Human behaviour and decision making

This week we will be using a tool called Agent Based Modelling (ABM). In this step we will introduce you to how it operates.

ABM is a computational tool allowing for the modelling of many interacting individuals, called agents. To use it we need:

1. agents – this basically can be any individual: be it an ant, a tree, or a human
2. a set of rules and characteristics of the agents – these can tell us what agents are like (e.g. are they hungry, what they believe in) and what they do (e.g. how they move, how they make decisions)
3. a set of rules and space/network for interactions -. Many of these agents together can represent an ant-nest, a forest or a community/society. they can communicate, influence or …. eat each other.

The unique possibility ABMs offer is that we can equip the agents with rules for interaction. This means that the simulated ants, trees or people can react to the behaviour of other simulated ants, trees or people. Due to such interactions between individuals (micro-level), group phenomena (macro level) may grow (emerge). This opens new possibilities for studying group dynamics and processes of self-organisation.

As an example, the standard ant model in netlogo shows how ants, by just leaving a pheromone trail when they bring food to the nest, are capable of displaying collective intelligent behaviour by first harvesting the food source closests to their nest. Whereas the simulated ants are not aware of their environment, as a collective they self-organise in such a way that the nearest food source is harvested first.

Human behaviour and agent based modelling

The behaviour of humans is targeted by many ABM’s. For example, innovative behaviour and practices may spread, and opinions may polarise between subgroups.These group behaviours set the boundaries for individual behaviour. For example, if all agents believe in X, an agent believing in Y may experience a normative pressure to start believing in Y as well

Agent Based Modelling has been one of the most interesting methodological developments of the last decades for the behavioural sciences. Starting from computer science, artificial intelligence, and complexity science, it has matured to deal with genuinely social science aspects. Conducting experiments on artificial populations, social simulation provides a computational methodology to systematically explore how collective behaviour arises from interactions between many individuals (emergence), and how in turn the behaviour of a collective influences individual behaviour (downward causation). The seminal work of Nobel laureate Thomas Schelling (1971) for example demonstrated that in an ethnically mixed society, where the individuals prefer merely not to be in a minority in their local neighbourhood, a completely segregated society emerges as a result. Schelling’s social computational model was one of the first where individual decision making, in this case the choice to stay or to move away from a location, was made dependent on the behaviour of the other individuals, in this case neighbours staying or moving away.

More information and an exercise with the Schelling model of segregation can be found in the course on People, Networks and Neighbours: understanding social dynamics.