I returned from California to three packages on my desk:

If you want the philosophy, read Dennett; if you want the economic application, read Beinhocker; if you want to analyze your problems and make decisions, get the Palisade products.

I spent a peaceful Saturday in the sun listening to Simon Boccanegra and reading with delight the books that come in the Palisade package. Here is my summary of them; more learned summaries and discussion are readily available on the web.

@RISK enables you to develop a model by defining your problem in an Excel worksheet, identifying uncertainty by specifying variables and their possible values along with probability distributions, analyzing the model via simulations to determine the range and probabilities of possible outcomes, and finally making a decision based on the model results and your own preferences.

PrecisionTree enables you to compile and analyze problems using influence diagrams and decision trees-you can apply probability distributions to any uncertain values in the decision tree, let Monte Carlo run, and get the range of possible results-hence you can make your decisions.

RISKOptimizer uses genetic algorithm-based optimization and Monte Carlo simulation to find optimal solutions to problems that are "unresolvable" by standard linear and non-linear optimizers. This is the tool if you seek to use the insights of Darwin to simulate your real-world problems and hence make decisions.

Many years ago, I participated in day-long workshops to generate probability distribution curves of the costs to remediate a number of sites containing radio-active waste. The remediation costs were potentially (and in practice) very high and so our effort was justified. We defined work activities, likely quantities and possible ranges of quantities, likely unit rates and possible ranges of unit rates, and we passed this information to a kindly and wise gentleman from Kansas City who with his son had written a computer code that did Monte Carlo simulations with our estimates and produced probability distribution curves of costs. Then we sat with the client and helped formulate strategy. The most interesting part was arrival of the lawyers and their integration of our findings with their take on the law. I wonder if our recommendations would have differed if we had had @RISK. Maybe, maybe not; but certainly we could have done it faster and with greater facility and maybe even precision.

Even now I am involved in helping a client construct Decision Trees so they may decide whether to remediate a site by in situ treatment, or pump-and-treat, or mass excavation. I will recommend PrecisionTree to them. The probability is that I will not be permitted to record the outcome of this recommendation or, hopefully, its implementation.

Let us turn then to what I have not done, namely use evolutionary theory in decision making. By way of background, here is how Palisade summarizes the relevant theory - see pp. 153-154 of the RISK Optimizer Users Guide, which comes with the DecisionTools Suite Industrial Edition.(I prefer the description by Dawkins in "The Selfish Gene"; a more recent description is to be found in "Breaking the Spell", once again by Dennett):

  1. "Evolution takes place at the level of the chromosome. The organism doesn't evolve, but only serves as the vessel in which the genes are carried and passed along. It is the chromosomes which are dynamically changing with each rearrangement of genes.
  2. Nature tends to make more copies of chromosomes [than can survive] which produce a more "fit" organism. If an organism survives long enough, and is healthy, its genes are more likely to be passed along to a new generation of organism through reproduction. This principle is often referred to as "survival of the fittest". Remember that "fittest" is a relative term; and organism only needs to be fit in comparison to others in the current population to be "successful".
  3. Diversity must be maintained in the population. Seemingly random mutations occur frequently in nature that ensure variation in the organism. These genetic mutations often result in a useful, or even vital feature for a species' survival. With a wider spectrum of possible combinations, a population is less susceptible to a common weakness that could destroy them all (virus, etc.) or other problems associated with inbreeding."

Palisade tells us "the popularity of genetic algorithm is now growing exponentially … The International Conference of Genetic Algorithms is already focusing on practical applications, a sign of maturity that eludes other 'artificial intelligence' technologies."

Least I leave you with the idea that this is a biological issue only, here are some situationsoptimized using RISKOptimizer:

  1. A senior executive wants to find the most effective way to distribute funds among the various departments of the company to maximize profit.
  2. ZooCo is thinking of marketing a new drug used to make hippos healthier; they need to select the capacity level for a new plant that maximizes profits.
  3. A university must assign 25 different classes to 6 pre-defined time blocks. Since the schedule must be developed prior to student registration, the actual number of students per class is uncertain.
  4. It is June 8, 2000. GlassCo needs to purchase 500,000 gallons of heating oil on November 8, 2000. The current spot price of oil is 42 cents per gallon. GlassCo is hedging the price risk inherent in future oil purchase by buying oil futures that expire on December 8, 2000. How many futures should they buy?
  5. A broker has a list of 80 securities of different types that will be worth a different and uncertain amount of money in the future. The broker wants to group these securities into five packages (portfolios) that will be as close to each other in total value as possible one year form now.
  6. A salesman is required to visit every city in the assigned territory once. What is the route with the shortest travel time possible that visits every city?
  7. And finally one that has aggravated each and every one of us: what is the optimal number of reservations to accept in excess of the number of available seats-the classic overbooking problem.

For the mining engineer, these are the classic problems you may solve with Palisade's suite of codes: Will minerals be found? If a deposit is found, will it be economical, or a bonanza? Will the costs of developing the deposit be as forecast? Will some political event like an embargo, tax reform, or new environmental regulations drastically alter the economic viability of the project? If you do solve any of these problems using these codes, let both Palisade and me know and we will pass along the success story.

Maybe I am lucky: I have no acknowledged problem of sufficient complexity to warrant use of RISKOptimizer. Except perhaps the distribution of a two million dollar estate in an impending divorce-yes maybe I will try the code to find an optimum strategy. In the meantime I can only revel in the beauty of the theory, the code, and the potential. But before you take off, here are some problems culled from Wikipedia that may affect the answer:

  1. In complex problems, genetic algorithms (GAs) may converge towards local optima rather than the global optimum. The likelihood of this occurring depends on the shape of the fitness landscape. The Palisade technical folk note that they have always found that GAs do better than traditional linear solvers at finding global solutions rather than local optima.
  2. Operating on dynamic data sets is difficult, as genomes begin to converge early on towards solutions which may no longer be valid for later data. Recent research has also shown the benefits of using biological exaptation (or preadaptation) to solve this problem.
  3. GAs cannot effectively solve problems in which the only fitness measure is right/wrong, as there is no way to converge on the solution. (No hill to climb). In these cases, a random search may find a solution as quickly as a GA.
  4. Selection is clearly an important genetic operator, but opinion is divided over the importance of crossover versus mutation. Some argue that crossover is the most important, while mutation is only necessary to ensure that potential solutions are not lost. Others argue that crossover in a largely uniform population only serves to propagate innovations originally found by mutation, and in a non-uniform population crossover is nearly always equivalent to a very large mutation (which is likely to be catastrophic).
  5. For specific optimization problems and problem instantiations, simpler optimization algorithms may find better solutions than genetic algorithms (given the same amount of computation time). Alternative and complementary algorithms include simulated annealing, hill climbing, and particle swarm optimization.

And so it goes!