causal models
Internal representations of cause-and-effect structure that allow predicting the results of interventions.
As defined in 308 | Alison Gopnik on Children, AI, and Modes of Thinking
Where it comes up · 1
Most important claims · 3 of 3Search all claims for “causal models” →
- Child development should be understood as analogous to theory formation in science: children construct intuitive theories or causal models of the world, gather new data, and revise their theories in response to that data, and from a computational perspective this is simply a very good way of acquiring information about the world.308 | Alison Gopnik on Children, AI, and Modes of Thinking
- Following the philosophy of science from Hume onward, causal inference cannot be reliably extracted from patterns of correlation alone; what distinguishes a genuine causal claim is the ability to intervene—to actually act in the world and observe the outcome—and children appear to do exactly this kind of interventionist causal learning.308 | Alison Gopnik on Children, AI, and Modes of Thinking
- Large models require enormous amounts of training data yet generalize poorly, especially to out-of-distribution cases not present in their training, whereas children with vastly less data generalize very well; LLMs notably lack the capacities to do real-world experiments and build abstract causal models that support generalization.308 | Alison Gopnik on Children, AI, and Modes of Thinking