Value of wrong models

Jump to navigation Jump to search

This article by Dr Piotr Wozniak is part of SuperMemo Guru series on memory, learning, creativity, and problem solving.

Wrong models contribute value to science. The geocentric model was religiously inspired, but it provided valuable reference needed for the falsification of its central premise. In the creative evolutionary process of modelling, all models considered wrong will naturally be deemed inferior, and costly in terms of human wasted brain processing power. This changes on a dime during a paradigm shift that is only possible with a never-ending supply of new models, which should always be considered new value, as all new species, memes, fashions, philosophies, or cultures that are subject to an evolutionary process (see: Value of diversity). The emergence, competition, and death of models is an essential part of human swarm intelligence that underlies the survivability of mankind.

In the never-ending process of generalization, a brain will build a model that is largely consistent with the rest of that particular brain's knowledge. This is why it is healthy to devote a brain to a model, while having many models in many brains. This is also why the confirmation bias is of tremendous value in building models (see: How brains protect wrong models).

Novelist Ian McEwan put it best:

There are ways of being wrong that help others to be right

My favorite example of a wrong model is: Memory overload hypothesis of Alzheimer's. It is a fantastic inspiration in the quest for the hygiene of learning.

Wrong models can serve as inspiration, even if they never lead to a paradigm shift

Naturally, the conceptual value of a wrong model is of little consolation when a wrong model is translated into immeasurable harm to humanity. A wrong model can easily be taken as the basis of action that may cost lives of millions. For that reason, we should always pay attention to the probabilistic estimates of validity that stand behind any model. All models can be used in theoretical reasoning. At worst, the conclusion will be as feeble in probabilistic terms as the underlying premises. However, when investing billions or risking human lives, we need to apply much stricter probabilistic standards.

When employing models in practice, we can factor in their probabilistic validity by computing the expected cost of being wrong

See also: