User:ZuluPapa5/Causal Learning
Causual learning is the process by which individuals react to causes and effects, and learn to understand them. In philosophy and computer science, there are two principle modeling classes, i) associative models and ii) normative models which are evaluated on statistical rigor and mathematical elegance. Descriptive and expectational approaches are predominate in psychological theories, relying on explanatory power and empirical adequacy.
Associative models
[edit]This view claims that casual judgments are associated with trial and error empirical clues such as: i) regular succession; ii)temporal contiguity; and iii) spatial contiguity. Associations are strengthened in continuous events and weakened as event become independent. The process order of the clue's evolution is significant. These are computationally simple models; however past event history (i.e episodic memory) is generally not included. Associative models are appropriate to overlapping domains, but not identical domains. It's assumed that they all converge to a single domain at some level.
Normative models
[edit]These models include contiguity and conditionality, what happens when the cause is Counterfactually removed. Implying, the cause makes the difference. The relationship may be probabilistic rather then deterministic. The cause and effect correlation is an empirical clue to causality, which does not directly imply casual knowledge. These models support a claim that individuals compute contingency tables showing four events in a binary classification cause and effect scenario. The four events are represented with the frequency for each conjunction.
| Normative Relationship | Effect | No effect |
|---|---|---|
| Cause | A | B |
| No Cause | C | D |
These models imply the subject stores information about their experiences and performs normative calculations at the point of decision making. Thus, the domain and temporal time-frame over which the models represent is considerably important.
Intervention
[edit]Intervention assumes that a cause can be manipulated to achieve their effects, whereas empirical observation by itself would have no intervention. When an experimental intervention is introduced in the cause and effect relationship, the causation question becomes iterative, to establish how the intervention affected the original cause. The difference between observational and interventions approaches roughly corresponds to backtracking an non-backtracking counter-factuals.
See also
[edit]Necessary and sufficient condition
Bibliography
[edit]- Causal learning edited by David R. Shanks, Keith James Holyoak, Douglas L. Medin [1]
- Causal learning: psychology, philosophy, and computation By Alison Gopnik, Laura Schulz [2]
- Interdisciplinary Perspectives on Causation By Michael May [3]
- Causality: models, reasoning, and inference By Judea Pearl [4]