Presentation on theme: "Forecasting the Future using Computer Simulation Models Presentation by Jon Roland April 18, 2006"— Presentation transcript:
Forecasting the Future using Computer Simulation Models Presentation by Jon Roland April 18, 2006
Methods of Forecasting/Conjecture Vision/Revelation Divining Scenario construction Trend extrapolation Constraint projection Mental models Delphi Mathematical models Physical models Computer simulation models Only the complete Universe has the compute-power to simulate any open subsystem of itself.
Mental Models Can integrate much information Difficult to extract details Irreproducible Distortable by emotion Not good for complex interactions Limited by education Limited by cognitive capacity Slow Resistant to change Susceptible to herd influences Often the best available alternative
Varieties of computer simulation models Discrete or Continuous Linear or Nonlinear Open or Closed Systems Automata Simulation games Statistical best-fit analysis Bondgraph Agent-Based Stock-Flow Evolving Complex Network Hybrid
Feedback Loops +/- InputOutput Effector Damped negative feedback Positive feedback Positive feedback and negative combined Feedback Loop Timing delays can cause oscillation
System Dynamics, Model of U.S. Economy
System Dynamics, Run of Economy Model
World3 Model Standard Run (Scenario 1) (Beyond the Limits by Meadows, etc., Scenario 1, p.133)
Biosphere: Closed Systems for Materials Biospheres can be small and simple if lifeforms are small, few, and with a simple ecology. If we seek to build biospheres for people, we begin by enclosing them inside a membrane that must be made impermeable for materials. The technology available to maintain such impermeability then becomes the key consideration in habitat design. If an infrastructure is assumed to be unable to endure a loss of more than 63% of its stock of materials, and if losses cannot be replenished, it can endure no longer than n years, where 1/n is the annual loss of materials. To endure millions of years, annual losses must be < , which means going underground.
Biospheres, Compact Cities, Sited on Earth Biosphere II, Arizona Kaymakli, Turkey, Ancient Underground City Build Cities Downward Ultimately, only building underground can avoid losses of materials critical to the infrastructure, to the surrounding environment, diluted to the point they cannot be recovered economically.
Paolo Soleri’s Arcology Designs
Space Habitats Asteroid Ida (and Dactyl) Space TorusBernal Space City Earthlike Conditions withinCosmic rays may require more shielding
Space Cities: Star Trek Starbase
Space Cities: Star Trek Deep Space 9
Space Cities: Babylon 5
Prisoners’ Dilemma, 2-Player, 1-Round The average payoff for cooperation is (3 + 0)/2 = 1.5. The average payoff for defection is (5 + 1)/2 = 3. Therefore the rational strategy for each is to defect. But if both play the individually rational strategy, they both come out worse than if they cooperate. There is a conflict between what is rational for each individual and what is rational for them in combination. This game, extended to multiple players and iterations, is fundamental to understanding a society. Scenario: Two separated prisoners are being tortured to get each to betray the other. If one rats out the other, the other is executed, and the betrayer goes free, but if both rat out the other, they both get life in prison. If they both remain loyal, they eventually will be released.
Prisoners’ Dilemma, n-Player, Iterated A number of players n are allowed to encounter one another at random, through multiple rounds m. Each player starts with a stock of points, which are incremented or decremented, depending on the payoffs from each encounter. If the stock of points of a player falls below 0, that player “dies” and is removed from the game. Each player has a memory of the move made by each of the other players it encountered last. Each player has a strategy for whether to cooperate or defect, depending on its memory of past encounters. By giving each player different strategies and playing multiple rounds, the survival of the players provides a measure of the rationality of each strategy. The most successful strategy has been found to be “Tit-for-Tat”, cooperating on the first move with another player one has not previously encountered, and otherwise making the same move made by the other player at the previous encounter. This strategy can result in universal cooperation, internecine vendetta, or a pattern of most cooperating while a few take advantage of that to defect and gain an advantage, like a criminal class, typically about 6% of the whole.
More Results of Iterated Prisoners’ Dilemma Gaming If the players have information about how many moves m will be played, and keep count of which move it is, it can be a rational strategy to defect on the last move. If the players have information that the number of players is so large they are unlikely to encounter the same player again, it can be a rational strategy to defect with players not previously encountered. A “society” of persistent cooperation is most likely to develop if players are initially confined to small groups, then the groups of cooperators combined into larger groups. This resembles human social development, beginning with families, then extending social bonds to larger groups. The kind of computer simulation used is called “agent-based” simulation, because each of the players acts as an independent agent operating in a field of random encounters among them. Minor extensions of the capabilities and modifications of the payoff tables for the players can produce runs that resemble the complex behaviors of members of societies of all kinds. These simulations explain why cooperative behavior and society has conferred survival advantages, but also explains how cooperation can fail and rivalries develop. The emergence of a subgroup of persistent defectors resembles the emergence of a criminal class.
Evolving Complex Networks: Internet Network of nodes and links evolves by links being added at random between pairs of nodes. If Probability of adding a link to a node increases with number of links the node already has, network tends to evolve into a set of hubs with large numbers of links to them. Models explain how “rich get richer”, and challenge market models of economic, social, and biological behavior.