercise: []

“Post Publication” ODD Protocol: The model used in this paper (DomWorld) was designed to provide an explanation for systematic differences in social organization that has been observed in closely related primate species. The authors of the paper specifically use the model to analyze its strength when making generalizations about the genus Macaca.
 * 1) Purpose

The environment for this model is described as being a square at first glance, with wrapping enabled for the top and bottom edges, as well as the left and right sides. In NetLogo, enabling this kind of environment is very easy to implement.
 * 1) Entities, state variables and scales

Populating this environment are 8 agents (turtles in NetLogo), 4 male and 4 female. In addition to a sex for each agent/ turtle, there must also be variables regulating the agent’s visual range. This visual range is critical because the modeled environment is reportedly large enough scale that individuals can stray from the larger social group and become lost (that is, too far out of visual range of other individuals to find their way back to the group.) An additional state variable that is similar to the visual range of each agent is the personal space variable. This is the minimum distance two agents can have between one another without initiating a dominance interaction.


 * 1) Process overview and scheduling

In the DomWorld model, there must be processes that govern the action of each agent/ turtle. These actions are reported to include movement, waiting, and performing dominance interactions. However, more “basic” functions such as eating, dying or reproducing are not modeled, so there is no need for processes governing those behaviors.

Unfortunately, the authors of this article do not go into detail about what goes into these processes. This leaves each reader of this paper a lot of freedom in which to design their own processes if they were to design their own model using this report as a base of understanding.

Likewise, other processes that would have to be developed to run during the setup command (if NetLogo were used to create this model). These setup variables would change for each run of the experiment, meaning they cannot be considered State Variables, and therefore require processes to determine these variables before a run or go command is given to initiate the experiment. Processes that fall under this category would include: the setup of a Dominance value (determining the each agent’s dominance rank within the social group of individuals. This process must take sex of the individual into account, but also must be able to change as the result of dominance interactions), determining when and how long to remain in a waiting period that is meant to simulate foraging, eating, resting and grooming in the wild, and the setup of a //StepDom// or aggression level for each individual that determines how the individual will respond during dominance interactions (aggressively or relatively more passive).


 * 1) Design Concepts
 * 2) Basic Principles- The authors of the article describing the DomWorld model include equations that were used to determine how some processes are implemented, particularly the Dominance Interactions. However, there are very few specific examples of other research or sources used when the original programmer developed this model.
 * 3) Emergence- Because the authors of this article did not include how processes were developed, but instead opted to only report on what the processes accomplished, it is very difficult to determine which results of the model are a result of adaptive behavior within the model and which results are simply the result of rules imposed on the model’s agents.
 * 4) Adaption- Agents in DomWorld do have a number of behaviors that are a result of “decisions” made by those agents. A number of these decisions are related to the model’s Dominance Interaction system, which determines the frequency of these interactions based on a number of variables and processes.

However, there are some behaviors that are not the result of adaption, but are instead the result of rules built into the model. The clearest example is the process that states that all individuals will turn until they are within view of another individual as opposed to walking away from the group in random directions.
 * 1) Objectives- actions such as eating and mating are removed from this model, minimizing the amount of objectives an individual agent must account for/ worry about.
 * 2) Learning- Similar to (d.), the agents within this model do not learn.
 * 3) Prediction- This is not implemented in the DomWorld model.
 * 4) Sensing- Variables that can be “sensed” by agents are their proximity to other individuals (other agents), which governs their behavior. If agents are too far apart, they will move closer together, but if the agents are too close together and in one another’s personal space a Dominance Interaction will take place, resulting in the agents spreading out.
 * 5) Interaction- Agents directly interact with one another by sensing their proximity to other individual agents. These direct interactions are meant to model the complex social structure of primates in a way that is (hopefully) easier to understand, providing insight for researches of these primate social groups.
 * 6) Stochasticity- “pseudorandom” numbers are used to determine the base level for each individual’s aggression level, as well as their initial dominance rank within the social group.
 * 7) Collectives- This entire model is built with the idea that individuals will collect and organize themselves in a way that reflects the social interactions of primates in the real world. These social interactions within the model require that agents influence the “amount” of aggregation based on every interaction with other agents.

The collective/ aggregation of agents is brought about by the agents themselves, as determined by the result of multiple social interactions amongst individuals, and is not influence by pre-determined qualities as defined by the model/ program.
 * 1) Observation- Another failure of the authors of the article is their neglect to include what sort of output is created by the model as it is running. This makes it very difficult to determine what sort of hard data is provided to the model users, if any.
 * 2) Initialization- Although not stated specifically, it can be presumed that the aggression level and the initial dominance rank are determined at the setup stage of the model before it is run. How each of the 8 agents is placed within the environment of the model is unspecified, but a process governing their placement would also have to be in place during the setup stage.
 * 3) Input Data- There is no reference to input data for the DomWorld model.
 * 4) Submodels- Submodels could include some of the processes governing movement and social interactions.

3. There are numerous different bird species, many of them coastal and/or marine species, which have very specific nesting arrangements. Many of these species nest on a beach or other suitable location, and have integrated complex social behaviors that determine the spatial distribution of mating and nesting sites within the larger “Mating area”. Modeling these social behaviors and the resulting next distribution would require an ABM that explicitly models individuals.

A model such as this would require both male and female agents. It is likely that relative measures of fitness and other metrics used by the individuals to select a mate can be removed from the model for simplicity. However, the processes and variables that govern the social interaction of these birds are absolutely essential. Variables such as these could include, but are not limited to: Relative measures of dominance, vision ranges, measures of personal space etc. All of these variables must come together to accurately replicate what happens in nature before the model can be implement to understand this behavior.


 * Chapter 4**

Setxy random-pxcor random-pycor
 * 1) As predicted, when setting q to equal 0.0, the movement of the butterflies is 100% random, and they do not follow the protocol that dictates they should move uphill. Conversely, setting q to equal 1.0 means the butterflies move directly uphill as quickly as possible without any random movement at all. These patterns are more easily observed by changing the number of butterflies to 50, as opposed to the original one.
 * 2) Implemented the following to give the butterflies a random distribution as a result of the setup protocols, as opposed to having all butterflies start from the same fixed point:


 * 1) The paths the butterflies take may look generic because of the simplicity of the environment. Once a butterfly finds a patch that makes up one of the two hills, there will always be another patch with an elevation higher than its current position until the butterflies have all congregated at the peak of either hill. As suggested, completing exercise 4 provides the additional insight necessary to answer this question. It should also be noted that if/ when actual landscape data is put into this model, this artificial flight path is likely to be much more difficult to observe due to landscape “noise”.
 * 2) Adding noise to the landscape adds a sense of realism to the flight path of the butterflies. They no longer progress straight towards the hilltop, but instead manage to appear “flighty” and more butterfly-like by backtracking and veering in seemingly random directions before eventually accomplishing their overall goal of reaching the hilltop.