Causality
The search for a causal understanding of the world is at the heart of human cognition. It shapes learning throughout development and guides intelligent behaviour by allowing cognisers to predict outcomes, selectively gather information, attribute blame and credit, and imagine hypothetical and counterfactual situations. Causal reasoning is also central to the scientific method, underpinning how we, as scientists, design experiments, build and evaluate theories including the ones we use to describe and understand our own minds.
In this session, we will think about how to model the cognition involved in learning, representing and exploiting a causal model of the world.
Primary Readings
Everyone should read these and be prepared to discuss.
| Coenen, A., Rehder, B., & Gureckis, T. M. (2015). | Strategies to intervene on causal systems are adaptively selected. Cognitive Psychology, 79, 102-133. |
| Pearl, J. (2019). | The seven tools of causal inference, with reflections on machine learning. Communications of the ACM, 62(3), 54-60. |
Secondary Readings
The presenter should read and incorporate at least one of these.
| Bramley, N. R., Lagnado, D. A., & Speekenbrink, M. (2015). | Conservative forgetful scholars: How people learn causal structure through sequences of interventions. Journal of Experimental Psychology: Learning, Memory and Cognition, 41(3), 708.This article explores how people learn causal structures through interventions. The researchers developed a computer task where participants learned the structure of probabilistic causal systems. They developed models to understand participants’ intervention choices and judgments, considering memory and processing limitations. The study found that successful participants used a model that maximized information gain, forgot evidence from earlier trials, and had conservative beliefs. The study highlights the importance of active learning and the use of simple heuristics in causal structure identification. Like Coenen et al, this is an example of human causal learning through interventions, at rung 2 of Pearl’s ladder. |
| Quillien, T., & Lucas, C. G. (2023). | Counterfactuals and the logic of causal selection. Psychological Review.Quillien and Lucas explore the relationship between counterfactuals and causal selection. They propose a counterfactual theory, suggesting that people imagine alternative possibilities and judge causal responsibility based on the correlation between factors across these simulations. The theory is supported by empirical data and experiments. The article discusses the use of a computational model, the Counterfactual Effect Size Model (CESM), to explain human judgments in various studies related to causal attributions and judgments. The CESM is found to have a good fit with the data and is able to predict participants’ intuitions accurately. |
| Pearl, J. (2000/2009) | Causality: Chapters 1-2. (On Learn)This important book laid formal/mathmatical groundwork for causal modelling in data science but also use of bayesian networks as representations of structural knowledge about the world, particularly causal, combining this with principles of probabilistic inference and statistical dependence. It is technical but readable. Chapter one of the book introduces the concepts of probabilities, graphs, and causal models. It highlights the importance of probabilities in studying causality, especially in disciplines such as economics, epidemiology, sociology, and psychology. Probability theory helps determine the strength of causal connections and make inferences from noisy observations. The chapter also discusses the use of probabilities in handling exceptions that cannot be processed by deterministic logic. The passage goes on to introduce various concepts within probability theory such as axioms, conditional probabilities, Bayesian inference, joint distribution functions, and graphical models. It explains the properties and terminology associated with probabilities, graphs, and causal models. The chapter also explores the application of probabilities and causal models to hypothesis testing, interventions, and counterfactual analysis. Overall, these chapters provide a comprehensive introduction to probabilities, graphs, and causal models, highlighting their applications and implications in studying causality. |
Questions under discussion
- What are the levels of Pearl’s Ladder of Causation?
- What makes an intervention valuable?
- What are counterfactuals and what role do they play in reasoning?
