How objectively rational is it to believe that God exists or that the sun will rise tomorrow or that my girlfriend will cheat on me?

I understand that the concept of credence and subjective probability exists but that just tells you how you feel about a certain matter, not necessarily how the world actually is. There is no direct logical principle connecting your credence to reality.

How does one then, similar to credence, come up with a way to objectively figure out how (objectively) rational it is to believe in something? Or is this impossible?

  • 2
    There's an SEP article on the logic of belief revision that might be relevant. C.f. dynamic and static epistemic logic, or even justification logic. Aug 12, 2023 at 0:52
  • there is probably some kind of consensus that some things, like infidelity, are more likely than others. how would you respond to waking up one day and seeing all your thoughts being broadcast via telepathy to everyone around you? if your answer would be to work out how possible it is via what you have learnt about the world, rather than consensual reality, then you would likely go insane
    – user67155
    Aug 12, 2023 at 1:03
  • 1
    I don't think there's a standard definition that allows for quantification beyond "rational" or "not rational", but if you wanted one, you could subtract your subjective credence from the bet that would be correct given your priors. I.E. assuming perfect play, believing that you will break even gambling at a casino game could be assigned a "rationality rating" of 1 minus the house edge.
    – g s
    Aug 12, 2023 at 1:30
  • That's the way to approach the problem, si! How about a quote, from days past? "Better to be silent and thought of as a fool than to open your mouth and remove all doubt". Aug 12, 2023 at 2:04
  • 🤔 ... We're not stones, of that much I'm certain. There are, unfortunately for me, at least 35 things nonstones do. I'm a guesstimator and looks like I'm not exactly in jannat/swarga. Aug 12, 2023 at 5:40

4 Answers 4


From the scientific perspective, rational belief is linked to observable evidence and reproducible experiments. For example, we believe the sun will rise tomorrow based on our understanding of the Earth's rotation and consistent observations over millennia.

But things get trickier when we move away from physical phenomena to more abstract or personal ones. For instance, predicting personal behavior, such as whether your girlfriend will cheat on you, is a lot more challenging. This is because humans are influenced by a lots of factors, many of which aren't easily quantifiable or predictable. In such a case, you may rely on trust and your understanding of her character to form your belief.

Regarding the existence of God, this is a deeply personal thing that falls outside the realm of scientific inquiry. People's belief or disbelief in God is usually based on their personal experiences, upbringing, cultural context, and philosophical inclinations.

So you give three very different examples.

I'd also add that rational belief requires openness to changing one's mind as new evidence contradicts prior beliefs. The less willing someone is to recognize they could be wrong, the less rational their thought process.


There probably isn't an objective way to measure rationality.

So can we all agree on a general consensus to come to the truth? We know that wouldn't work.

But what we can do is, to work onto reducing the biases that arise from the subjective nature of rationality. How can we do that? Well, getting a "lot" of people to discuss and argue on a certain topic, some would come up with ways of why it can be rational while some would do the inverse. The best way to obtain truth is not just to try and figure out why something might work, but why something might not.

This literally is how the scientific method works.


One obstacle to (or complication with) the idea of quantifying rationality is the question of whether the word "rationality" (and its cognates) is a mass noun or a count noun. Does it apply all-at-once-or-not, to a given set of beliefs, or can it be partially smeared over/worked into a belief?

Or imagine a belief that is implicitly a conjunction of 100 "smaller" beliefs. If one has direct evidence for 50 of those, indirect evidence for 25, no evidence either way for 10, indirect counterevidence for 10 more, and direct counterevidence for the remaining 5, one might think of describing the integrated 100-fold set as being rational as a percentage of the total (though how one would compute this versus the ratio of direct/indirect (counter)evidence for the subsets, I don't know).

On the other hand, that is a fanciful hypothetical example and unlikely to come up "in practice." There are some theists who try to "average" the justification from different arguments for belief in God (Mitchell and Swinburne come to mind, though I'd have to triple-check to make sure they're the major examples I have in mind...), which might seem impious; and Kant spent a little bit of time arguing against such a form of reasoning:

For the assertion that the reality of such ideas [such as God] is probable is as absurd as a proof of the probability of a proposition in geometry. Pure abstract reason, apart from all experience, can either cognize nothing at all; and hence the judgements it enounces are never mere opinions, they are either apodeictic certainties, or declarations that nothing can be known on the subject. Opinions and probable judgements on the nature of things can only be employed to explain given phenomena, or they may relate to the effect, in accordance with empirical laws, of an actually existing cause. In other words, we must restrict the sphere of opinion to the world of experience and nature. Beyond this region opinion is mere invention; unless we are groping about for the truth on a path not yet fully known, and have some hopes of stumbling upon it by chance.

  1. Find a very successful system of updating beliefs. This would be a system that, if you followed it, would help you act effectively to achieve the outcomes you desire. The system need not be particularly easy to follow - just that if you were able to follow it, you would form beliefs that help you act effectively.
  2. Ask what this system of updating beliefs would dictate that you believe.

One such idealized system of updating beliefs is ideal Bayesian inference with a minimum-description-length prior. In this system, you should hold the set of probabilities that you would hold, if you were capable of ideal Bayesian inference with a minimum-description-length prior. (Which, by the way, no one is, but it is still useful to think in terms of approximations to it.)

We could also imagine other systems of updating beliefs - logical inference in ZFC is one, though it is silent on the majority of real-world issues. Or we might think of various versions of computation-time-limited Bayesian inference.

One thing common to all such systems is that we should avoid inconsistency in the probabilities we assign. For example, the probabilities of an exhaustive set of mutually exclusive outcomes should sum to 1. We can often incrementally step in the direction of being more rational, by:

  1. Find a set of credences we assign
  2. Determine these credences are inconsistent with each other (e.g. they're mutually exclusive but the probabilities sum to more than 1)
  3. Modify the credences to become consistent.

Of course, this modification in step 3 might cause the credences to become inconsistent with some other credences we hold that we haven't considered. So this is not always a step forward. In general, it's NP-hard to find a consistent set of credences.

  • The problem is if coherence is the only criteria, all sorts of credences are permissive, which defeats the purpose of it. The more you restrict the criteria, the more you deal with the reference class problem. Each event can be placed into one of infinite classes which has different probabilities
    – user62907
    Aug 14, 2023 at 9:10
  • @thinkingman Coherence is one thing I said, the first thing I said is that you pick a theoretically successful system of updating beliefs and then what you should believe is what that system would infer. Coherence is subordinate to efficacy of the system. And the system we can pick is specifically ideal Bayesian inference with MDL prior.
    – causative
    Aug 14, 2023 at 15:22
  • Also, I personally believe that if you try to eliminate inconsistencies in the beliefs you hold, and sincerely examine the foundations, all humans will eventually end up with the same set of credences given the same information. The credence you "should" hold is the credence you "would" hold if you considered the issue long enough.
    – causative
    Aug 14, 2023 at 15:24
  • The problem is the correct credence is untestable, partially because the correct “credence” doesn’t exist. How can it? Reality isn’t partially true.
    – user62907
    Aug 14, 2023 at 18:49
  • @thinkingman And yet, if you want to design a robot to navigate a maze or play a game, the robot is going to represent things probabilistically. A robot that doesn't use probabilities is not going to work as well as one that does. That is the ultimate authority: how well does it work?
    – causative
    Aug 14, 2023 at 18:54

You must log in to answer this question.