This page is part of an in-progress commentary by Roger
Frye
on High Altitude
Thinking: The International Informatics Summit.
3 features of complexity science: 1st a system of heterogeneous objects, 2nd interacting, 3rd emergent phenomena. We are more interested in the emergent behavior than the objects themselves. E.g. today's Nasdaq index. E.g. the traffic rather than the drivers.
Favorite example of a complex adaptive system -- football. Space=playing field, agents=players, info=behavior of nearby players, interaction=blocking, tackling, kicking, etc. Emergent properties = the points scored. All you need to know about your bet. But you can not understand how without digging down into how the game was played. Trying to decide how to bet on superbowl, he played a few thousand games on his computer. See book Would Be Worlds for how it turned out.
Fingerprints of complexity: 1st medium sized number of agents, 2nd intelligent and adaptive, 3rd local information. Example the 3 body problem. Know initial positions and velocities of 3 bodies under gravitation. Will they collide or will a ball escape from the system? No closed solution, but can solve 2 body. Therefore medium-sized = 3in this space.
Liter of gas. Since Boltzmann, we can statistically summarize. Here medium-sized is less than 10^23. Actually, medium-sized usually works out to be a few hundred or more, evn though football world had 22 players.
By intelligent, he means rule following. By local info, he means no global information, just your neighbors.
Examples of complex systems. Stock markets, road traffic networks, evolutionary ecosystems, supermarkets (example of project for Sainsbury's), national economies, health care delivery systems, communications networks, insurance industry (Insurance World simulator). Graph of a reinsurer winning the game, another surviving, others losing.
The mathematics of complex adaptive systems does not exist. Until Fermat and Pascal came along, there was no mathematical machinery for gambling. Now we look it up in a probability book. We are still waiting for the Fermat or Pascal of complexity to appear.
What is data mining. Mining wisdom out of a vast sea of numbers. Discern the underlying patterns or rules in complex systems. 2,4,6,? We thing 8, but you could choose any number and construct a rule for which it would be true. Psychologists sometimes give a random sequence. People see patterns nevertheless. Tools: Neural nets, self-organizing systems, genetic algorithms, ant algorithms.
Emergent patterns from very simple local rules. A school of fish. 3 rules. minimum distance, direction of neighbors, steer toward average position. www.kewlschool.com demo. Introduce red sharks to break apart the school, so we can see how they tend to reform the school.
QUESTION could look at intent of the fish, what is the non-inertial frame as in physics. Do we need to know the intent to find the frame. Answer: Talk off-line.
Question do you see further advancement of the science and convergence. Answer This is his favorite question, the can you trust it question. The answer is sort-of. He lost his first superbowl simulation, but got better the next year. The art of modeling. Very counter-intuitive behavior. Need deeper understanding which can only come from experimenting.
Question tying non-inertial frames to insurance world. Hardest problem is to find that initial reference frame of the consumer. Frames are changing over time and adapting. ANSWER the objects in the system have a goal. The actions they take are directed toward that goal. and they have to evaluate how they are doing. Decide whether the rules they are using are helping move toward the goal. That is the adaptive part. Rules for changing rules. Mathematics and computing are not magic. They transform implicit information.
END