Sub Prob Does Not Exist

This version extracts the Anti-Bayesian Arguments and Isolates them into a separate paper.

The OTHER PART REMAINS to be written up separately on Models as Metaphors and a NEW Definition of Probabilty = for this other part, use my own notes on models and reality, and also Kay's article on The Map is not the Territory.

SSRN CURRENT VERSION: Subjective Probability does not Exist --- Submission History

UPDATE on 25 Feb 2019 - rejected by E&P.

23-Feb-2020

Dear Dr. Zaman,

I am writing to you with regard to manuscript # EAP-AR-2019-0160 entitled "Subjective Probability Does Not Exist" which you submitted to Economics and Philosophy.

I regret to inform you that your manuscript has been rejected as it is not suitable for publication in Economics and Philosophy. I read your manuscript with interest, but ultimately found many of your core arguments wanting. Just to give some examples: first, I was not convinced by the contrast between objective and subjective probability that you assume in his argument, claiming that by showing ontic probability to exist, the notion of subjective probability becomes meaningless. This flies in the face of ecomenical accounts, e.g. Lewis principal principle. Second, you posit a close connection between certain forms of positivism and subjective probability, and then claim that by rejecting the positivist position, you also reject subjective probability. That does not follow. Third, some of the arguments against positivism seem to be based on notions of non-fallible introspection of knowledge and preference, which strike me as implausibly strong assumptions.

Thank you for considering Economics and Philosophy for the publication of your research. I hope the outcome of this specific submission will not discourage you from the submission of future manuscripts.

Yours sincerely,

Dr. Till Grüne-Yanoff

Editor, Economics and Philosophy

You do many things in one paper. They include:

Giving a taxonomy of interpretations of probability.

Comments on the distinction between risk and uncertainty and why it is important.

Providing arguments that limiting relative frequency cannot provide a foundation for an interpretation of probability.

Exegeses of many important figures, such as de Finetti.

Comments on the history of ideas. You give an account of how different ideas have shaped the course of philosophy and economics over the past 100 years.

Criticising a version of the subjectivist account of probability.

Criticising positivism.

…

You might consider turning this into more than one paper.

In the philosophical literature, the notion of subjective probability has been developed slightly differently from the picture you paint of a positivist Ramsey-de-Finetti-Savage account. I will not comment on the influence it has had in economics, as this is not my area of expertise. But let me explain how the problems that you raise nonetheless arise for philosophers and how contemporary subjective Bayesian philosophers think about them. I hope these reflections may help make your paper more influential in these circles.

Many subjectivist philosophers would want to add a fourth kind of probability to the three you list. They would call this credence, referring to an intermediate state of belief. This would not be an intermediate state of knowledge, in that a credence could be based on error or fantasy. So credence is distinct from your epistemic probability. Credence also need not be revealed in choices or actions (especially in cases of forced choice, for just the reasons you suggest). So, it need not be accessible to an outside observer and is distinct from your revealed probability. It seems plausible that this kind of epistemic mental state exists and is an attitude that agents form to some propositions.

With mental states of credences distinguished, your point about risk and uncertainty might be put as follows: The most important feature of epistemic mental states is that they ought to reflect an agent's epistemic situation. But credences are not able to reflect the epistemic situation of agents with respect to many propositions. Credences can only properly reflect the epistemic situation of agents under risk, not under mild or radical uncertainty. As a result, agents will not have credences about many propositions and in fact they ought not to. Some philosophers acknowledge this point and consider broader kinds of epistemic representation, such as interval-valued credences. But even these broader representational states may not be able to reflect all agents' states of uncertainty. It seems to be a mistake to insist that agents do, or ought to, have a particular kind of epistemic representation that cannot reflect their situation.

Nonetheless, the notion of subjective probability is very useful in that it allows us to be clear about certain kinds of norms on epistemic states - for example, a norm that one's credences ought to be able to be extended to at least one complete probability function over all propositions.(Note that this is often phrased as a "coherent extendability" norm to avoid the assumption or norm of completeness. See e.g. Joyce 1999 p102-3. De Finetti spoke this way, too. This allows gappy or imprecise credences. The approach of using credences to state norms has proved fruitful even without a complete account of objective probabilities. For another example, Lewis' Principal Principle (Lewis 1980) might be construed as a norm for how credences ought to align with objective chances (ontic probabilities), even if we cannot say exactly what objective chances are. Within the subjectivist framework, these kinds of norms provide guides to epistemic representations (and hence guides to life). Introspection of credences without constraints is fruitless, just as you say.

Your arguments seem only to go as far as showing that people do not always have a definite subjective probability (degree of belief) in every proposition. This seems right. But that does not show that subjective probabilities do not exist. (Anymore that the fact that people suspend judgement on some propositions shows that belief and disbelief do not exist.)

Given all of this, I might temper your conclusion that "Subjective probabilities do not exist". Rather, subjective probabilities exist, but they may not be revealed by choices, agents may not hold this kind of attitude towards all propositions, and they cannot be a good guide to life unless constrained by substantive norms.

I’m also attaching a few further comments, inserted into your text. I hope they are helpful!

20-Dec-2019

10/20/19 - Revise to SELL better - Incorporate Kay and King about Embrace Radical Uncertainty

Explain three obstacles - What is single case probability, what does it mean to "Know" probability of an event, and how to differentiate between "knowing probability" and "acting as if you know" the probability.

Submitted to Metroeconomica on 26th June 2018

Submitted 5Jun18 The British Journal for Philosophy of Science: https://mc.manuscriptcentral.com/thebjps Username MIT+eorS

15/12/17 Submitted to: Journal of Econometric Method: https://mc.manuscriptcentral.com/dgjem Username MIT+eolj

https://mail.google.com/mail/u/1/#inbox/16059bbfb7a7ce58 acknowledgement of submission.

Submitted to Games & Economic Behavior -

Submitted to Statistical Science --username asadzaman+e https://www.e-publications.org/ims/submission/STS/

As Editor, I have read your submission carefully. However, I regret to tell you that I have decided not to send it for formal review. You can find our editorial policy described in some detail at http://imstat.org/sts/policy.html, where you will find that our principal role is as a statistics review journal, and "A high standard of scholarship is expected, and ideally an intellectually rigorous analysis that helps to set the scene for future work in the area." You paper asserts that the assignment of (subjective) probability to an urn with unknown composition of balls of different colors cannot be always consistent with preference to balls of a certain given color. This point is valid, philosophical but also trivial. From my perspective, your paper is unfortunately not the kind of broad scholarly review that we publish.

Cun-Hui Zhang

Editor of Statistical Science

Things to add to next veryuson

ONE: Distinguish betwee all models are false and current paperl A: Metaphors are OBVIOUSLY and PATENLY false as literal truths, so this is not a revalation, It is a revelation to people who confuse fingers with the moon or mix model with reality to asseret that the two are not the same. HOWEVER, this still leave the question of HOW and WHY can false models be useful. Having discovered that the finger is not the mooon, one still needs to learn what the finger is pointing to . Our metaphor model provides a clear and precise ARTICULATION of how models are separeate and also precisely hwo a false model can nonewtheless be useful.

TWO: An explikct aritucualte of how Bayesian probability is a different TYHPE of object fropm conventional and we need differfent names as well as difernt updating mecahnings. RECONSIDER BAYES RULE (as we have done,wioth TIMING) and then look at Bayes Prob as a qunatifiable measure of EVIDENCE. Looup belief ufntinds Surprise functions etc.