- Created by: GemmaBruce
- Created on: 16-04-19 19:46
What is Philosophy of Science?
- Science is a cultural phenomenon with an unparallelled prestige and influence on our lives
- Science is extremely successful at generating knowledge, what are its methods for doing so?
- Science introduces thousands of new technical terms into our language, how does it supply them with meaning?
- Science relies on evidence from observation and experimentation, what kind of inference is used in science to get the evidence to theory? Are these justified?
- Philosophy of science has grappled with all of these questions
Science and Induction
- Science is empirical: its findings are based on observation and experiment
- Science does not merely record observations and experiments, it also extrapolates from them to predict new ones
- Science draws inference about the unobserved from the directly observed
- Extrapolation cannot proceed by deductively valid arguments: in such arguments, the conclusion is already entailed by the premises
- Induction is such a process of extrapolation from given data to new ones from the observed to the unobserved
- It seems induction is central to scientific reasoning
- When we infer a prediction from given data, when we infer soemthing about the unobserved from the observed, we are reasoning in accordance with an inductive argument
- Inductive arguments can take different forms (the sun will rise tomorrow)
- This method of reasoning is called enumerative induction
- The premises enumerative cases where things that have a given property A also have another property B
- The last premise says in effect that all relevent observed cases have been enumerative in the earlier premises
- another form of argument by enumerative induction has a stronger conclusion - not about the nect case to be observed, but about all things or events of the relevent kind (the sun rises everyday)
- Arguments by induction are not deductively valid - it is possible for the premises to be true and the conclusion false (all swans are white, we will be fed tomorrow arguments for examples by Russell)
The Problem of Induction
- Is the use of induction reasonable and justified?
- What would it take for induction to be justified? We would have to have an argument to the effect that their conclusions are always true - hence no sound argument could show that they are
- We presumably believe that most such arguments that have true premises have true conclusions, the premises of inductive arguments make their conclusion likely or probable
- Hume's dilemma is one of the most popular objections
Is Induction Irrational
- Does Hume's dilemma suggest that it is irrational to use induction? No but it appears to show that using induction is not rationally required
- Compare someone who infers by induction with three other characters:
> Cautious Martian: has no opinion on whether the next instance will be like her past experience
> The dogmatist: forms beliefs about the future that are not informed by beliefs about past regularities
> The counterinductivist: holds that regularities in her past experience so far will not continue in the next instance
- Are all these attitudes on a par as far as rationality is concerned?
The Definitional Response (Russell's response)
- According to this response, popular amongst "ordinary language philosophers" - words like rational and justified just mean being in conformity with the principles of induction
- Why do Russell and others think there is a problem of induction?
- When Russell says that observed instances are never by themselves a reason for an inductive conclusion, he is guilty of an ignoration elenchi ( alogical fallcy which consists in refuting something that the opponent has not asserted) by redefinition
- His assertion that the premises of an inductive argument never by themselves constitute a logically conclusive reason for an inductive conclusion in no way contradicts the commensence assertion that they frequently constitute a reason in the ordinary sense of the word
The Definitional Response (Problem's With The Resp
- It is not clear that the ordinary meaning of 'reason', 'rationality' ect is as suggested by Ayer and Edwards and that the induction skeptic redefines them
- We can ask whether we should be rational, or seek to make justified inferences; or whether we should expect rational inferences to have true conclusions
The Inductive Justification of Induction
- One response is that induction is considered as just one form of inductive inference, and that we could non-circularly use other such forms to justify it
- Sticking to induction is not clear whether the objection works, asuume that induction has no counter-example, we could then reason that inductiuon has not failed and so will not fail next time
- Observations about the argument:
> The argument relies on empirical premises that might be false. This suggests that it does not beg the question
> There is no premise in the argument to the effect that induction gets us from the true premises to true conclusions. Reasoning in accordance with a rule it is not the same as asserting that the rule gets us from true premises to true conclusions. This also suggests that the argument does not beg the question
> The main problem with the strategy of using a self-supporting argument relying on past instances is that it is not only available to the would-be justifier of induction, but also of counterinduction. A counterinductionist holds that regularities in her experience so far will not continue in the next instance
The Pragmatic Justification Of Induction (The Just
- A pragmatic justification does not establish that most instances of the rule with true premises have true conclusions
- It tries to establish that if any method is reliable, then induction does
- Suppose that any method to make inferences from observed to unobserved or not yet observed cases is reliable. Then induction will successfully establish the reliability of that method. So induction is successful too
- We can argue that it is reasonable to use induction:
> If the universe is so chaotic that no method is reliable, then it does not matter what method we use
> If the universe is not so chaotic that no method is reliable, then induction is reliable
> Either way we are no worse off using induction
The Pragmatic Justification Of Induction (The Obje
- Two problems with the argument:
> if another method is reliable, induction has one notable success: showing that the other method is reliable: it does not clearly follow it is generally reliable, or as successful as the other method
> Suppose the universe is so chaotic that no method is reliable. It is then perhaps better to be cautious, in the sense of suspending judgement about whether the next instance will be like the observed ones
the Context of Discovery and The Context of Justif
- Theoretical Hypothesis
> How did people come up with it?
> What reasons do we have to accept it as true?
- Context of discovery = asks about a historical account of what led to the formulation of the hypothesis
- Context of justification = asks about how well the hypothesis stands up to tests, and whether it has other features that a good hypothesis is supposed to have
- Do facts about the context of discovery have a bearing on the context of justification? Standard answer = no
Induction as a Method of Discovery
- The problem of induction casts doubt on whether the premises of an inductive argument justify the acceptance of the conclusion
- Given the seperation of the context of discovery from the context of justification. It is a different question whether the inductive method is a good method of discovery
- Inductive method = first gather data through observation and experiment and then formulate a hypothesis that extrapolates from them
- On the inductivist view, observation is typically held to be objective, publically available, and yield beliefs we can hold with certainty
- two problems for the view that induction is a method of discovery in science:
> Many scientific hypotheses are no mere extrapolations of data, hypotheses about unobservable entities are never inferred from observations or experiments by a simple form of inductive argument
> Without some guidance by theoretical hypotheses, there is no fruitful observation or experimentation there are too many things to observe, or to test out and we don't know which ones are relevent
The Role of Creativity in Discovery
- There is no algorithm for inventing hypotheses
- Irrational factors like dreams, hallucinations ect may play a role
Hypotheses and Their Test
- According to the hypothetico-deductive model of science (HD model), what scientists do can be broken down in three parts:
> Come up with a theoretical hypothesis
> Deduce testable predictions from the hypothesis
> Test whether those predictions are true or not
- According to the HD model, there is not that much to say about the first step. Coming up with a hypothesis requires creativity on the part of the scientists, the philosopher of science annot formulate rules guiding that sort of creativity
- The arguments by which testable predictions are derived from theoretical hypotheses may be complex and involve a great deal of mathematics still, the logic used in such deviations is ordinary deductive logic. No inductive arguments are involved at all
- A hypothesis can be tested by either observations or experiments. In advanced sciences, experiments are typically needed, involving special techniques and machines
- Due to specialization, the three steps may be performed by different people
- The Dilemma: there cannot be either a deductive or an inductive argument for the claim that most inductive arguments with true premises have true conclusions
- It is possible that induction fails to lead from true premises to true conclusion in most cases, there cannot be a deductive argument that induction works in most cases
- You could argue induction has worked in the past and will work again but this is circular and begs the question
- Hume's dilemma does not turn on at all on observed failures of induction, it would still apply even if induction had always worked in the past
Hypothetico-Deductivism and Inductive Support
- On the HP model, scientists come up with a hypothesis H and deduce testable consequences, such as E: H/E
- Suppose that the argument from H to E is deductively valid, and that we have tested E
- If E is false, then H as been falsified, a deductively valid argument with a false conclusion needs to have at least one false premise, and H is the only premise here. Hence H needs to be rejected
- If E is true, we cannot conclude that H is true, a deductively valid argument with a true conclusion need not have a true premise
- It is plausible that the truth of E is some evidence for H - that it provides inductive support for H,or confirms it
- This suggests a general answer to question 3, which we may call the "hypothetico-deductivist account of inductive support"
- HD account of inductive support: a hypothesis H is inductively supported by evidence E if and only if H deductively entails E
Four Questions about Induction (The Four Question
1- How, if at all, do scientists use induction to come up with hypotheses?
2- Is it reasonable to use induction to come up with hypotheses?
3- Under what circumstances do we take a hypothesis to be inductively justified?
4- Is it reasonable to take hypotheses to be inductively justified?
- The Questions are related but distinct
Four Questions about Induction (Explaining the Fo
- Regarding question 1: we saw that there is no algorithm for coming up with interesting scientific hypotheses
- Question 2 is closely related to 1: induction is very limited in scope
- Question 4: poses the traditional problem of justifying induction
- Question 3 , sometimes called the "descriptive problem of induction". For the sake of the argument, we will grant that we legitimately use induction to justify hypotheses. The descriptive question is then a question about the relationship between a given hypothesis and a given body of evidence: when does the evidence inductively support the hypothesis?
- We can also put this questiuon in terms of arguments. A good inductive argument is one where the premises inductively support or confirm the conclusion. So the question is what the good inducive arguments are
The Paradox of The Ravens (Part 1)
∀x (x is a raven → x is black)/ Fred is a raven → Fred is black
- This is a valid deductive argument
- Suppose Fred is a raven, and turns out to be black. Then the hypothesis that all ravens are black has gained some inductive support
- Now suppose that Fred is not a raven, but a shoe. By the truth table for Fred is a raven -> Fred is black is true
- By the HD account of inductive support, observing that Fred is a shoe- and therefore not a raven, confirms the hypothesis that all ravens are black
- Recall, the truth table for -> is such that if the antecedent (P) is false, then the overall conditional is true
- Thus, the HD account of inductive support combined with the basic logical truths seem to entail that observing a shoe provides confirmation of the hypothesis that all ravens are black
- This is very intuitively implausible and so seems to count against the HD account of inductive support
The Paradox of The Ravens (Part 2)
- Suppose we give up the HD account of inductive support. We still have not solved the problem, because other plausible assumptions still entail that any observation of a shoe that is not black - or more generally, of a non-black non-raven- inductively confirms that all ravens are black
- Consider the following apparently independently plausible principles:
> A hypothesis of the form ∀x(Fx → Gx) is confirmed by each statement of the form Fa ∧ Ga
Equivalence: if E confirms H and H0 is logically equivalent to H, then E confirms H0
-This is an instance of the logical rule of contraposition
The Paradox of The Ravens (Part 3)
-Hence we can reason:
> "This non-black thing is a non-raven" confirms "All non-black things are non-ravens" (by positive instance)
> "All non-black things are non-ravens" is logically equivalent to "All ravens are black"
> "This non-black thing is a non-raven" confirms "All ravens are black" (by1,2 Equivalence"
- Thus, we seem to arrive at a paradox: all of the principles we employed appear true, the rules of inference we employ appear valid, yet the conclusion reached appears at first glance intuitively absurd: surely observing a white shoe doesn't provide any confirmation of the claim that all ravens are black
A Solution To The Paradox of The Ravens?
- [W]hat is wrong [lies] in our reliance on a misleading intuition in the matter: The impression of a paradoxical situation is not objectively founded; it is a psychological illusion. (Carl G. Hempel, “Studies in the logic of confirmation”, Mind 54 (1945)
- According to Hempel, we mistakenly take the "all ravens are black" to be about ravens only, rather than about everything our domain of discourse
- Intuition is less strong if we need observations to figure out that a non-black object is not a raven
- Ituition less strong if we consider a world with lots of ravens and very few nonblack things
The Grue paradox
- What has come to be known as the "grue paradox" casts doubt on one of these principles:
> A hypothesis of the form ∀x(Fx → Gx) is confirmed by each statement of the form Fa ∧ Ga
> Now consider the hypothesis that all emeralds are green. This seems to be confirmed by the each observation of an object which is an emerald and is green
> But now think about how observing green emeralds could equally confirm other, less intuitvely acceptable hypotheses. Such as the hypothesis that all emeralds are grue
> Define an object to be grue if and only if: It is examined before noon on 12 March 2020 and green, not examined before noon on 12 March 2020 and blue
> Intuitively, it seems that observations of the form "a is an emerald and a is grue" do not confirm the statement "all emeralds are grue". So positive Instance appears false
> But 'grue' is a very strange predicate. Maybe Positive Instance holds for all "normal" predicates F and G. It is possible to articulate the conditions that F and G need to meet?
The Grue paradox (Part 2)
- One idea: F and G need to stand for natural kinds or natural groupings of things. Natural kinds or groupings are here opposed to arbitrary kinds or groupings. Roughly, two thing belong to the same natural kind or grouping if they are similar to each other. Different things can be grue without being similar to each other
Falsification (Part 1)
- Scientific hypotheses are typically universal claims (of the form ∀xAx). Such claims can be conclusively falsified. But they cannot be conclusively verified
- Inductive logic tries to specify conditions under which such hypothesis are confirmed, or inductively supported. The raven and grue paradoxes show that this is difficult to do
- A radical response: there is no such thing as confirmation, or inductive support
- Karl Popper, one of the most influential philosophers of science in the 20th century, combines the hypothetico-deductive view of science with a rejection of the idea that hypotheses can be positively confirmed
-According to Popper, confirmation has no role to play in scientific practice. Scientists do not perform the same experiments all over again to pile up confirming instances. Rather, they:
> Try to formulate an interesting hypothesis
> logically deduce consequences from it
> try to make observations and devise experiments that might show these consequences to be false
Falsification (Part 2)
- In his words: "proposing bold hypotheses, and exposing them to the severest criticism, in order to detect where we have erred"
The Demarcation Problem
- The demarcation problem =the boundary between scientific and non-scientific ways of forming beliefs about the world. What are the distinctive characteristics of scientific theories and methods that distinguishes them from pseudo-science or non-science and superstition?
- Popper’s example of non-falsifiable theories: Astrology, Historical Materialism, Psychoanalysis
-Getting the demarcation of science right seems important because of the prestige and authority of science, and because its influence on education and public policy
- Popper’s answer to the demarcation problem: it is distinctive of science to seek falsifications of its theories
Popper on The Problem of Induction
- Popper claims, in various places, to have solved the problem of induction, and that his solution has not been sufficiently appreciated
-Popper’s response is to hold that
> There is no such thing as positive confirmation or inductive support for a hypothesis
> That observation and experiment still bear on the question of whether we should accept or reject a hypothesis, because it may falsify them
-Regarding the first point: Popper emphasizes that there are plenty of positive instances of pseudo-scientific hypotheses, and that we should not take these hypotheses as confirmed
- Regarding the second point, Popper allows that a hypotheses may be indirectly favoured by observation and experiments, namely if its rivals are eliminated
Against Falsificationism: Existential Statements
- Popper claims that scientific hypotheses can be conclusively falsified, but never conclusively verified
- This is true of hypotheses that take the form of universal quantifications. Hypotheses that take the form of existential quantifications can be conclusively verified, it would seem, but never conclusively falsified?
- How important are existential hypotheses in science?
Against Falsificationism: The Problem of Auxiliary
-Typically, no observation statements can be deduced from hypotheses alone
-Such deductions require auxiliary hypotheses about:
> The functioning of the measuring apparatus or conditions of observation
> Other variables influencing the measurement or observation
- If the deduced observation statement is false, it might be an auxiliary hypothesis that is at fault
- Indeed, scientists do not always reject theories that would be falsified by Popper’s lights. They may modify the theories, or question auxiliary assumptions
- Example: Leverrier’s defence of Newton’s theory in the face of irregularities in the motion of Uranus
Against Falsificationism: Accepting Hypotheses
- If ... ‘this law is scientifically accepted’ ... merely meant ‘this law has withstood severe tests’—and there were no suggestion at all that a law which has withstood severe tests is likely to withstand further tests, ... then science would be a wholly unimportant activity. It would be practically unimportant, because scientists would never tell us that any law or theory is safe to rely upon for practical purposes; and it would be unimportant for the purpose of understanding, since in Popper’s view, scientists never tell us that any law or theory is true or even probable. (Hilary Putnam, “The ‘Corroboration’ of Theories”, in Paul Arthur Schilpp (ed), The Philosophy of Karl Popper, vol. I, Open Court Publishing, 1974, p. 222-223.)
Induction and Inference to The Best Explanation
-Simple forms of inductive inference, like enumerative induction, are “horizontal”: from observations we infer observable facts. This makes enumerative induction very limited in use in inferring interesting scientific hypotheses
- Inference to the Best Explanation is “vertical”: from observation we infer something “deeper”, perhaps something that is unobservable, that explains them
- Inference to the Best Explanation is also called “abduction”
- Terminological warning: some authors take inference to the best explanation to be a form of induction, others take it to be a rival
How Does IBE Work?
- A distinction:
> An actual explanation of E is a true hypothesis that explains E
> A potential explanation of E is something that would explain E if it were true
- If we know that H is an actual explanation of E, we already know that H is true. So the question whether we should infer H does not arise
- But sometimes we know that H is a potential explanation of E without knowing whether H is true
- Inference to the best explanation would be more aptly called “inference to the best potential explanation”
- Schematically, IBE can be represented as follows: E/H
- This is not a deductively valid argument form, of course. Is it still a legitimate one?
What Makes an Explanation Good?
-Two different virtues of potential explanations:
> Being likely: having a high probability of being true
> Being lovely: being highly explanatory, providing a great deal of understanding, provided that it is true
- These virtues may come apart. The explanation that opium puts people to sleep because of its dormitive power is likely but not lovely
- Some conspiracy theories provide explanations that are lovely but not likely
- If IBE would make us infer the likeliest potential explanation, it would not be very helpful. An interesting version of IBE takes the best potential explanation to be the loveliest one
- Still, some restriction of the pool of potential explanations is needed— otherwise IBE might sanction inferring the truth of conspiracy theories. Before we infer to the best explanation, we need to decide which hypotheses we consider to be “live options”. Arguably, such “live options” should not be too unlikely. Roughly speaking, the phrase “the best potential explanation” can be understood as “the loveliest explanation that meets a certain threshold of likeliness”
Enumerative Induction and IBE
- Recall enumerative induction, the poremises are about what we have observed
- Consider a pool of potential explanations of the premises:
> All As are B
> As that are not B cannot be observed by us (Perhaps they are smaller than anything our eye can detect.)
> Someone has made sure that we do not come across any As that are not B
>All As are B up until the end of the year 2015
- Then enumerative induction may be justified by IBE, provided that the first one of these explanations is the best one
Objections To IBE
> Explanation has to stop somewhere, otherwise we are embarking on an infinite regress of explanations: “turtles all the way down”
> We cannot assume that everything that happens makes sense, in the sense of there being a good explanation for it. Maybe there are sometimes only bad explanations
> The best explanation we are considering may not be the best explanation there is. So using IBE may give us false beliefs even though the best explanation is in fact true
> “Beauty is in the eye of the beholder.” What the best or loveliest explanation is may be different for different theorists. So they may use IBE to infer different, incompatible hypotheses.
Is Observation Objective?
- Is observation an objective source of evidence? Of course, there are always skeptical worries (dreams, hallucinations, evil demons). But philosophers of science have found much less far-fetched reasons to question the objectivity of observations
- Experiments show that what people observe may depend on what they expect to see. It is not just a matter of light patterns arriving at the retina, but also of their background beliefs
- we know that effective observation needs to be guided by theory. But what theory someone holds—part of their background beliefs—may in turn influence what they observe. This is sometimes called the “theoryladenness of observation”
The Demarcation Problem Again
- Criteria proposed in the history of philosophy:
> Certainty / infallibility
> Specific methodology
> Falsifiability (Popper)
- In “Demise of the Demarcation Problem” Larry Laudan contends that they all fail. (Arguably, the Inference to the Best Explanation is not specific to scientific reasoning):
[W]e have managed to conflate two quite distinct questions: What makes a belief well founded (or heuristically fertile)? And what makes a belief scientific? The first set of questions is philosophically interesting and possibly even tractable; the second question is both uninteresting and, judging by its checkered past, intractable. If we would stand up and be counted on the side of reason, we ought to drop terms like ‘pseudo-science’ and ‘unscientific’ from our vocabulary; they are just hollow phrses which do only emotive work for us. (Larry Laudan, “The demise of the demarcation problem”, p. 125)
Feyerabend's Epistemological Anarchism
- While Laudan suggests that the distinction between scientific and nonscientific belief formation is unimportant, he still thinks that belief formation can be assessed by epistemological criteria (well-founded or not, supported by substantial evidence or not)
- Feyerabend is more radical: no such epistemological criteria should be held to be generally applicable. His epistemological anarchism rejects all general methodological rules and principles:
[M]y thesis is that anarchism helps to achieve progress. ... Even a law-and-order science will succeed only if anarchistic moves are occasionally allowed to take place. [T]he idea of a fixed method, or a fixed theory of rationality, rests on too naive a view of man and his social surroundings. To those who look at the rich material provided by history, and who are not intent on impoverishing it in order to please their lower instincts, their craving for intellectual security in the form of clarity, precision, ‘objectivity’, ‘truth’, it will become clear that there is only principle that can be defended under all circumstances and in all stages of human development. It is the principle: anything goes. (Feyerabend, Against Method, pp. 18- 19)
Against Philosophical Accounts of Scientific Metho
- Examples of rules that should occasionally be broken:
> Avoid logical inconsistency in your theory!
> Reject theories that are in conflict with observations!
> Avoid ad-hoc hypotheses to save theories!
> Prefer theories with better empirical support!
“[T]heories cannot be justified and their excellence cannot be shown without reference to other theories.” (“How to Defend Society against Science”, p. 159)
Against the Pretensions of Science
- Feyerabend does not merely think that philosophers’ accounts of science are wrong. He also thinks that the prestige and authority of science is not fully deserved
- First, the results of science are not as great as is sometimes claimed: “the competence, the complications and the successes of science are vastly exaggerated” (p. 162). Example: forensic science
- Second, non-scientific practices also get results, and the credit often wrongly goes to science alone. Examples of non-scientific influences on putative successes of science:
> Philolaos (pythagorean) on Copernicus
> Artisans on mechanics and optics
>Midwives and witches on medicine and science has no monopoly on knowledge
Science as an Ideology
- Even if science were unrivalled in giving us knowledge—which Feyerabend disputes— we should still not treat with as much respect and deference as we do. For knowledge is just one valuable thing among others:
Human life is guided by many ideas. Truth is one of them. Freedom and mental independence are others. If Truth, as conceived by some ideologists, conflicts with freedom then we have a choice. We may abandon freedom. But we may also abandon truth. (p. 158)
- Feyerabend values freedom above truth. “My criticism of modern science is that it inhibits freedom of thought.” (p. 158)
- He does not think that there is anything wrong with science as such, only with the role it is allowed to play in our society
- In the society he envisages: ... many will become scientists ... without having been taken in by the ideology of science, they will be scientists because they have made a free choice. ... [S]cientists will not play any predominant role in the society I envisage. They will be more than balanced by magicians, or priests, or astrologers. (pp. 164-5)
Objections to Feyerabend
> Regarding methodology: he may be right that no single methodological maxim should be sacrosanct at all times. It does not follow that “anything goes”, that every epistemic move is permissible at every time
> Regarding results: he understates the extent to which science has been successful, and produced results
> Regarding rivals: he is right that philosphers of science are usually unfairly dismissive of so-called “pseudo-sciences”. But he makes no detailed study of them either, so the claim that voodoo can rival science is left unsubstantiated
> Regarding freedom: Feyerabend’s view of science has been criticized as resting on a naive conception of freedom as the mere absence of constraints