Falsification

structure

intro → Hume problem of induction and induc / conv. vs Popper → falsification positive case → why certain conjectures → Duhem-Quine problem → Salmon’s rational prediction problem → conc / lakatos

Intro

• Science appears to rely on induction, generalising from observed instances to universal laws

• But Hume shows that inductive inference cannot be rationally justified…Popper falsificationism is the most ambitious attempt to reconstruct scientific method without induction - replacing confirmation with falsification and verification with corroboration

• Rough thesis - Popper correctly diagnoses the problem of induction and his falsificationist criterion captures something genuinely important about scientific practice - bold conjectures and severe tests. But falsificationism faces two decisive objections: the Duhem-Quine problem shows individual theories cannot be definitively falsified, and Salmon demonstrates that without induction Popper cannot account for rational prediction. These failures reveal that falsificationism solves the problem of induction only by leaving science unable to do what we need it to do

problem being solved / Hume prob of induc

• Inductive inference passes from singular observation statements to universal conclusions

  • ‘This swan is white’ to ‘all swans are white’

  • No matter how many white swans we observe, this cannot logically justify the universal conclusion as any conclusion drawn this way may always turn out false

• Two horns of the problem

  • If we try to justify induction by appeal to experience we face infinite regress of justifying principle of induction based on instances of the success of induction etc.

  • If we try to justify induction deductively by arguing that ‘if induction is justified on many occasions, then it is justified in general…’, this is circular as we are using induction in the first premise to justify induction

  • assigning probability to statements based on inductive inferences can only be justified by invoking another principle of induction…inductive reasoning always causes regress

inductivism / conventionalism / popper’s approach

• The obvious response to the problem of induction is inductivism — simply accept induction as the method of science and generalise from observations. But inductivism cannot justify the principle of induction without circularity and has no internal explanation of why scientists choose which facts to investigate or why scientific errors occur so frequently. 

• Conventionalism — the view that theories are frameworks adopted by community agreement — sidesteps the problem but produces a code of honour Lakatos calls dishonourable: with no rigorous principle distinguishing legitimate scientific frameworks from pseudoscience anything can count as good science.

• Popper’s approach

  • Reject inductive reasoning, and instead hold that the only scientific statements are statements of falsification

  • Rather than asking ‘what confirms a theory’, we should ask ‘what would falsify a theory’

  • A theory is only scientific if and only if it is falsifiable…it makes predictions that can, in principle, be shown false by observation

positive case for falsification

• Demarcation criterion - a theory is genuinely scientific iff it is falsifiable; it forbids certain possible observations and would be refuted if those observations occurred

  • Einstein’s general relativity predicts that light bends near massive objects, which is a specific prediction that could in principle be false and was tested by eclipse observation…paradigm example disproving Newtonian theory of gravity

  • Freudian psychoanalysis has a justification to explain any possible human behaviour…they make no predictions that could be falsified and as a result they are unfalsifiable and unscientific because they cannot in principle be shown to be wrong…say nothing informative about the world

• Corroboration vs confirmation - Popper says a theory that survives tests without being falsified is not confirmed or verified as this would be inductive reasoning

  • Instead, a theory is corroborated by current testing and evidence - corroboration is purely a report on past performance

  • Science proceeds by bold conjecture and attempted refutation…goal is not to accumulate confirming instances but to expose theories to the most severe tests possible in the hopes of falsifying it and achieving a development

  • A theory that survives tests is preferred because it has not been shown to be false, not because it is true

• Code of honour ( key methodological norm of the framework) is novelty

  • Conjectures should not be ad hod modifications designed to avoid falsification, but instead ideas that are genuinely new

  • Adding an epicycle to save Ptolemaic astronomy from anomaly is poor science as it does not make any independently testable predictions

  • Einstein’s prediction that light may bend new the sun is good science as it is a bold new claim about the world that goes beyond the evidence motivating it

obj - why certain conjectures

• Popper struggles to explain why scientists make certain conjectures rather than others… his view is compatible with the social need hypothesis (scientists choose what to study on the basis of social factors) but Popper would hate this implication…if scientists choose what to conjecture based on funding pressures, cultural assumptions and political priorities, the starting point of scientific inquiry is not rationally determined and not objective

• Falsification cannot give a complete account of scientific rationality if he cannot explain where conjectures come from

• RESPONSE - logic of knowledge is separate from psychology of knowledge…how scientists may come up with conjectures is an empirical psychological question and what instead matters for the logic of science is how theories are tested… the act of inventing a theory involves creative intuition and therefore outside the logic of science entirely - this does not threaten the logical structure of falsification as the conjecture (whether irrational or social) can be tested

  • But if there is no rational account of conjecture generation and they are influenced via social factors, he cannot explain the entire process of scientific progress with reason and the line between science and pseudo-science is slightly blurred

obj - Duhem-Quine problem - definitive falsification impossible

• Popper requires that a negative test result definitively falsifies a theory…but as Duhem and Quine showed, theories face a tribunal of experience not individually but as a web…when a prediction fails, how can we know which part of the web to falsify?

• Take an experiment like boiling water - if the thermometer reads 99 degrees rather than 100, why should we conclude the hypo of boiling point is false when it could be that the thermometer is inaccurate, the altitude is affecting the result, impurities are present or the room pressure is different…these auxiliary assumptions could be revised instead of the core theory

• Scientists can always save a theory from complete falsification by modifying auxiliary hypotheses rather than the core claim…so under this, adding epicycles or slightly tweaking Newton’s theory is allowed

• RESPONSE - ad hoc modifications are methodologically illegitimate… they do not have the novelty that is part of the code of honour…

  • But then what counts as a legitimate auxiliary adjustment vs an illegitimate ad hoc modification? This happens all the time; Neptune was discovered via a postulation of an unobserved planet to save Newton’s laws when they predicted Uranus’ orbit wrong…Popper says this is a legitimate auxiliary adjustment since it made a specific prediction about where to look with testable predictions, but the same logical structure could be used to postpone falsification of a false theory ‘this is not ad hoc; i’m predicting unknown factor x that we will eventually detect’ which could go on infinitely…

• This is more of an issue than conjecture obj bc it strikes at core of what falsification requires; the ad hoc adjustment issue is not clearly defined enough for Popper

obj - Salmon’s rational prediction objection

• Popper claims that corroboration (surviving severe testing) provides a rational basis for preferring one theory over another for practical prediction..but Popper also explicitly says corroboration statements have no bearing on future events and they are purely based on past performance / say nothing about future performance

• This creates a tension illustrated by the balloon example…the friendly physicist predicts the balloon will move forward when the plane accelerates…his prediction is based on physical theory corroborated by past tests…when he wins the bet, what rational basis did he have for making the prediction on Popper’s account?

• If corroboration has no bearing on future events preferring a well-corroborated theory gives no rational grounds for expecting it to perform well in the future

• Salmon’s dilemma - either Popper’s science embodies essential inductive aspects (past testing performance is a guide to future reliability) or science on Popper’s account lacks genuine predictive content

• RESPONSE - we should prefer best tested theory because there is nothing more rational than well-conducted critical discussion

  • Reply - question-begging as saying the best tested theory is the most rational choice for prediction assumes that past testing performance is a guide to future predictive success, which is the inductive assumption Popper claims to have eliminated

  • If corroboration has no relevance to future predictions then preferring the best tested theory for practical action is no more rational than flipping a coin

• Implication - Popper’s position may ultimately rest on a covert commitment to scientific realism (belief that the world has regularities making past testing performance a reliable guide to future behaviour); so Popper’s rejection of induction depends on a metaphysical assumption he has not justified and his epistemology is not as far from inductivism as claimed

• DECISIVE OBJ shows falsificationism solves problem of induction only by leaving science unable to account for its own predictive success…Popper’s response to prefer best tested theory because nothing is more rational either tacitly assumes inductive principles or provides no rational basis for prediction at all

Lakatos’ four theory points

• failure of falsificationism does not mean Popper's core insight was wrong; the emphasis on bold conjectures and severe tests captures something genuinely important about scientific practice

• Lakatos's methodology of scientific research programmes preserves this insight while answering Duhem-Quine

• his account distinguishes the hard core of a research programme from its protective belt of auxiliary hypotheses, and evaluating programmes as progressive when modifications generate novel predictions or degenerative when modifications are purely ad hoc

• This gives the principled distinction between legitimate auxiliary adjustment and illegitimate ad hoc manoeuvre that Popper lacked, while acknowledging that individual falsification is impossible

conc

• Demarcation criterion captures difference between theories that risk falsification and those that can accommodate any observation

• Emphasis on bold conjectures and severe tests as hallmarks of good science is real contribution…cautious accumulation does not advance science

• But falsification as a complete philosophy of science fails due to Salmon dilemma for rational prediction and Duhem-Quine problem

• Most defensible position for the view acknowledges demarcation but recognises that scientific practice depends on inductive elements as a feature of how evidence supports theories