Modeling Dispersal Processes

In order to use agent-based simulation for understanding the reason why hominins migrated from Africa to Eurasia and to reconstruct the actual route taken by them, a sophisticated agent model needs to be created.

Assuming that hominins had a concrete reason to leave their original habitat, a detailed consideration of potential influencing factors seems necessary. Utility-based agents, in contrast to simpler agent structures, consider changes of their environment and evaluate the consequences of their actions in advance. Furthermore, the happi­ness regarding new states created by performing an action will be considered as well. Transfered to the challenges hominins faced when crossing Africa towards Eurasia, this happiness might be equated with the sufficient availability of food and other

Fig. 2.7 Sequence of a BDI agent’s reasoning process (Timm 2004)

resources of vital importance. Alternatively, the level of happiness was likely being influenced by the absence of predators or other hostile creatures. With other words, we can assume that the path of hominins was influenced by the pursuit of multiple goals.

When transferring a multi-goal scenario into agent-based modeling the so called BDI architecture for software agents has proven to be suitable. In the belief-desire- intention software model, as shown in Fig. 2.7, a distinction is made between the selection and the execution of the agent’s plan. According to the model of human practical reasoning developed by Bratman, the BDI architecture consists of three components: (Rao et al.

• Beliefs: The agent’s beliefs represent its perception. All information collected about the environment and other agents located inside this environment as well as information about the agent itself are stored in a beliefset.

• Desires: The motivational state of an agent is stored as its desires. Desires may be perceived as the objectives agents want to accomplish. However, the term goal is not equivalent to the term desires when talking about BDI agents. A goal is a specific desire which is actively pursued by an agent. Hence, the desires of an agent may be contradictory whereas the goals of an agent must not be.

• Intentions: The deliberative component of a BDI agent is represented by its inten­tions. In order to attain its goal, agents are provided a set of plans. The plans purposefully chosen by the agent are referred to as the agent’s intentions. A plan consists of a sequence of actions which can be performed by the agent, yet, plans may be composed of other plans.

Transfered to the modulation and simulation of dispersal processes, hominins may be modeled using a BDI architecture. As hominins were able to perceive their environment and had to choose certain actions according to their needs, this way of modeling seems suitable. Hence, the recurrent reasoning process of BDI agents (see Fig. 2.7) needs to be fit to the behavior of hominins.

Initially, the beliefs already known to the agent need to be revised using recent sensor input. New perceptions need to be added or changes in the environment need to be considered for further planning. Based on the updated beliefset, agents need to consider whether the execution of the current plan is still expedient or even possi­ble. Changes in the environment or new observations may cause the current goal to become unattainable pursuing the current plan or the selection of another goal might have become more reasonable. For instance the observation of a group of carnivores close to a nearby river crossing might require a tribe of hominins to discard their current goal of crossing a river course at a certain place.

Generating the agents’ new options by considering the updated beliefset, desires and feasible plans, a new list of goals is being created as well. Based on this goal list a new plan is chosen and the actions which are necessary to accomplish this plan are executed. Executing BDI agents requires this process to be performed over and over again in order to consider changes of the world.

However, hominins are not the only actors which are part of the Out-of-Africa- Hypothesis that deliberate their behavior in regard to their actions. The behavior of carnivores might for example be modeled by using a BDI architecture as well. Choos­ing appropriate prey as well as selecting, defending and marking their territory are processes which can be modeled using intelligent software agents. Yet, specialized domain models is a need to be provided by domain experts.

1.5