Travelling into the past and seeing a human did it is no guarantee of correctness and still relies on inference. Seeing itself is theory laden. You theorize you’re not hallucinating. You theorize that in the past, people look enough like they do now that you’re recognizing one and not an alien species with identical appearance. That you’re not a brain in a vat.
Since you brought up fallibilism elsewhere, I’ll assume you already know about the relativity of wrong and tentative knowledge. Within that framing, it isn’t necessary to verify adductive inferences, only to iteratively and progressively become less wrong over time by testing our inferences against critical reasoning and evidence.
So when we justify that IBE is reliable in general, do we have to rely on past track record of IBE being successfull to do that?
Edit: Sorry for not being clear enough. I meant to ask if we need a inductive argument to justify IBE. Like how we make inductive inference based on past accurate instances of it. Do we need to show that IBE has produced reliable conclusions in the past in order to know it is generally reliable?
Abduction doesn’t lead to “correctness” or truth. It produces explanations which are either more or less reasonable or warranted. Conclusions (or conclusively-stated inferences, to be more precise) can be disproved by counter-corroborating evidence but they can’t be proven to be correct or true, only shown to be in greater correspondence with corroborating evidence. Deduction deals with (propositional or logical) truth, or what must be true; induction proposes patterns from data, stating what seems to be generally true; abduction infers explanations for particular effects, proposing hypotheses.
So, do we need to know past instances of abductive inference to be successfull in order to justify abductive inferences in general? If so, how do we know past instances of abductive inferences were successfull?
It depends on the epistemic context. In science, you won’t know if your AI was successful in the sense of being true; but you’ll know whether it comports with other evidence, if it points to testable predications that bear out; etc.
“Past instances of abductive inference which we regard as true” is a long way of saying “observations”, “evidence”, “data”, or “corroboration.”
So, for example if I see a cat at one spot, look away, look back and see that it's in a different spot with food, how do I know the inference that the cat moved from one place to the food's place because it was hungry is successfull? I can never see what happened to the cat when I looked away.
Some trhings you just can't know with absolutre certainty. In your example though, you can make that inference with a very high level of probability that you are correct.
What helped me a lot with these kind of questions was understanding the difference between "empirical scientific methodology" and "formal science methodology"
If you google it you'll see various scientific methods these and others have emerges that can be distinguished
. The empirical method is the most common. Formal Scientific method lands in the realm of hard sciences like physics
fox-mcleod | 20 hours ago
We don’t verify contingent knowledge.
Travelling into the past and seeing a human did it is no guarantee of correctness and still relies on inference. Seeing itself is theory laden. You theorize you’re not hallucinating. You theorize that in the past, people look enough like they do now that you’re recognizing one and not an alien species with identical appearance. That you’re not a brain in a vat.
Since you brought up fallibilism elsewhere, I’ll assume you already know about the relativity of wrong and tentative knowledge. Within that framing, it isn’t necessary to verify adductive inferences, only to iteratively and progressively become less wrong over time by testing our inferences against critical reasoning and evidence.
[OP] --Estel-- | 12 hours ago
So when we justify that IBE is reliable in general, do we have to rely on past track record of IBE being successfull to do that?
Edit: Sorry for not being clear enough. I meant to ask if we need a inductive argument to justify IBE. Like how we make inductive inference based on past accurate instances of it. Do we need to show that IBE has produced reliable conclusions in the past in order to know it is generally reliable?
knockingatthegate | a day ago
Abduction doesn’t lead to “correctness” or truth. It produces explanations which are either more or less reasonable or warranted. Conclusions (or conclusively-stated inferences, to be more precise) can be disproved by counter-corroborating evidence but they can’t be proven to be correct or true, only shown to be in greater correspondence with corroborating evidence. Deduction deals with (propositional or logical) truth, or what must be true; induction proposes patterns from data, stating what seems to be generally true; abduction infers explanations for particular effects, proposing hypotheses.
[OP] --Estel-- | a day ago
So, do we need to know past instances of abductive inference to be successfull in order to justify abductive inferences in general? If so, how do we know past instances of abductive inferences were successfull?
knockingatthegate | a day ago
It depends on the epistemic context. In science, you won’t know if your AI was successful in the sense of being true; but you’ll know whether it comports with other evidence, if it points to testable predications that bear out; etc.
“Past instances of abductive inference which we regard as true” is a long way of saying “observations”, “evidence”, “data”, or “corroboration.”
[OP] --Estel-- | a day ago
So, for example if I see a cat at one spot, look away, look back and see that it's in a different spot with food, how do I know the inference that the cat moved from one place to the food's place because it was hungry is successfull? I can never see what happened to the cat when I looked away.
knockingatthegate | a day ago
How are you defining “successful”?
[OP] --Estel-- | a day ago
What I mean is how we know that inference's conclusion is more likely to be true than not.
knockingatthegate | a day ago
Crucially — how are you defining “know”?
[OP] --Estel-- | a day ago
I mean that we have fallible true beliefs about things. Beliefs that are shown to be highly probable.
knockingatthegate | a day ago
“True” is not a property which we can apply to conclusions obtained via abductive reasoning.
[OP] --Estel-- | a day ago
I get that. I'm asking if the IBE's conclusion can be shown to be probable.
LazarX | a day ago
Some trhings you just can't know with absolutre certainty. In your example though, you can make that inference with a very high level of probability that you are correct.
seedpod02 | a day ago
What helped me a lot with these kind of questions was understanding the difference between "empirical scientific methodology" and "formal science methodology"
rodrigorootrj2 | a day ago
what is the difernce between empirical and formal?
seedpod02 | a day ago
If you google it you'll see various scientific methods these and others have emerges that can be distinguished . The empirical method is the most common. Formal Scientific method lands in the realm of hard sciences like physics