Methodology

13.2.1 ZambeziLand: An Agent-Based Model (ABM)

of Politogenesis by Canonical Processes

The methodology of this study consisted of building and analyzing an organiza­tional agent-based model (ABM) of the region of interest, called ZambeziLand.

The Canonical Theory provides an integrative framework for linking micro­level, short-term political activity by individuals and groups in a given society (fast processes) with macro-level sociopolitical changes experienced over longer periods of time (slow process). All societies experience numerous fast processes, each ini­tiated by situational changes, but they realize a single slow process resulting from iterations of canonically varying fast processes. The main structure of the fast process is universal and invariant, but the exact branching paths realized vary, depending on contingencies such as a situational change having endogenous or exogenous causes, a society perceiving or not the situational change, collective action occurring or not, success or failure in collective action being realized: hence, the term canonical.

An ABM was chosen for implementing the Canonical Theory because one of the hallmarks of such formal, computational models is their ability to generate macro­level behaviors caused by micro-level decisions of individual agents characterized by bounded rationality, decision-making autonomy, sociality, and dynamic interac­tions among them Epstein and Axtell (1996). These are also features assumed in the theory’s canonical fast process of situational changes and societal responses that result in the slow process produced by each simulation run.

In the current model (version 1.0), agents represent individual members of soci­ety. Each individual can join a group, and each has a level or amount of two attributes: fealty and leadership. Fealty in the ZambeziLand model is a measure of how attached or loyal a person feels towards one’s group in general and its leadership in particular. Fealty is a measure of attachment in that if it drops too low for members of a group, they will seek to move to another group with stronger leadership. All members in a group have a leadership score; however, when group decisions or actions need to be made, only the individual with the highest leadership score counts as the group leader.

13.2.2 Model Details

ZambeziLand 1.0 is an ABM consisting of a society comprised of groups of individ­uals. The model is initialized with 100 groups, each with 50 members, so N = 5,000 total population. These features were chosen to represent an egalitarian, undifferen­tiated society as would have existed prior to the origin of social complexity in the region (i.e., hunter-gatherer tribes). At the start of the simulation, each actor-agent is given an initial value for fealty and leadership.

The model was implemented in Python 2.7.1, which allows for setting fealty and leadership adjustments as input parameters. However, to clarify analysis, all runs are reported here with the same leadership adjustment parameter. Runs of the model were made on a Macbook Pro with four processor cores.

The model takes on average 9 s to run. Four minutes and 30 s were required for executing 30 runs.

13.2.3 Model Action

ZambeziLand 1.0 runs as event loops, where each group of agents has an opportu­nity to act on one or more of its behaviors at each clock tick. Each event loop starts with a situational change occurring (e.g., drought, attack, or other societal threat or opportunity) and each group deciding if collective action should be undertaken. The situational change is left as generic in the current model version, but can be made specific in subsequent versions. This implements the causal fast process of the Canonical Theory, which links situational changes, societal awareness, collective action, and political results: “[w]hen a society correctly perceives and understands a given situational change, it may or may not be willing and able to undertake collec- five action...in response to such a change” Cioffi-Revilla (2005, p. 144).

Specifically, a group will undertake collective action if the average fealty score for the group is code walk-through, debugging, profiling, and sensitivity analysis Cioffi-Revilla (2014, pp. 235, 297). Although complete sweeps of the entire para­meter space were not conducted, numerous parameter settings for initial conditions yielded consistent and replicable results across 30 runs for any given set of initial conditions (parameter settings). All issues encountered were resolved until the model ran as intended.

13.3