Maybe “personal” is a better word than “subjective”. All probabilities are conditional, and each individual has different knowledge to condition on.
A problem with de Finetti is that the only rational aspect of probability is coherence — otherwise they can be completely arbitrary with no connection with the real world. This is the negative connotation to “subjective”. But this is a reductionist argument. Probability judgments don’t start from nothing. Every individual has a wealth of experience on which to condition his or her probability. Coherence requires using Bayes’ rule to incorporate experience into the “subjective” probability. Real human minds don’t actually do that, but the ideal is that that is how one ought to think. It requires total amnesia for de Finetti’s subjective probabilities to be free of empirical knowledge.
Assigning probabilities based on symmetry (coins, cards, dice, etc.) — on other words, probabilities as discussed in elementary (and advanced) courses are not empirical, but can be verified empirically. They are still essentially subjective — based on reasoning, but hopefully a subjectivity on which reasonable people will agree.
Mik Bickis
On Feb 19, 2021, at 12:06 PM, Jay Kadane <kadane(a)andrew.cmu.edu> wrote:
CAUTION: External to USask. Verify sender and use caution with links and attachments. Forward suspicious emails to phishing(a)usask.ca
Dear Serena, I think that our assignment of probabilities to the outcome of flipping a fair coin stem from the definition of "fair", and hence are neither subjective or frequentistic. Jay
On 2/19/21 12:35 PM, Serena Doria wrote:
Dear Professor Kadane and dear all,
thank you, all comments were for me very interesting and stimulating!
From a mathematical point of view, it seems to me that the crucial point, in the probabilistic field, is not the different role and meaning of the words "objective" and "subjective" but between the procedures used in the subjectivist approach (based on the notion of coherent bet) and in the axiomatic approach (based on Kolmogorov axioms). In both cases we assign 1/2 head and 1/2 cross in the toss of a fair coin. One of the differences could be that in the subjectivist approach such assignment can be interpreted as that of a person betting on the given event in a coherent bet. Another important difference could be that a "coherent" person who knows the result \omega of a random experiment assess probability 1 to any event containing \omega and 0 to any event which does not contain \omega.
Best regards, Serena
Jay Kadane <kadane(a)andrew.cmu.edu> ha scritto:
Dear Serena, Thank you for this clarification of de Finetti's cultural surroundings. I've heard it said that the data are the data, and everything else we made up. (Sometimes it turns out that what one thought were data , turn out not be the data, but that's another issue). We may have better or worse reasons for the model, prior, etc. used in an analysis, and to be persuasive, it is necessary to explain the thought process behind the choices made. Sometimes the word "objective" is used to try to intimidate people from challenging the choices made. All the best, Jay Kadane
On 2/18/21 3:36 AM, Serena Doria wrote:
Dear all
the term "subjective" for probability was introduced by Bruno de Finetti, and to understand the motivation it is necessary to understand the cultural environment in Italy in that period. In 1934 Luigi Pirandello received the Nobel Prize in Literature and his vision of life and knowledge affirmed that there was no 'single truth' but reality was what the individual subject perceived. And this is also evident from some of the best-known titles of his plays ( Cosi e' se vi pare, Uno nessuno centomila). de Finetti grows in this cultural context but gives to the term "subjective" a value more rigorous and "objective" through the concept of coherent betting.
Personally, I do not believe that the term subjective is to be understood in a negative way because it highlights how the knowledge an individual has about a phenomenon depends on the information the subject has.
For this reason, in the subjectivist approach to probability, the concept of conditional probability and conditioning events are fundamental to represent respectively partial knowledge and different information that individuals have.
Best regards,
Serena Doria
"Kreinovich, Vladik" <vladik(a)utep.edu> ha scritto:
Dear Friends,
We often talk and write about objective and subjective probabilities, about objective and subjective measures of uncertainty. However, at a recent conference on uncertainty, Yakov Ben-Haim made an important observation -- based on his experience of working on applications with colleagues from many different areas.
His experience is that in many application areas, the word "subjective" has a negative connotation: it means unjustified estimates based on gut feeling only, prone to bias and wild variations.
Such gut-feeling-based estimations sometimes happen, but mostly, when we talk about "subjective", we mean judgmental estimates, estimates which are not just coming out of gut feeling, but which can be usually provided with some justification. For example, if we estimate to what extent someone is young (one of Zadeh's original examples) we can usually explain the degree we assign to "youngness" of an individual by referring to features which are present and which are typical young age - and features of this individual which are more typical for mature-age folks.
For example, subjective probability often means simply probability that is not coming from the analysis of frequency, but from expert estimates.
Yakov's recommendations is to use words like "judgmental" (or "expert-based") instead of "subjective" in such situations, especially when working on applications - and applications are the main goal of uncertainty studies in the first place.
SIPTA mailing list -- sipta(a)lists.idsia.ch To unsubscribe send an email to sipta-leave(a)lists.idsia.ch
SIPTA mailing list -- sipta(a)lists.idsia.ch To unsubscribe send an email to sipta-leave(a)lists.idsia.ch
SIPTA mailing list -- sipta(a)lists.idsia.ch To unsubscribe send an email to sipta-leave(a)lists.idsia.ch