Thomas Kuhn, answered from the texts and cited to the page.
Theory choice in mature science is rational but not algorithmic. There is no rule, no procedure, no neutral calculus that takes the evidence as input and yields the choice as output. What there is, and what scientists actually use, is a shared set of values: accuracy (the theory's empirical fit with experiment and observation), consistency (internal logical coherence and consistency with other accepted theories), scope (the range of phenomena the theory addresses beyond those it was designed to explain), simplicity (the theory's clarity and economy of structure), and fruitfulness (the theory's capacity to suggest new lines of inquiry, new predictions, new applications).
These five values are widely shared across the scientific community across periods.12 They are also, characteristically, in tension with one another: a theory may be more accurate but less simple, more fruitful but less consistent with the rest of accepted science, broader in scope but with poorer empirical fit in central cases. There is no neutral way to weight them.
Different scientists, working from different parts of the field, with different research interests, with different temperaments and trainings, will weight them differently — and this is not a defect to be remedied by sharper philosophical analysis but a structural feature of how scientific judgment works under conditions of evidential underdetermination.345
The areas I have in mind are: value invariance, subjectivity, and partial communication. If my views of scientific development are novel-a matter about which there is legitimate room for doubt-it is in areas such as these, rather than theory choice, that my main departures from tradition should be sought.The Essential Tension, p. 355
If they did not hold values like these, their disciplines would develop very differently. Note, for example, that the periods in which the history of art was a history of progress were also the periods in which the artist's aim was accuracy of representation.The Road Since Structure, pp. 164–165
In the absence of criteria able to dictate the choice of each individual, I argued, 1. The Structure of Scientific Revolutions, 2d ed. (Chicago, 1970), pp. 148, 151-52, 159. All the passages from which these fragments are taken appeared in the same form in the first edition, published in 1962.The Essential Tension, p. 341
After that concession he has no recourse, and his opponent's proof is then compelling. Only if the two discover instead that they differ about the meaning or application of stipulated rules, that their prior agreement provides no sufficient basis for proof, does the debate continue in the form it inevitably takes during scientific revolutions.The Structure of Scientific Revolutions, pp. 159–160
Though they function at all times, their particular importance emerges when the members of a particular community must identify crisis or, later, choose between incompatible ways of practicing their discipline.The Structure of Scientific Revolutions, p. 149
The mature view is in 'Objectivity, Value Judgment, and Theory Choice' (1973). The earlier rhetoric in Structure — including the analogy between paradigm change and religious conversion — was rhetorically heavier than the position required, and I refined it. The 1973 paper makes the position philosophically explicit: theory choice involves values, the values are rational, the values legitimately conflict, and the conflict is not a problem to be dissolved. Feyerabend's slogan anything goes misreads the situation as license; the actual situation is that not just anything goes, the values are shared, the standards are real, but algorithm and unique-correct-answer are not what the rationality of science consists in.