Causal Reasoning. Note. Being g is necessary for being f iff being f is sufficient for being g

Similar documents
Scientific Explanation- Causation and Unification

Math 38: Graph Theory Spring 2004 Dartmouth College. On Writing Proofs. 1 Introduction. 2 Finding A Solution

The problem of disjunctive causal factors. Following Hitchcock: fix K and do everything within a single cell K (which we don t mention).

Chapter 2. Mathematical Reasoning. 2.1 Mathematical Models

A New, Universal Frame Of Reference

Philosophy 240 Symbolic Logic. Russell Marcus Hamilton College Fall 2013

Physic 602 Conservation of Momentum. (Read objectives on screen.)

Bounding the Probability of Causation in Mediation Analysis

Contrastive Causation

Tooley on backward causation

The Conjunction and Disjunction Theses

Why Care About Counterfactual Support? The Cognitive Uses of Causal Order Lecture 2

Physicalism Feb , 2014

Statistics 251: Statistical Methods

LIVING IN THE ENVIRONMENT 17 TH

1.1 Statements and Compound Statements

CS 453 Operating Systems. Lecture 7 : Deadlock

Day 15. Tuesday June 12, 2012

10.2 PROCESSES 10.3 THE SECOND LAW OF THERMO/ENTROPY Student Notes

Why write proofs? Why not just test and repeat enough examples to confirm a theory?

18 : ( ( ( ( ( ( ( ( ( ( ( (3-4. (1. (2. (3. (4-5» «. (4 (3 (2

Science. Science. Passage I

As you scroll through this review, you move your hand; this causes the

Précis of Modality and Explanatory Reasoning

(1) If Bush had not won the last election, then Nader would have won it.

Symbolic Logic 3. For an inference to be deductively valid it is impossible for the conclusion to be false if the premises are true.

Lesson 3-1: Solving Linear Systems by Graphing

3 The Semantics of the Propositional Calculus

PHYSICS 107. Lecture 8 Conservation Laws. For every action there is an equal and opposite reaction.

Introduction to Basic Proof Techniques Mathew A. Johnson

Debunking Misconceptions Regarding the Theory of Evolution

The Philosophy of Physics. Is Space Absolute or Relational?

MI 4 Mathematical Induction Name. Mathematical Induction

Discrete Structures Proofwriting Checklist

You will toggle between Rutherford atom and plum pudding atom.

The paradox of knowability, the knower, and the believer

Unit 1: Introduction WHAT IS BIOLOGY, NATURE OF SCIENCE, BIOCHEMISTRY

Chapter 11 Heat Engines and The Second Law of Thermodynamics

Chapter 7 Rocket Propulsion Physics

Objective probability-like things with and without objective indeterminism

Faculty of Philosophy, University of Cambridge Michaelmas Term Part IA: Metaphysics Causation

Chapter 12- The Law of Increasing Disorder

Introducing Proof 1. hsn.uk.net. Contents

THE SIMPLE PROOF OF GOLDBACH'S CONJECTURE. by Miles Mathis

Simultaneity And Time Dilation

Response to Kadri Vihvelin s counterfactuals and dispositions

Guest Speaker. CS 416 Artificial Intelligence. First-order logic. Diagnostic Rules. Causal Rules. Causal Rules. Page 1

Regularity analyses have failed; it is time to give up and try something else: a counterfactual analysis.

Laws of Nature. What the heck are they?

Statistics 301: Probability and Statistics 1-sample Hypothesis Tests Module

6.080 / Great Ideas in Theoretical Computer Science Spring 2008

HPS 1653 / PHIL 1610 Introduction to the Philosophy of Science

CS 188: Artificial Intelligence Spring Announcements

Physical Matter and Entropy Were Made

Weather and climate. reflect. what do you think? look out!

Lesson 6-1: Relations and Functions

EC 331: Research in Applied Economics

Electrical measurements:

Correct Resolution of the Twin Paradox

GALILEAN RELATIVITY. Projectile motion. The Principle of Relativity

Mathematics 1a, Section 4.3 Solutions

You are given two carts, A and B. They look identical, and you are told they are made of the same material. You put A at rest on a low-friction

Introduction. Introductory Remarks

Chapter 6. Net or Unbalanced Forces. Copyright 2011 NSTA. All rights reserved. For more information, go to

What is the "truth" about light? Is it a wave or is it a particle?

STATISTICS Relationships between variables: Correlation

Writing proofs. Tim Hsu, San José State University. May 31, Definitions and theorems 3. 2 What is a proof? 3. 3 A word about definitions 4

Russell s logicism. Jeff Speaks. September 26, 2007

Writing Mathematical Proofs

Explanation and Argument in Mathematical Practice

Preptests 55 Answers and Explanations (By Ivy Global) Section 4 Logic Games

Lecture 6: Finite Fields

LECTURE FOUR MICHAELMAS 2017 Dr Maarten Steenhagen Causation

Propositional Logic: Part II - Syntax & Proofs 0-0

P (E) = P (A 1 )P (A 2 )... P (A n ).

THE P-ADIC NUMBERS AND FINITE FIELD EXTENSIONS OF Q p

Conditionals. Daniel Bonevac. February 12, 2013

David Lewis. Void and Object

Core Chemistry UNIT 1: Matter & Energy Section 1: The Law of Conservation of Mass Section 2: States of Matter & Intro to Thermodynamics

Understanding Exponents Eric Rasmusen September 18, 2018

Climate Change. Presenter s Script

CS1800: Strong Induction. Professor Kevin Gold

PHYS:1200 LECTURE 18 THERMODYNAMICS (3)

A Brief Introduction to Proofs

CS 360, Winter Morphology of Proof: An introduction to rigorous proof techniques

Indicative conditionals

DAY 28. Summary of Primary Topics Covered. The 2 nd Law of Thermodynamics

LECTURE 15: SIMPLE LINEAR REGRESSION I

Long division for integers

Lecture 34 Woodward on Manipulation and Causation

Exam 2--PHYS 101--Fall 2014

Lecture 3: Sizes of Infinity

0 questions at random and keep in order

INTRO TO SCIENCE. Chapter 1. 8 th grade

Properties of Sequences

Atmosphere - Part 2. High and Low Pressure Systems

Calculus: What is a Limit? (understanding epislon-delta proofs)

Where linguistic meaning meets non-linguistic cognition

Name Period Date. D) density. E) speed.

Discrete Mathematics and Probability Theory Summer 2014 James Cook Note 5

Transcription:

145 Often need to identify the cause of a phenomenon we ve observed. Perhaps phenomenon is something we d like to reverse (why did car stop?). Perhaps phenomenon is one we d like to reproduce (how did I get extra money in my bank?). Causal claims often rest on special kind of generalization (a causal generalization) C (event 1) caused E (event 2) e.g. firing a gun caused someone to have a bullet wound e.g. running out of fuel will cause any car to stop running (this is a generalization) Causal generalizations have predictive force : when you combine a causal generalization with facts in a particular case, you generate a prediction about what will happen in that case Causal generalization can be represented using a conditional : a general condition of form of all x, if x if F, then x is G A shorthand way of talking about these conditionals in terms/necessary (or sufficient) conditions Necessary Conditions X s being G is a necessary condition for x s being F iff for all x, if x if F, then x is G [for all x: (F x ) Gx )] Sufficient Conditions X s being F is a sufficient condition for x s being G iff for all x, if x is F, then x is G [for all x: (Fx ) Gx)] Note Being g is necessary for being f iff being f is sufficient for being g G F e.g. being a square is sufficient for being a rectangle

145 e.g. being a rectangle is necessary for being a square True/ false? 1. Being a car is sufficient condition for being a vehicle (T) 2. Being a car is necessary 2 and condition 3 are for pretty being much a vehicle same (F) 3. Being a vehicle is a sufficient condition question for being a car (F) 4. Being a vehicle is necessary condition for being a car (T) 5. Being an integer is a sufficient condition for being an even number (F) 6. Being an integer is a necessary condition for being an even number (T) Important: not all relationships of necessity/sufficiency are causal (e.g. relationship b/w being a square and rectangle). Cause PRECEDES the effect this does not apply to a square as it does not precede to be a rectangle it happens at the same time. But plausibly all causal generalizations encode sufficient conditions (the cause is sufficient for effect) The sufficient condition test By looking at specific cases we can gather evidence about whether something is (or not) a sufficient condition for something else The sufficient condition test (SCT) If feature F is present when feature G is not, then F is eliminated as a possible sufficient condition for G Look at where target condition is NOT present Why? If F were sufficient for G, then any time F were present, G would need to be present too, so if G = absent, then F must be absent too Note : eliminating a condition as possibly sufficient for G by the SCT is deductively valid method of arguing Target Case 1: A B C D G Case 2: A B C D G Case 3: A B C D G Case 1 : irrelevant, no possible violation of SCT target feature(g) is present in this case therefore, does not matter Case 2: SCT eliminates B and C as possibly sufficient for G Case 3: SCT eliminates A as possibly sufficient for G Does it follow that D is sufficient for G? (not necessarily) o Case 4 exists: A B C D G

145 There are too many cases to check for sufficiency, but must simply notice that it is not possible to come up with counter examples Case 4: SCT eliminates D as possibly sufficient for G But: the SCT can inductively support the inference that something is sufficient for G. If The Necessary Condition Test (NCT) By looking at specific cases, we can gather evidence about whether something is (or not) a necessary condition for something else If feature F is absent when feature G is present, then F is eliminated as possible necessary condition for G e.g. Look at where the target condition IS present Case 1: A B C D G : irrelevant, no possible violation of NCT Case 2: A B C D G : NCT eliminates A as possibly necessary for G Case 3: A B C D G: NCT eliminates B and D as possibly necessary for G Does it follow that C is necessary for G? not necessarily Case 4: A B C D G o NCT eliminates C as possibly necessary for G BUT: the NCT can inductively support the inference that something is necessary for G. if we never are able to eliminate a condition via NCT, there are inductive grounds for thinking it is necessary for G *as long as no counter example, have good proof that generalization is true * Note: if C is necessary for G, Using SCT to reach positive conclusion To infer that C is sufficient for G [for all x : (C x) Gx)]. We must give conclusion plenty of chance to be falsified SCT to generalization 1. Test cases where C is present 2. Test Cases where G is absent 3. Fail to find any case where C is present and G is absent 4. Test enough cases of various kinds that would be likeliest to include a case where C is present and G is absent *condition 4 relies heavily on background knowledge : look for counter examples of where they would be the most likely to arise*

145 Using NCT to reach positive conclusion To infer that C is necessary for G [for all x: (Gx)Cx)], we must give conclusion plenty of chance to be falsified SCT to Generalize 1. Test cases where C is absent 2. Test cases where G is present 3. Fail to find any case where C is absent and G is present 4. Test enough cases of various kinds that would be likeliest to include a case where C is absent and G is present *condition 4 relies heavily on background knowledge: look for counter examples of where they would be the most likely to arise* Combining SCT and NCT Ideally, applying SCT and NCT, will gather inductive evidence that identifies the cause behind given event Inferring Causation 1. Observation of phenomenon P requiring causal explanation 2. Use SCT and NCT to isolate some Q necessary and sufficient for P 3. Conclusion : infer that Q was causally responsible for P Such reasoning can be presented as either justificatory or explanatory Inference is defeasible at 2 steps: inferring that something is necessary/sufficient (since new cases could always arise) as well as inferring from this that it was a cause (since many conditions that are necessary/sufficient are not themselves causal) Combining results of SCT and NCT is important because: Being sufficient for something isn t nearly enough to have caused it o Causal Preemption: imagine 2 guns firing, one right after another, at same deer. Both shots are sufficient to cause death but only one does (cannot infer which bullet killed deer) Being necessary for something isn t nearly enough to have caused it

145 o Causal Underdetermination: being in hotel where Legionnaire s Disease originated was necessary for contracting it then. But this isn t what caused ppl to contract Legionnaire s disease (need both necessity AND SUFFICIENCY) Complication: Assuming Normality Very natural to think that striking a match causes it to light, b/c it is both necessary and sufficient for getting the match to light. But there are exceptions Not physically necessarily to strike match to light it : a match can light if surrounding environment gets hot enough Nor is it physically sufficient : match can fail to light when struck (e.g. if struck underwater) It might seem impossible to write all these exceptions into useable causal claim: match lit b/c it was struck in an environment that was (i) relatively cool, (ii) not underwater Generally, we get around this by simply assuming that conditions were normal : striking match in normal conditions is necessary and sufficient to light match Assuming normalcy introduces another point of defeasibility into our causal reasoning : conditions that are normal (and that license thinking of something as a cause) can become abnormal when further Co variation Suppose want to know whether CO2 emissions from burning of fossil fuels are causing melting of polar ice caps. Can we use NCT and SCT to guide us? NO Causal feature seems to be introduction of CO2 into atmosphere from terrestrial sources ( C ), while target feature seems to partial melting of polar ice caps (M) Testing whether C is necessary for M (NCT) requires looking at cases where C. but never observe such cases there is always introduction of CO2 into atmosphere from terrestrial sources. Testing whether C is sufficient for M (with SCT) requires looking at cases where M is absent. But can never observe such cases every summer, polar ice caps melt significantly NCT and SCT are not well suited to cases like this. Instead we need tests that track whether w/ abundance of causal feature co varies (is correlated) w/ target feature Method of Co Variation Q1 : does change (+/ ) in the causal feature C imply change (+/ ) in target feature G?

145 Q2: does change (+/ ) in target feature G imply change (+/ ) in causal feature C? Yes answers to both questions will often inductively support concluding that a change (+/ ) in C is cause of change (+/ ) in G Co variation and Causation Important to note that changes in C can be very highly correlated w/ changes in G without there being any causal relationship b/w them. Co variation does not imply causation For one, co variation is symmetric: A co varies w/ B iff B co varies w/ A. but causation is non symmetric : if A is cause of B, then B cannot be cause of A o Correlation is symmetric For two, co variation between A and B if often explained by fact that both A and B have common cause, while A and B have no causal relationship to one another Last, co variation is sometimes accidental (for e.g. causes of autism Background Knowledge Need extensive background knowledge to disentangle co variation from causation: method of co variation is good way to discover potential causal connections, but further investigation is generally needed to identify a causal mechanism.

145

145