(missing reference)

Summary

Helmholtz is often credited for explicitly stating that the goal of the mind is to infer what processes in the world give rise to our perceptions through a mechanism he terms “unconscious inference”. It could be that, in performing these unconscious inferences, we simply learn a series of associations between senses and things in the world. This would correspond, I suppose, to a discriminative view of perception. Helmholtz however doesn’t stop with associations, but goes further to assert that the way we perform unconscious inferences is by determining “generic notions” and “laws of nature”. In other words, our mind constructs a generative model of the world:

Our intellect is the faculty of forming general conceptions. It has nothing to do with our sense-perceptions and experiences, unless it is able to form general conceptions or laws. These laws are then objectified and designated as causes. (pg. 34)

Interestingly, this idea is almost exactly the one put forth by Craik about 50 years later:

My hypothesis then is that thought models, or parallels, reality—that its essential feature is not ‘the mind’, ‘the self’, ‘sense-data’, nor propositions but symbolism, and that this symbolism is largely of the same kind as that which is familiar to us in mechanical devices which aid thought and calculation (pg. 57)

I think the major difference between these two views is, perhaps, that Craik seems more committed to the idea that our mental models are exceedingly accurate, while Helmholtz doesn’t seem to care about their accuracy – they will be as accurate as the regularities in our perceptions lead us to believe. That is, if there is ambiguity in our perceptions, then our model of reality will be similarly ambiguous, lacking in detail, or inaccurate.

Helmholtz also discusses the relevance of the experiment in our ability to understand the world, foreshadowing approaches in active learning and optimal experiment design:

We are not simply passive to the impressions that are urged on us, but we observe, that is, we adjust our organs in those conditions that enable them to distinguish the impressions most accurately. (pg. 14)

He also predicts the idea of thought as being symbolic:

Our ideas of things cannot be anything but symbols, natural signs for things which we learn how to use in order to regulate our movements and actions. Having learned correctly how to read those symbols, we are enabled by their help to adjust our actions so as to bring about the desired result; that is, so that the expected new sensations will arise. (pg. 19)

To sum up, Helmholtz discusses a lot of the ideas that are really core to the way that we think about things today in cognitive science, and also to probabilistic, generative models of cognition.