A model is a simplified representation of some aspect of the world, which can be more easily analyzed than the world itself. When assessing and assimilating evidence, it will often be necessary to construct models.

For instance, economists who are interested in predicting how people will behave often analyze models in which everyone acts perfectly in accordance with decision theory. Although the existence of cognitive biases means that no real people actually act this way, economists are able to make mathematically precise conclusions about these models in a way they could not about the actual world. They then argue that their models are similar enough to reality to be a useful way to structure thinking about the relevant aspect of the world.

See game theory and 80,000 Hours’ three-factor framework for two examples of models being used within the context of effective altruism.

Perhaps the main difficulty when building a model is dealing with uncertainty. First, in cases where there is substantial certainty about the relevant aspect of the world, it may be necessary to incorporate uncertainty into the model. Second, there may be [uncertainty about whether the model is appropriate, or whether it diverges too substantially from reality to be instructive.