Margherita Benzi Università S. Raffaele di Milano, University of Genoa (Italy) Contexts for Causal Models.

Slides:



Advertisements
Similar presentations
Moral Relativism and Conceptual Analysis David J. Chalmers.
Advertisements

The Contextual Character of Evidence for Causal Claims Mauricio Suárez CaEitS conference, University of Kent, 7 September 2012.
WHAT IS THE NATURE OF SCIENCE?
Summer 2011 Tuesday, 8/ No supposition seems to me more natural than that there is no process in the brain correlated with associating or with.
Bayesian Network and Influence Diagram A Guide to Construction And Analysis.
Stuart Glennan Butler University.  The generalist view: Particular events are causally related because they fall under general laws  The singularist.
Causality Causality Hill’s Criteria Cross sectional studies.
GROUNDED THEORY © LOUIS COHEN, LAWRENCE MANION & KEITH MORRISON.
Soc 3306a Lecture 2 Overview of Social Enquiry. Choices Facing the Researcher What is the problem to be investigated? What questions should be answered?
The Case Study Approach By, Mark Traina. A Brief Case Review First reading: “Case-Oriented Comparative Methods” –All case studies are designed to uncover.
Introduction to Research Methodology
Testing Hypotheses About Proportions Chapter 20. Hypotheses Hypotheses are working models that we adopt temporarily. Our starting hypothesis is called.
Introduction of Probabilistic Reasoning and Bayesian Networks
Hume on Taste Hume's account of judgments of taste parallels his discussion of judgments or moral right and wrong.  Both accounts use the internal/external.
From: Probabilistic Methods for Bioinformatics - With an Introduction to Bayesian Networks By: Rich Neapolitan.
Designing Case Studies
Empirical Analysis Doing and interpreting empirical work.
Kent Where causal dualism comes from Monika Koeppl Causality, Cognition and the Constitution of Scientific Phenomena Department of Philosophy University.
Scientific method - 1 Scientific method is a body of techniques for investigating phenomena and acquiring new knowledge, as well as for correcting and.
Argumentation - 1 We often encounter situations in which someone is trying to persuade us of a point of view by presenting reasons for it. We often encounter.
Scaling and Attitude Measurement in Travel and Hospitality Research Research Methodologies CHAPTER 11.
Qualitative research in psychology. A distinct research process Inquiries of knowledge that are outside the framework prescribed by the scientific method,
Comparative Research.
Framing and testing hypotheses
Copyright © Cengage Learning. All rights reserved. 8 Tests of Hypotheses Based on a Single Sample.
RESEARCH A systematic quest for undiscovered truth A way of thinking
Causation and the Rules of Inference Classes 4 and 5.
CHAPTER 4, research design
Ch. 2: Planning a Study (cont’d) pp THE RESEARCH PROPOSAL  In all empirical research studies, you systematically collect and analyze data 
Methodology and Philosophies of research Lecture Outline: Aims of this session – to outline: what is meant by methodology the implication of adopting different.
WELNS 670: Wellness Research Design Chapter 5: Planning Your Research Design.
Psy B07 Chapter 4Slide 1 SAMPLING DISTRIBUTIONS AND HYPOTHESIS TESTING.
1 A Conceptual Framework of Data Mining Y.Y. Yao Department of Computer Science, University of Regina Regina, Sask., Canada S4S 0A2
LOGIC AND ONTOLOGY Both logic and ontology are important areas of philosophy covering large, diverse, and active research projects. These two areas overlap.
The Problem of Evil: McCabe, “The Statement of the Problem”
Gile Sampling1 Sampling. Fundamental principles. Daniel Gile
WHAT IS THE NATURE OF SCIENCE?. SCIENTIFIC WORLD VIEW 1.The Universe Is Understandable. 2.The Universe Is a Vast Single System In Which the Basic Rules.
Thomson South-Western Wagner & Hollenbeck 5e 1 Chapter Sixteen Critical Thinking And Continuous Learning.
Invitation to Critical Thinking Chapter 9 Lecture Notes Chapter 9.
Scientific Inquiry There will be a quiz tomorrow on the following 7 statements.
Uncertainty Management in Rule-based Expert Systems
Shoemaker, “Causality and Properties” Events are the terms involved in causal relations. But all causal relationships seem to involve a change of properties.
A Comparison of General v. Specific Measures of Achievement Goal Orientation Lisa Baranik, Kenneth Barron, Sara Finney, and Donna Sundre Motivation Research.
Research Methodology and Methods of Social Inquiry Nov 8, 2011 Assessing Measurement Reliability & Validity.
AOT Lab Dipartimento di Ingegneria dell’Informazione Università degli Studi di Parma Unifying MAS Meta-Models ADELFE, Gaia & PASSI Carole Bernon, Massimo.
Experimental Control Definition Is a predictable change in behavior (dependent variable) that can be reliably produced by the systematic manipulation.
Miracles: Hume and Howard-Snyder. * For purposes of initial clarity, let's define a miracle as a worldly event that is not explicable by natural causes.
What’s Ahead for Embedded Software? (Wed) Gilsoo Kim
Transient Unterdetermination and the Miracle Argument Paul Hoyningen-Huene Leibniz Universität Hannover Center for Philosophy and Ethics of Science (ZEWW)
Pag. Jan Lemeire ETRO dept., Vrije Universiteit Brussel Causality in the World vs Causality in the Mind Examples that puzzle philosophers 1.Me not watering.
CHAPTER 1 THE FIELD OF SOCIAL PSYCHOLOGY. CHAPTER OBJECTIVES After reading this chapter, you should be able to: Offer a definition of social psychology.
Formulating a research problem R esearch areas and topics.
Causal truthmakers vs Causal interpretations Federica Russo Philosophy, Kent.
WHAT MODELS DO THAT THEORIES CAN’T Lilia Gurova Department of Cognitive Science and Psychology New Bulgarian University.
HAYEK AS A METHODOLOGICAL INDIVIDUALIST Francesco Di Iorio Southeast University (Nanjing) 1.
Research Design
Validity in epidemiological research Deepti Gurdasani.
Chapter Nine Hypothesis Testing.
CHAPTER 5 Handling Uncertainty BIC 3337 EXPERT SYSTEM.
How to Research Lynn W Zimmerman, PhD.
Formulation of hypothesis and testing
MODULE 2 Myers’ Exploring Psychology 5th Ed.
THE FIELD OF SOCIAL PSYCHOLOGY
CASE STUDY BY: JESSICA PATRON.
If You Aren’t Dong Arguments, You Aren’t Doing Evidence
Comparative Method I Comparative methods deal primarily with finding and/or eliminating necessary and/or sufficient conditions that produce a given outcome.
Nature of Science Understandings for HS
Comparative Research.
Introduction.
Transformational Psychology
Presentation transcript:

Margherita Benzi Università S. Raffaele di Milano, University of Genoa (Italy) Contexts for Causal Models

Plan of the talk  Contrast two views of causality:  absolutistic view = causation is an absolute relation  relativistic view = causes are relative to contexts  Distinguish different kinds of contexts in causal models  Argue against the ‘absolutistic’ view

Relativistic positions: P. Suppes  What factors should be considered in investigating the causes of a certain effect?  “Perhaps one of the most puzzling and difficult aspects of the analysis of causality is the problem of how to handle the background that serves as a framework for the occurrence of the particular events under study. If, for example, we study the cause of a match lighting, to what extent must we consider meteorological conditions, the composition of the atmosphere, the absence of meteorites, etc.?”  Dependence on a ‘conceptual framework’:  With respect to one field, framework or background, one event may be a cause of another, and yet, when the field is changed and the framework is extended by the consideration of additional variables, the cause may turn out to be spurious.

Absolutistic positions: J. S. Mill  No real distinction between causes and conditions  The real cause is the whole of these antecedents; and we have, philosophically speaking, no right to give the name of cause to one of them exclusively of the others.  Nothing can better show the absence of any scientific ground for the distinction between the cause of a phenomenon and its conditions, than the capricious manner in which we select from among the conditions that which we choose to denominate the cause.

Types of Contexts [Hart and Honoré]  Context of Inquiry: different conceptual framework  Context of Occurrence: same conceptual framework, different ‘circumstances’

Probabilistic Causality  Test Situations (Cartwright)  Causal Background Contexts (Eells)  Context Unanimity requirement  Context Unanimity vs Objective Homogeneity [Twardy and Korb]

(Causal) Bayesian Nets  Relativity wrt the choice of possibly relevant variables (context of inquiry)  Relativity wrt variables in the DAG [‘V’]  Background causal field in bayesian Network: partial reductionism

Causal Nets  Background Context In assessing the causal role of X relative to Y, the appropriate background context consists of all variables which are 1.Direct parents of Y or of any intermediate variable between X and Y, and 2.Non-descendants of X.

The standard example A1 A2 A3 A4 A5  Variable A1: Season of the year, values  spring, summer, fall, winter   Variable A2: Rainfall during the season, values  yes, no   Variable A3: Sprinkler during the season, values  on, off   Variable A4: Wet pavement, values  yes, no   Variable A5: Slippery pavement, values  yes, no 

Circularity and Reduction  Circularity can eliminated in the model but it can still be present in the considerations that orient the choice of variables (context of Inquiry)  If we eliminate the dependence from the context of Inquiry, we obtain a complete reduction of causes to probabilities

Reductionism and Absolutism: The All Embracing Causal Net  Failures of faithfulness can disappear in a more fine-grained description of the domain [Papineau 2001]  Metaphysical Assumption: Nature is not “conspiratorial”  Causal relationships at higher level are fixed by those at the lowest level. Patterns of correlation can thus be misleading about causal structure at any higher level.  But at the bottom level there is no metaphysical room for such failures of faithfulness, since there the causal order is simply constituted by the correlational order.  Causal structures are determined by an extremely rich underlying network of correlations, from which God can read off the causal facts  What if there is no lowest level, if there is no limit to how fine we can cut up our mechanisms? Then reductionists can adopt a limiting procedure.

Reductionism and Absolutism: The All Embracing Causal Net  Two relativisations to the granularity of the description [Spohn 2001]:  property of being a direct rather than an indirect cause  where there appears to be a direct or indirect causal dependence within a coarse-grained frame, there may be none within a more fine-grained frame, and vice-versa.  Relativity to the agent

Sprinkler Example again A1 A2 A3 A4 A5  Variable A1: Season of the year, values  spring, summer, fall, winter   Variable A2: Rainfall during the season, values  yes, no   Variable A3: Sprinkler during the season, values  on, off   Variable A4: Wet pavement, values  yes, no   Variable A5: Slippery pavement, values  yes, no 

Context Dependencies  L 0 (General background knowledge). the most extensive description of the universe. (“the whole prior state of the universe”, or in “the whole environment”)  L 1 (Context of Inquiry). a subset of the set of factors in L 0 : ‘L 1 -factors’In probabilistic causality and Causal modelling, L 1 -factors seem identifiable as the variables of the domain of interest in a given causal inquiry (factors which are potentially statistically associated.  L 2 (Causal models). the selection of a subset of the set of L 1 -factors the set V of the variables in the DAG).  L 3 (Set of the Explicitly Relevant Factors). the set of factors which are more directly ‘causally relevant’ for the assessment of a causal relation (the components of ‘Causal Background Contexts’ in probabilistic causality).

Problems with Absolutism  Duplications of causes: truth w.r.t. All-embracive Net and truth w.r.t. Particular models  Forced commitment to physicalism

Relativity and Objectivity How to reconcile ‘relativity to the context of Inquiry with objectivity?  Williamson: objective mind-dependent notion of causality (epistemic causality), where objectivity is based on the notion of the ‘most complete’ background knowledge β*  Advantages: β* is well defined (as long as the indicators of causality can be delimited) Causality is not tied to a particular agent – indeed This view is compatible with relativistic view if we don’t require that the different ‘Contexts of inquiry’ collapse

Conclusion: consequence of AA Epistemological point of view: acceptance of the ‘Absoluteness Assumption’ (a) duplication of causal judgments (w.r.t the true All-embracive Net, and w.r.t the local nets), (b) commitment to some form of very strict physicalism.

Conclusion: Avoid Arbitrariness a good reason to reject Absoluteness Assumption Absoluteness assumption is not the only option to avoid arbitrariness. Relativity to context as “relativity to context of Inquiry” objectivity without absoluteness

Conclusion objectivity without absoluteness: causal statements will be as objective as any other scientific statement. The price to pay is renouncing to a complete reduction of causes to probabilities, but it is highly dubious that, from an epistemological or a methodological point of view, this reduction offers any real advantage.