intelligence: Book reviews

By Judea Pearl. 2000

Subfields: AI Ethics, Fairness & Society

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In AI Ethics, Fairness & Society

Citations in the corpus

A Causal Bayesian Networks Viewpoint on Fairness (2018, neutral): “The back-door criterion can also be derived from the rules of do-calculus [39,40], which indicate whether and how p→A(Y |A) can be estimated using observations from G: for many graph structures with unobserved confounders the only way to compute causal effects is by collecting observations directly ”
Fair Inference on Outcomes (2017, positive): “iven a joint distribution p(Y,X) either modeled fully via a generative model,orpartlyviaadiscriminativemodelp(Y,X\W|W). In addition, we assume that p(Y,X)is induced by a causal model in the sense of (Pearl, 2009; Spirtes et al., 2001), that the presence of discrimination based on some sensitive feat”
Invariant Causal Prediction for Nonlinear Models (2017, neutral): “tion.1 1.2 Causal discovery In causal discovery (also called structure learning) one tries to reconstruct the structural causal model or its graphical representation from its joint distribution (e.g. Pearl, 2009; Spirtes et al., 2000; Peters et al., 2017; Chickering, 2002; Peters et al., 2014; Hecke”
Causal Discovery from Changes (2001, neutral): “Several such algorithms have been developed, including IC al gorithm [Pearl, 2000, section 2.5] (initially intro duced in [Pearl and Verma, 1991]) and PC algo rithm (Spirtes et al., 1993].”
Causal Transfer in Machine Learning (2015, neutral): “17). 1.3 Contribution Taking into account causal knowledge, our approach to DG and MTL assumes that covariate shift holds only for a subset of the features. From the point of view of causal modeling (Pearl, 2009), assuming invariance of conditionals makes sense if the conditionals represent causal m”
Causal Structure Learning (2017, positive): “graphs (SWIGs) (Richardson and Robins, 2013) provide a unied framework for potential outcome and graphical approaches to causality. 2 2 THE MODEL We formulate the model as a structural causal model (Pearl, 2009). In particular, we consider a linear structural equation model (e.g., Wright, 1921) for ”
Learning What Makes a Difference from Counterfactual Examples and Gradient Supervision (2020, neutral): “on beyond its training distribution should ideally capture the causal mechanisms at play behind the data. Acquiring additional training examples from the same distribution cannot help in this process [49]. Rather, we need either to inject strong prior assumptions in the model, such as inductive bias”
Identifiability of Path-Specific Effects (2005, unknown): “The ’forcing’ of the variables toz is called an intervention, and is denoted by do(z) in [Pearl, 2000].”
Reasoning in Non-probabilistic Uncertainty: Logic Programming and Neural-Symbolic Computing as Examples (2017, unknown): “The structural ‘common core’ of the last LP model in this sequence can provide, for instance, the basis for creating a Bayes Net by adding distributional information [Pearl, 2000, Pinosio, in prep.].”
The seven tools of causal inference, with reflections on machine learning (2019, unknown): “” Such sentences can be computed only when the model is based on functional relations or structural.(18) This three-level hierarchy, and the”
Inferring causal networks from observations and interventions (2003, unknown): “…of causal relationships, as opposed to mere statistical associations, gives us a sense of deep understanding of a system and a sense of potential control over the system's states, stemming from the ability to predict the consequences of actions that have not yet been performed (Pearl, 2000).”
Causes and explanations in the structural-model approach: Tractable cases (2002, unknown): “A different family of approaches evolved from the area of Bayesian networks, such as Pearl’ s approach to modeling causality by structural equations [1, 5, 19, 20 ].”