Models for real-life

As DOND around the world is now in its 5th year, the principles of the game (at least as it is played in Australia) have been the subject of a number of university papers. This is because there is a belief that DOND models key elements of social/business/economic interactions, and could be used to predict certain outcomes.

There have, for many years, been computer scientists and mathematicians who specialise in game theory. These people are most often employed (where they ARE employed) in decision-making groups within large organisations. Typically game theory assists senior executives determine what their clients/competitors might do if they take certain courses of action.

What makes DOND interesting from this point of view, is the reactions of the players to the decisions they have made, the level of risk that they are prepared to take on as events unfold. Of course individuals can behave very differently, but over many shows extremes can be filtered out, allowing strong rules of thumb to be derived.

For example, a casino chain starting out might conclude that where they have men between the ages of 30 and 50 entering their premises for the first time they can be become 'hooked' by allowing them to win an amount of money which is just short of their leave-immediately level, and then stop them from winning again. Why, because in DOND we often see men in this age group chase a previous offer from the banker, and lose everything by following that strategy. On the other hand women from this age group appear to act differently, tending to shutdown if they suffer a major setback, as if they are following a strong damage limitation instinct.

You may be surprised at just how prevalent predictive models are in your daily lives, from scheduling your bus and train timetables, to the way you use soap powder over a week, or even use electricity. About ten years ago a program was developed to determine the location and dimensions of fire doors for a given building. This used a large population of virtual people who had each been given a small set of criteria and properties e.g. virtual person A is 6ft tall, 180lbs and needs to get out quickly but cannot be more than two metres from virtual people B and C, while B and C might be programmed to fight the fire. It was found that the resulting program could be used to predict behaviour on the stock market, and I believe it is still used to predict the rise and fall of certain shares.

Some years ago I worked with genetic algorithms, an artifical intelligence technique to solve extremely complex problems, the model was life itself - if one considered the world to be the problem environment, one might conclude that the human race has evolved to be the most successful solution to that problem. In the same way, we created a model of the problem environment under consideration - a production plant - and created 'chromosomes' which were then evolved over many generations until their 'success' reached a peak. The values of the genes in the most successful chromosome were then translated to their assocaited variables within the factory, making it considerably more efficient.

It is the human element which makes prediction difficult in any model, and I think DOND is a game show which explores the human element really well.