Divisão de Observação da Terra e Geoinformática

I hope the book demonstrates that the agent-based generative approach can be explanatory even in […] cases […] where “the equilibrium approach,” […] is either infeasible or is devoid of explanatory significance. (p.xiii)

Every computational model is a computer program. Therefore it is Turing computable and then it can be described by a recursive function (gargantuan in size and imposingly complex, but they exist). Agent models are expressible as equations. Which representation (equations or programs) is most illuminating? Agent model is immediately intelligible as such. (p. xiv)

I use the term “computational” to distinguish artificial societies from various equation-based models in mathematical economics, n-person game theory, and mathematical ecology that (while not computational) can legitimately be called agent-based.

The agent-based model […] is especially powerful in representing spatially distributed systems of heterogeneous autonomous actors with bounded information and computing capacity who interact locally.

Even if one conducts a statistical analysis over some distribution of runs - using different random seeds - each run is itself a deduction. From a technical stand point, generative implies deductive.

[talking about sugarscape] when an initial population of [such] agents is released into an artificial environment in which, an with which, they interact, the resulting artificial society unavoidably links [lots of social areas]. Because the individual is multidimensional, so is the society.

Equations vs. ABM: equations only describe the global behaviour of the model, and it cannot conclude how the model evolves to that situation. it devoid of explanatory power depict (?) its descriptive accuracy.

one more feature of abm:

The author uses Godel's incompleteness theorem to argue that not all equilibrium is computable. Also, following computational complexity, some of them would require too much effort to be solved. These considerations, when combined with powerful psychological evidence, cast severe doubt on the rational choice picture as the most productive idealization of human decision-making, and serve only to enforce the bounded rationality picture insisted by Simon (1982).

Empirical ABM can be seen as induction over a sample of realization, each one of which is a strict deduction, or theorem, and comparison of the generated distribution to statistical data. This differs from inductive survey research where we assemble data and fit it by aggregate regression, for example.

In many quarters, “explaining an observed social pattern” is understood to mean “demonstrating that it is the Nash equilibrium of some game”.

The game: players move in a 1-D space to achieve satisfiability. One moves to the rightmost cell if there is someone lower than him in his right.

Most of the equilibrium points of the presented game are unattainable (as the number of players grows), given any initial configuration. If someone finds out in the real world one configuration which belongs to the equilibrium set,

in classic social science she would argue that the players have followed that game, what is indeed false, because that equilibrium is unreachable.

Implicit claims that equilibrium analysis is explanatory or predictive should be challenged and require the most careful defence. A successful defence of any such claims must include a demonstration of attainability on the scales of interest, by agents employing plausible rules.

Two related theoretical issues:

1. the connection between individual rationality and aggregate efficiency
2. role of social interactions and social networks in individual decision-making and in determining macroscopic outcomes and dynamics.

Perhaps the main issue, then, is not how much rationality there is at the micro level, but how little is enough to generate macro-level patterns in which most agents are behaving “as if” they were rational, and how various social networks affect the dynamics of such patterns.

In 1961, Congress reduced the minimum age at which workers could claim social security benefits form 65 to 62. Collected data shows that it took nearly 3 decades for the downward shift from 65 to 62. If agents were rational, they wouldn't have taken that time. With social network and imitative dynamics, very little individual rationality may be needed for society as a whole ultimately to exhibit optimal behaviour.

Three types of agents:

1. Rational: retire at the earliest possible age allowed
2. Random: retire with probability p each period, once they reach the age
3. Imitator: retire when the majority of its neighbours is retired

The paper studies the contrast between two types of norms:

• discriminatory norms: allocate different shares of the pie according to gender, age, race, etc.
• equity norms: do not so discriminate

How such classes can emerge and persist, given a norm-free, classless world initially.

Evolutionary game theory and ABM. Each person has tags (merely distinguishing features) and play the Nash Demand Game (each player gets his demands if the sum of the two demands is not more than 100. players may demand 30, 50 or 70).

A social norm is a self-perpetuating state in which players' memories, and hence their best replies, are unchanging. The agent plays the best reply according to his memory, or a random choose according to an error probability. There are two regions of the state space - one equitable (50,50), the other fractious (30,70 and 70,30) - that are very persistent: once the process enters such a region, it tends to stay there for a long period of time.

The authors extend the model to best replies according to the tag of the opponent. The model may evolve to different strategies against each class. Various kinds of social orders - segregated, discriminatory, and class systems - can also arise through the decentralized interactions of many agents in which accidents of history become reinforced over time (a path-dependence dynamics).

The question: why rational choice leads to defecting in the PD? why do we need a one-round memory to attain cooperative behaviour?

Demographic games seems an appropriate name for this class of models because they involve spatial, evolutionary, and population dynamics. (p. 203)

Agents move around this space, interact with neighbours, and have offspring. They are not fixed cellular automaton sites, as in Nowak and May 92, Feldmand and Nagel 93, and others). Each agent is an object whose main attributes are vision, wealth, age and strategy. Behaviour: choose a random unoccupied site within your vision; go there and play your strategy against each neighbour. Payoffs accumulate. Since our payoff matrix has negative entries, an agent's accumulated wealth may go negative. In that event the agent “dies”. If an agent's accumulated wealth exceeds some positive threshold and there is an unoccupied site within the agent's vision, the agent has one offspring. (p. 204)

Perhaps it is worth emphasizing that, in adopting this assumption of a fixed agent strategy, we are not claiming that human strategies are literally hard-wired genetically. Rather, for modelling purposes, we are assuming that they are culturally transmitted from parents to children with high fidelity, like certain religious or ethnic affiliations, tastes and native language (on cultural transmission, see Cavalli-Sforza and Feldman 81, Boyd and Richerson 85) (p. 204)

Cooperation can emerge and flourish in a population of tagless agents playing zero-memory fixed strategies of cooperate or defect in this demographic setting.

Variations: maximum age and mutation. some results are similar to predator-prey cycles Cooperators function to separate defectors from one another.

For non-negative payoff, defectors spread while cooperators sharing until extinction.

Cooperation acquires selective advantage as selection pressures increase. Or, more prosaically, necessity is the mother of cooperation.

To me, […] one of the more remarkable aspects of homo sapiens is how much of our social behaviour is entirely thoughless. Individual though is inversely related to the strength of a social norm.

The agent follows a boolean “norm”. She looks to her neighbours in order to update her norm. If increasing the observation radius changes the statistics then she increases the radius. Otherwise, if decreasing the radius does not change the statistics, she decreases it.

If you had asked a 14th Century European if the earth were round or flat, he would have said “flat”. If, today, you ask the average American the same question, you will certainly get a different response: “round”. But I doubt that the typical American could furnish more compelling reasons for his correct belief than our 14th Century counterpart could have provided for his erroneous one. Indeed, on this test, the “modern” person will likely fare worse: at least the 14th Century “norm” accorded with intuition. Maybe we are going backward! There is a model with a re-equilibrium after a noised period for capturing that basic phenomenon.

In the rest of the chapter, the author investigates different levels of noise.