3 What’s Important About Causality? ExplanationAssociation provides prediction, but not explanationIdentifying causal mechanisms may uncover underlying reasons for relationshipsControlUnderstanding causality allows us to predict the effects of actions without performing themAllows more efficient exploration of the space of possible solutions
7 Are Feathers Associated with Flight? Do they have a casual relationship with the ability to fly?
8 Related Fallacies Common (Questionable) Cause Fallacy Post Hoc Fallacy This fallacy has the following general structure:A and B are regularly associated (but no third, common cause is looked for).Therefore A is the cause of B.Called “Confusing Cause and Effect” fallacy, if in fact, there is not common cause for A and BPost Hoc FallacyA Post Hoc is a fallacy with the following form:A occurs before B.
17 Examples of Abductive Reasoning A Medical DiagnosisGiven a specific set of symptoms, what is the diagnosis that would best explain most of them?Jury Deliberations in a Criminal CaseJurors must consider whether the prosecution or the defense has the best explanation to cover all of the evidenceNo certainty about the verdict, since there may exist additional evidence that was not admitted in the caseJurors make the best guess based on what they know
18 “… when you have eliminated the impossible, whatever remains, however improbable, must be the truth.”- Sherlock Holmes
19 Abductive Reasoning in Science Abduction selects from among the hypotheses being considered, the one that best explains the evidenceNote that this requires that we consider multiple alternative hypothesesAbductive Reasoning is closely related to the statistical method of Maximum Likelihood EstimationPossible threats to validitySmall hypothesis spacesSmall amounts of evidence to explain
20 Challenges in Abductive Reasoning Creating hypothesis spaces likely to contain the “true” hypothesisApproach: create large hypothesis spacesKnowing when more valid hypotheses are missing from the hypothesis spaceApproach: constantly evaluate and revise the hypothesis space (multiple working hypotheses)Creating good sets of evidence to explainApproach: seek diverse and independent evidence with which to evaluate hypotheses
21 Why use multiple working hypotheses? Objectivity: Helps to separate you from your hypotheses; shift from personal investment in hypotheses to testing the hypothesesFocus: Reinforces a focus in falsification rather than confirmationEfficiency: Allows experiments to be designed to distinguish among competing hypotheses rather than evaluating a single oneHarmony: Limits the potential for professional conflict and rejection because all hypotheses are considered and evaluated
22 “Strong Inference” John R. Platt, Science, October 1964 “Strong Inference - Certain systematic methods of scientific thinking may produce much more rapid progress than others.”Not all science/research is created equalDon’t confuse research activity with effective researchActivity: building systems; proving theorems; conducting surveys; writing and publishing articles; giving talks; obtaining grantsResearch: improved predictions; better understanding of relationships; improved control of computational artifactsMany researchers are active; only a subset do effective research
24 Arguments and Fallacies Aside from general reasoning methodologies, one must ensure the validity of all arguments used in any research endeavorAn argumentConsists of one or more premises and a conclusionA premise is a statement (a sentence that is either true or false) that is offered in support of the claim being made, which is the conclusion (also a sentence that is either true or false)Modus Ponens (and Modus Tollens)A fallacyGenerally, an error in reasoning (differs from a factual error),An "argument" in which a logically invalid inference is made (deductive) or the premises given for the conclusion do not provide the needed degree of support. (inductive)
25 Common Fallacies Ad Hominem Appeal to Authority Appeal to Belief Appeal to Common PracticeAppeal to PopularityBegging the QuestionBiased SampleHasty GeneralizationIgnoring A Common CauseBurden of ProofStraw ManSee: