Overfitting and heuristics in philosophy

• Verfasst von: Williamson, Timothy [VerfasserIn]
• Titel: Overfitting and heuristics in philosophy
• Verf.angabe: Timothy Williamson
• Verlagsort: New York, NY
• Verlag: Oxford University Press
• E-Jahr: 2024
• Jahr: [2024]
• Umfang: 1 Online-Ressource
• Gesamttitel/Reihe: The Rutgers lectures in philosophy
• Gesamttitel/Reihe: Oxford scholarship online
• Fussnoten: Includes bibliographical references and index. - Description based on online resource and publisher information; title from PDF title page (viewed on April 8, 2024)
• ISBN: 978-0-19-777924-8
• DOI: doi:10.1093/oso/9780197779217.001.0001
• Datenträger: Online-Ressource
• Sprache: eng
• Bibliogr. Hinweis: Erscheint auch als : Druck-Ausgabe
• Sach-SW: Philosophy
• Sach-SW: Philosophy: epistemology & theory of knowledge
• K10plus-PPN: 1904981712
• K10plus-PPN:
  Universitätsbibliothek
• Lokale URL UB: Zum Volltext

Abstract

In his Rutgers Lectures, Timothy Williamson explains how contemporary philosophy suffers from a widespread pathology known as overfitting to natural and social scientists, but little understood by most philosophers. Overfitting involves an insufficiently critical attitude towards data, which leads to over-complicated theories designed to fit what are in fact errors in the data. In philosophy, the data typically comprise verdicts on hypothetical or actual cases. Errors in such data can result from our reliance on heuristics, efficient cognitive shortcuts, simple to use but not fully reliable. Just as heuristics embedded in our visual system produce visual illusions, so heuristics embedded in our general cognitive systems produce philosophical paradoxes.

Abstract

"The main aim of this book is to encourage philosophers to take a more sophisticated and scientific attitude to their handling of evidence, both in theory and in practice, by introducing two categories neglected in current metaphilosophy. The first category is that of heuristics. These are typically efficient ways of solving problems of some kind, quick and easy to use, and mostly but not always reliable. Those most probably central to philosophical methodology are more or less humanly universal general cognitive heuristics which we employ without conscious reflection. In many plausible cases, they can be shown to be implicitly inconsistent, and so cannot be perfectly valid, though for evolutionary reasons they are likely to be fairly reliable in normal cases. Such heuristics may well generate false but convincing data as applied to some actual or hypothetical. Arguably, the heuristics also generate philosophical paradoxes, just as heuristics embedded in our perceptual systems generate perceptual illusions. The second category is that of overfitting. This is a recognized pathology in science, when models are allowed to become very complicated to achieve a close fit with data, and further complications typically keep having to be made as new data comes in. When the data contains errors, such theorizing provides little insight. Scientists guard against overfitting by strongly preferring comparatively simple models. Philosophers should learn to do the same. This approach is applied to debates between coarse-grained (intensionalist) and fine-grained (hyperintensionalist) theories of metaphysics, ascriptions of propositional attitudes, and other topics, in favour of intensionalism"--