Noise: A Flaw in Human Judgment
[FULL TEXT READ, the complete published PDF, open access under CC BY-SA 4.0] Szreder, M. (2022). 'Noise and bias, some controversies raised by the book Noise: A Flaw in Human Judgment.' Przeglad Statystyczny, 69(1), 39-49. doi:10.5604/01.3001.0015.8792. THE STRONGEST PUBLISHED CRITICISM AND IT LANDS ON EXACTLY THE LOSS FUNCTION THE BRIEF FLAGGED. He concedes the arithmetic and attacks the inference. From his abstract: he agrees that noise and bias contribute equally to mean squared error, but argues that in some circumstances reducing noise can make accurate inference harder rather than easier, and that the actual impact of noise cannot be determined without considering bias and noise together. The argument: hold bias constant and shrink the dispersion, and MSE falls exactly as the book says, but the probability that any given judgement lands close to the true value also falls, because the tightened distribution is centred on the wrong place. He argues that if bias is present in judgments, reducing noise makes it less likely that the true value will be discovered, because biased judgments absorb an increasingly large share of the probability. On the book's handling: he says the authors confine themselves to saying that noise reduction would be less of a priority if bias were much larger than noise, finds it hard to accept that this is only a question of priorities, and warns that stressing that bias and noise contribute equally to mean squared error, which is true, while paying little attention to how the two relate, may mislead. SECOND CRITICISM, bias should stay central for two reasons: it is what produces discrimination by race, religion, gender or wage, which noise does not; and bias differs from noise in a second respect.
[FULL TEXT READ, the complete published PDF, open access under CC BY-SA 4.0] Szreder, M. (2022). 'Noise and bias, some controversies raised by the book Noise: A Flaw in Human Judgment.' Przeglad Statystyczny, 69(1), 39-49. doi:10.5604/01.3001.0015.8792. THE STRONGEST PUBLISHED CRITICISM AND IT LANDS ON EXACTLY THE LOSS FUNCTION THE BRIEF FLAGGED. He concedes the arithmetic and attacks the inference. From his abstract: he agrees that noise and bias contribute equally to mean squared error, but argues that in some circumstances reducing noise can make accurate inference harder rather than easier, and that the actual impact of noise cannot be determined without considering bias and noise together. The argument: hold bias constant and shrink the dispersion, and MSE falls exactly as the book says, but the probability that any given judgement lands close to the true value also falls, because the tightened distribution is centred on the wrong place. He argues that if bias is present in judgments, reducing noise makes it less likely that the true value will be discovered, because biased judgments absorb an increasingly large share of the probability. On the book's handling: he says the authors confine themselves to saying that noise reduction would be less of a priority if bias were much larger than noise, finds it hard to accept that this is only a question of priorities, and warns that stressing that bias and noise contribute equally to mean squared error, which is true, while paying little attention to how the two relate, may mislead. SECOND CRITICISM, bias should stay central for two reasons: it is what produces discrimination by race, religion, gender or wage, which noise does not; and bias differs from noise in a second respect.
Written from the fULL TEXT READ, the complete published PDF, open access under CC BY-SA 4.0.
What our sources record
Recorded when this source was checked, figures included, with page references and quotation marks removed.
Szreder, M. (2022). Noise and bias, some controversies raised by the book Noise: A Flaw in Human Judgment. Przeglad Statystyczny, 69(1), 39-49. doi:10.5604/01.3001.0015.8792