Evolution of two ways of understanding causation: Humean and Kantian

Aaron Sloman


The current emphasis on causation as correlational/statistical, i.e. Humean, as in Bayesian nets, ignores a deeper notion of causation as structure-based and deterministic, i.e. Kantian. The history of science frequently involved moves from Humean (merely observed) to Kantian (intelligible) causation, and that also seems to happen in young humans and a subset of other animals.

A Kantian grasp of causation in our environment typically requires understanding of spatial structures and relationships and being able to reason about what happens when they change (often with multiple relationships changing concurrently). This uses geometrical and topological reasoning, while taking account of properties of different kinds of matter (e.g. rigid, flexible, brittle, elastic, tearble, etc.). A result is the ability to cope creatively with some novel problems. This is closely related to the development of some human mathematical competences. (The ability to do logical and statistical reasoning requires other forms of representation.)

The differences between ontologies and forms of representations required for Humean and Kantian understanding of causation will be discussed in relation to cognitive development and requirements for manipulation of 3-D structures (e.g. adding twigs to partially built nests, getting at flesh in animal prey, or playing with meccano and other construction kits). There are deep implications for various disciplines, including neuroscience, linguistics, education, philosophy.

Evolutionary considerations explain why varieties of causal competence are related to the altricial-precocial distinction: Kantian causal understanding seems to require characteristics of altricial species -- which acquire successive layers of orthogonal recombinable competences, using evolved mechanisms for learning by using playful, creative, exploration of the environment, together with what has previously been learnt, whereas precocial species have only pre-compiled causal competences. This extends some old ideas of Oliver Selfridge.

Some of the theoretical ideas, empirical data, and challenges for biological researchers, will be discussed in more detail in Jackie Chappell's talk.

Two views of child as scientist: Humean and Kantian
(PDF presentation)

Interactions between Philosophy and Artificial Intelligence: The Role of Intuition and Non-Logical Reasoning in Intelligence
Aaron Sloman, Proc IJCAI 1971, pp 270-278

Altricial self-organising information-processing systems
Proceedings IJCAI'05
Aaron Sloman and Jackie Chappell

Natural and artificial meta-configured altricial information-processing systems
To appear in Int Journ for Unconventional Computing
Jackie Chappell and Aaron Sloman

Computational Cognitive Epigenetics
To appear in Behavioral and Brain Sciences
(Commentary on Jablonka and Lamb: Evolution in four dimension)
Aaron Sloman and Jackie Chappell