How do I validate my confidence in things? [duplicate]

Question

This is honestly tripping me out the more I delved into it.

Of course, I feel more confident that my mother is my real mother than myself being kidnapped tomorrow. But how do I show that this is rational? How do I show that I have strong reasons to put higher credence in the former than the latter?

Of course, one can simply look at the rates of mothers being real and the rates of kidnapping and compare these rates. But okay, I now have rates.

How do I go from “the rate of mothers being real is X% and the rate of kidnappings in my city is Y%” to “I should have a higher credence in my mother being real than being kidnapped tomorrow.”

What is the reasoning step that takes you from rates to credences?

There is a second problem. Why should those rates be representative of me? There are multiple rates I can get. I can for example get the rate of kidnappings in my area, or how often people of my demographic are kidnapped, or any one of an infinite number of rates that may be plausible or relevant in the situation. Which one of these rates is relevant to me? I believe the term for this is the base rate problem.

After scouring through the internet, it seems that I’m not the only one that’s stumbled upon this issue. David Deutsch has an entire borderline tirade against this: https://youtu.be/wfzSE4Hoxbc?si=na80Dhd34th4YRH5

He has more extreme views on why confidence levels in propositions should be eliminated entirely. I’m surprised this isn’t talked about more. If taken to its full conclusion, it seems to imply that justifying that you should be more confident in X than Y is an arduous if not impossible task. But these confidences are what we use to base decisions upon almost everything in life. When deciding between A and B, we pick the option we feel more confident in.

Does this mean that our decisions are all ultimately unjustified and we’re blindly running along following our instincts (which may not be a bad thing)?

How does one deal with this?

Answer 1

An important epistemologically conundrum related to the issue at hand is the Münchhausen trilemma:

In epistemology, the Münchhausen trilemma is a thought experiment intended to demonstrate the theoretical impossibility of proving any truth, even in the fields of logic and mathematics, without appealing to accepted assumptions. If it is asked how any given proposition is known to be true, proof in support of that proposition may be provided. Yet that same question can be asked of that supporting proof, and any subsequent supporting proof. The Münchhausen trilemma is that there are only three ways of completing a proof:

• The circular argument, in which the proof of some proposition presupposes the truth of that very proposition
• The regressive argument, in which each proof requires a further proof, ad infinitum
• The dogmatic argument, which rests on accepted precepts which are merely asserted rather than defended

The trilemma, then, is the decision among the three equally unsatisfying options. Karl Popper's suggestion was to accept the trilemma as unsolvable and work with knowledge by way of conjecture and criticism.

Other variants include Fries's trilemma:

Jakob Friedrich Fries formulated a similar trilemma in which statements can be accepted either:

• dogmatically
• supported by infinite regress
• based on perceptual experience (psychologism)

The first two possibilities are rejected by Fries as unsatisfactory, requiring his adopting the third option. Karl Popper argued that a way to avoid the trilemma was to use an intermediate approach incorporating some dogmatism, some infinite regress, and some perceptual experience.

And Albert's formulation:

An English translation of a quote from the original German text by Albert is as follows:

Here, one has a mere choice between:

• An infinite regression, which appears because of the necessity to go ever further back, but is not practically feasible and does not, therefore, provide a certain foundation.
• A logical circle in the deduction, which is caused by the fact that one, in the need to found, falls back on statements which had already appeared before as requiring a foundation, and which circle does not lead to any certain foundation either.
• A break of searching at a certain point, which indeed appears principally feasible, but would mean a random suspension of the principle of sufficient reason.

With this background in mind, I'll respond to your specific questions:

Of course, I feel more confident that my mother is my real mother than myself being kidnapped tomorrow. But how do I show that this is rational? How do I show that I have strong reasons to put higher credence in the former than the latter?

You could argue dogmatically based on the assumption of uniformitarianism, that gives you the background context for expecting regularity in the universe, without which inductive reasoning and probability and statistics would be meaningless. Or you could also argue from personal perceptual experience, by claiming that your belief in your mother is properly basic. In contrast, your belief in being kidnapped tomorrow would fail to enjoy the same justifications: it's not based on your direct perceptual experience (I'm assuming you have not had the personal experience of being kidnapped yet), and the dogma of uniformitarianism upon which inductive reasoning and statistics are based would suggest that your being kidnapped tomorrow is statistically unlikely.

How do I go from “the rate of mothers being real is X% and the rate of kidnappings in my city is Y%” to “I should have a higher credence in my mother being real than being kidnapped tomorrow.” What is the reasoning step that takes you from rates to credences?

Back to the Münchhausen trilemma, you need to stop the infinite regress of justifications somehow. You could stop it, for example, by accepting uniformitarianism dogmatically, because it appears to be self-evident, and then you can develop your worldview based on that. So, if you assign a high degree of credence to uniformitarianism, it would make sense to consequently assign a proportionally high degree of credence to things that are statistically likely to happen. On the other hand, if you were to reject uniformitarianism as a dogma in the first place, then it seems to me that there would be no obvious reason to establish a mapping from past statistics to credence about future events.

There is a second problem. Why should those rates be representative of me? There are multiple rates I can get. I can for example get the rate of kidnappings in my area, or how often people of my demographic are kidnapped, or any one of an infinite number of rates that may be plausible or relevant in the situation. Which one of these rates is relevant to me? I believe the term for this is the base rate problem.

Again, if uniformitarianism is accepted dogmatically, then unless there is significant evidence that the kidnapping statistics of your area or of your specific demographic are particularly different from the average, there would seem to be no reason for expecting something significantly different from the norm.

Does this mean that our decisions are all ultimately unjustified and we’re blindly running along following our instincts (which may not be a bad thing)?

The Münchhausen trilemma seems to agree with this conclusion.

How does one deal with this?

You need to stop the infinite regress with some stopping criteria that you are comfortable with. Personally, basic beliefs grounded in perceptual experiences seem to be a reasonable stopping criterion.

---

Relevant related questions:

• Does externalism solve the Münchhausen trilemma?
• Is the Münchhausen trilemma really a trilemma?
• Why is the Münchhausen trilemma an unsolved problem?

Answer 2

Of course, I feel more confident that my mother is my real mother than myself being kidnapped tomorrow. But how do I show that this is rational? How do I show that I have strong reasons to put higher credence in the former than the latter?

The answer is really very simple. If you are able to think of good reasons, i.e., reasons that you believe are good enough, that the woman you believe is your mother is really your mother, then you will have changed your irrational belief into a rational one.

As long as you are unable to do that, your belief remains irrational.

Irrational doesn't mean wrong, though. Most of what we do in life, maybe 99% or more, is based strictly on irrational beliefs and yet it seems to work. We can stay comfortable, well fed, and even happy enough not to contemplate suicide every other minute. It is true that civilisation seems to involve quite a bit of rationalisation and there is no denying that this works well, but this shouldn't make us forget that, mostly, our view on things is essentially motivated by intuitive cognitive processes that we know very little about rather than on any extensive rationalisation.

Further, the idea is itself hopeless.

Rationalisation just means that you try to justify uncertain beliefs from more certain beliefs. This is intrinsic to how logic works and there is nothing you can do to improve the prospect of a 100% rationalised decision. We can only derive a conclusion from some premises we believe in. If we want to prove our belief, then we have to assume other premises to derive our initial premises from them. This is only interesting if we can find new premises that we can trust better than our initial premises, but there is apparently no useful premises of which we could be so certain that we would know that they are true.

Thus, rationalisation is only a marginal improvement on the confidence we have in our beliefs. Pragmatically, though, this is terrific, for it works really well. For example, the judge is able to form an opinion he or she can trust, and more importantly that we can all trust, only because the judiciary process is essentially a rationalisation of the decision.

We stop trying to rationalise further once we come to believe that the cost of any additional rationalisation would not be justified by any benefits we may believe it could result in.

There is no 100% rationalised belief.

Beyond that, you can consider what you actually know. You know you are in pain when you are in pain. You know that you believe the sun is shining when you believe that the sun is shining. You know your beliefs! Beliefs are more or less certain, so all you can do is reason from your more certain beliefs and hope for the best.

You can also cooperate with other people to improve the process of rationalisation, but this is only if you believe that these other people are reasonable people. Scientists do it, though, so why wouldn't you?

I don't know where people got the idea that we should be able to understand and explain everything there is in the real world!

Answer 3

Adapting my other answer..

How to justify our views and claims is both a human need and a philosophical problem.

There is no theory or criterion in a vacuum as absolute (as Munchhausen's trilemma and the problem of the criterion point out).

But one can make use of that and turn it into an advantage. In that sense, if a claim can be defended succesfully against all potential challenges and criticisms, then this means there is simply no better explanation available than that.

From an epistemological point of view, this is used by some approaches to epistemology and truth.

Another escape from the diallelus is critical philosophy, which denies that beliefs should ever be justified at all. Rather, the job of philosophers is to subject all beliefs (including beliefs about truth criteria) to criticism, attempting to discredit them rather than justifying them. Then, these philosophers say, it is rational to act on those beliefs that have best withstood criticism, whether or not they meet any specific criterion of truth.

In this sense, exposing a claim to more challenges makes us able to see its robustness or shortcomings easier.

An important way a claim or theory is exposed to challenge is through being put to practical use, since then a claim is exposed to the plethora of diverse conditions that make up a given reality and its robustness against these becomes evident. Both the marxist criterion of practice and the pragmatic criterion are similar in this respect.

Furthermore, jury systems, collective decision making and the like can be re-cast in that light as well, since these are also ways that a claim is exposed to more diverse challenges.

Hope these help defend and clarify your claims and beliefs.

Answer 4

Your questions are concerned with the issue that philosophers traditionally call the problem of induction. The philosophical literature on this subject is enormous. Given observations of events, is there a rational way to justify an inference to statements about unobserved events? These are called projections. Usually we are concerned with making projections about the future based on observations of the past.

The fact that things have happened a particular way in the past is usually considered to provide evidence about future events. A primitive tribesman from 10,000 years ago might notice that the sun always rises in the east and sets in the west. The elders of his tribe confirm that this has always happened during their lifetimes, but neither he nor they have any scientific knowledge to explain it. It seems reasonable to accept that he should expect the sun to rise in the east tomorrow, since it always has. The inference is not a certain one, but it is highly plausible. Explaining why it is plausible turns out to be surprisingly difficult.

The classic challenge to this kind of reasoning comes from David Hume. As I described in a recent answer, Hume argues that inductive reasoning is not demonstrative, since it lacks deductive certainty, and it cannot be justified based on the uniformity of nature without circularity.

There have been many proposed responses. Here are some of the main ones:

1. To base our projections on past observations is just part of what it means to be rational. It doesn't make sense to challenge the rationality of drawing inferences from experience.

2. Inductive reasoning is really a species of inference to the best explanation. We hypothesise some law-like causal explanations of the observed regularities and accept as plausible the one that provides the most satisfactory explanation according to our best scientific criteria.

3. We can use Bayesian methods to form probabilistic assessments of hypotheses based on prior probabilities and Bayesian conditioning. A common version of this approach does not attempt to place absolute probabilities on hypotheses but determines the extent to which given data provides evidence for one hypothesis relative to another.

4. There is no such thing as inductive reasoning. All we can do is formulate hypotheses and reject those that produce false predictions. This does not justify the remaining hypotheses, it just means they have survived attempts at falsification. This is Popper's proposal and is the one David Deutsch supports in the video you linked.

5. Statistical methods, particularly those that are used for testing statistical models, constitute a piecemeal solution. We can formulate models and test their predictions against observed data and reject those that are consistently inferior to their rivals. This might be considered a probabilistic version of Popper's falsificationism.

6. Induction can be justified inductively. Past inductions have mostly worked out pretty well. The reasoning is circular, but not viciously so.

7. The success of inductive reasoning is justified by its practical consequences. Individuals who are not good at inductive reasoning do not survive long enough to procreate. Inductive reasoning just works, accept it.

8. All inductive reasoning is justified on a case-by-case basis by examining the merits of the individual arguments. There is no need for general rules of induction or a general principle of the uniformity of nature.

9. Inductive methods are themselves subject to inductive assessment. We can perform a kind of meta-induction by formulating competing methods of induction and testing them to see which ones work best.

On the specific issues that you mention. Inferring a degree of credence c from a frequency of c is a feature of David Lewis' Principal Principle. It does seem somewhat arbitrary though, and it depends on an "absent any other considerations" clause. In practice there are always other considerations. Bayesians treat frequencies as evidence against which hypotheses are calibrated.

The fact that you can get completely different probabilities depending on how you frame or measure the frequency is called the reference class problem. It is a problem common to pretty much all statistical methods. A possible response is to say you need to formulate some specific hypothesis about which background data is relevant and then do your best to justify that hypothesis.

Answer 5

After scouring through the internet, it seems that I’m not the only one that’s stumbled upon this issue. David Deutsch has an entire borderline tirade against this: https://youtu.be/wfzSE4Hoxbc?si=na80Dhd34th4YRH5

He has more extreme views on why confidence levels in propositions should be eliminated entirely. I’m surprised this isn’t talked about more. If taken to its full conclusion, it seems to imply that justifying that you should be more confident in X than Y is an arduous if not impossible task. But these confidences are what we use to base decisions upon almost everything in life. When deciding between A and B, we pick the option we feel more confident in.

Does this mean that our decisions are all ultimately unjustified and we’re blindly running along following our instincts (which may not be a bad thing)?

Deutsch follows Karl Popper who sez that knowledge is created by conjecture and criticism. You notice a problem with existing ideas, guess solutions to it and criticise those solutions according to whether they solve the problem and are compatible with other knowledge, including experimental tests, see Deutsch's books "The Fabric of Reality" and "The Beginning of Infinity" for his take on this and the citations here

https://fallibleideas.com/books#popper

You should decide on ideas by considering alternatives and looking for criticisms until you have an idea with no criticisms. A feeling of confidence is a sensation that happens as a result of some idea meeting your internalised standards of criticism. That confidence may mislead you if your standards for some subject are bad: there are many confident creationists and flat earthers. For substantive advice on how to do this see

https://criticalfallibilism.com/yes-or-no-philosophy/

Trying to base your epistemology on feelings of confidence rather than adjusting your standards of confidence in the light of criticism may lead to bad results. For example, the Bayesian epistemology forum Less Wrong banned a critic because they couldn't answer his criticisms and decided to downvote him instead:

https://curi.us/2381-less-wrong-banned-me