A priori, a posteriori: Difference between revisions
mNo edit summary |
mNo edit summary |
||
Line 14: | Line 14: | ||
===A bit more practically=== | ===A bit more practically=== | ||
===Statistics=== | |||
In probability and statistics, particularly statistical inference, a priori is the prior knowledge of a population. | |||
Basically, it is anything factual that we can use to improve our model. | |||
It is more than just estimations or limited recent measurements. {{verify}} | |||
A priori probability http://en.wikipedia.org/wiki/A_priori_probability | |||
Posterior probability | |||
http://en.wikipedia.org/wiki/Posterior_probability | |||
===Modelling=== | |||
In machine learning and pattern recognition, and the models and math that backs it, | |||
a priori refers to factual/good/positive examples that make for supervised learning. | |||
(and a posteriori often short for 'a posteriori estimation' based on it) | |||
Without such a priori examples, the patterns would depend on data behaviour, clustering and such. {{verify}} | |||
===Knowledge (philosophically)=== | |||
Note that while a priori in the general sense can be translated as 'pre-existing', | |||
once you start saying 'a priori '' knowledge' '' you trip yourself into [[epistemology]] (a.k.a. 'what can we know'), and a bit of [[metaphysics]]. | |||
'''A priori knowledge''': are things that can be knowable independently of experience/evidence {{comment|(pedantry: ...aside from the experience of the language to communicate it)}}. | |||
: say, anything that follows from logic ''alone''. | |||
::: e.g. regardless of observation, we can say "all bachelors are unmarried", whereas for other things we need observation. | |||
'''A posteriori knowledge''' are things that can only be knowable, or verifiable, from empirical evidence. | |||
: that which is (or must necessarily be) deduced from epirical evidence, from experience, observation, or personal decision. | |||
The distinction is related to objective versus subjective observation.{{verify}} | |||
<!-- | |||
A priori knowledge often means you can reason it from principles or definitions. | |||
That sentence makes it really easy to be weasely about it, through, | |||
because a lot of principles and definitions didn't come from nowhere. | |||
More practically: | |||
: You might still want to verify it using emperical evidence | |||
: a priori tends to mean we are taking general princtiples/definitions/agreements {{comment|(in part just because a lot of knowledge exists in the form in generalisations)}} and doing something useful with it, involving more specific consequences/effects | |||
In many cases, a posteriori is almost synonymous with "empirical", | |||
because it often goes from particular observations | |||
to more generalized descriptions or theories. | |||
"Bachelors tend to be happy" or "George V reigned from 1910 to 1936" or "It is now raining outside" is something you cannot arrive at from reason, because it came from observation and/or requires emperical verification. | |||
--> | |||
===Law=== | |||
In ''law'', a priori refers to being based on hypothesis or deduction, rather than experimentation. | |||
It can still refer to subjective, semantic details: | |||
a | testimonials are automatically subject to a priori plausability - personal back knowledge. {{verify}} | ||
===Linguistics=== | |||
<!-- | |||
In linguistics, a priori [[constructed language]]s are those that are created from scratch. | |||
A posteriori constructed languages are those that mix and match from existing ones. | |||
--> | |||
====Why the terms are fuzzier than we pretend they are==== | |||
<!-- | |||
More pragmatically, a priori tends to translate as 'pre-existing', | |||
particularly in the statistical, modelling, and experimental-science sense. | |||
You might find statements like "there is no a priori knowledge of X", | |||
meaning there is nothing to go on yet, and you need some investigation in the real world. | |||
...'''but this is also where some people weasel in extra meanings'''. | |||
Some of them quite useful, | |||
some of them less so. | |||
In a setting like everyday scientific discussion, | |||
: a priori knowledge tends to mean "what is given without/before further inquiry" | |||
: a posteriori tends to mean "what we can conclude from this experiment" | |||
This is not as clear-cut as it looks. | |||
For example, note that "what we know already" can easily blur the line between logic and evidence. | |||
Science is ''all'' a posteriori, with science being very much about only being about evidence, and giving the best explanation for it. | |||
You could say the only things in science that are a priori are called mathematics. | |||
: ...but math cheats anyway, by only caring about internal consistency, not the real world at all. | |||
Everything ''interesting'' and everyday we are trying to do with science is likely to be based on posteriori. | |||
That said, the ''process'' of studying something with a scientific methodology often takes | |||
"well this is relevant" (a priori) and "what is under study" (about to be a posteriori) | |||
Since you can argue a priori barely exists, it's not a useful term/distinction at all, | |||
and it has been co-opted into softer distinctions. | |||
For example, in everyday science it's easy to say "you know that theory that's looked solid for the last hundred years? Let's treat that as fact, i.e. a priori knowledge, for this particular experiment". | |||
Or even to roughly split into "stuff I've finished checking" and "hypotheses I want to look at." | |||
'''Is this mostly just about timing?''' | |||
That is, a priori often seems use to mean "previous evidence" or "settled in the past". | |||
'' | But more importantly, what was a posteriori for us, once accepted widely ''enough'', | ||
is just evidence (a priori) for the next. | |||
Cited, sure, particularly if not just factual, sure, but still. | |||
'' | Most of what we consider 'everyday facts', | ||
and treat as a priori(-ish), | |||
are really neither of these things, | |||
but it's still useful to treat them that way. | |||
---- | |||
'''An a posteriori conclusion that is certain enough can be taken as a priori assumptions.''' | |||
If a statement is based on evidence, it would be a posteriori even after everyone considers it presumable or obvious. | |||
Even though the evidence is pretty damning, any useful answer to "does smoking cause cancer?" will be based on evidence. | |||
...while in a hundred years it may be so thoroughly proven that ''pragmatically'' it can be considered a priori assumption (in the non-scientific, everyday meaning of 'fact') to any discussion about smoking. | |||
Before people considered the idea, "Smoking causes cancer" was neither of these things - it was a possibility, then a proposition. The observation that cancer seemed to happen more in smokers isn't enough to call it a priori facto or a posteriori conclusion. | |||
--- | |||
It's easy to consider datasets (or just data in general) evidence, | |||
If you have a dataset, there is the assumption that there is | |||
"A dataset" is | |||
On the other hand, if you have a dataset, | |||
you will have questions that will call on a posteriori information. | |||
People ''can'' say they have some a priori knowledge about it, meaning you say you know things about how it was collected. | |||
This fuzzes the difference somewhat, because here, "a priori knowledge" is used in a "what we already know about this empirical information" sense. | |||
--- | |||
--> | |||
[[Category:Clarification]] | [[Category:Clarification]] |
Revision as of 16:34, 23 October 2023
Most generally
A priori roughly means something like "(from) that which goes before".
- Often used in a "prior to experience/measurement".
A posteriori roughly means "(from) that which comes after".
- Often meaning after experience, often using said experience
A bit more practically
Statistics
In probability and statistics, particularly statistical inference, a priori is the prior knowledge of a population. Basically, it is anything factual that we can use to improve our model. It is more than just estimations or limited recent measurements. (verify)
A priori probability http://en.wikipedia.org/wiki/A_priori_probability
Posterior probability http://en.wikipedia.org/wiki/Posterior_probability
Modelling
In machine learning and pattern recognition, and the models and math that backs it, a priori refers to factual/good/positive examples that make for supervised learning.
(and a posteriori often short for 'a posteriori estimation' based on it)
Without such a priori examples, the patterns would depend on data behaviour, clustering and such. (verify)
Knowledge (philosophically)
Note that while a priori in the general sense can be translated as 'pre-existing', once you start saying 'a priori knowledge' you trip yourself into epistemology (a.k.a. 'what can we know'), and a bit of metaphysics.
A priori knowledge: are things that can be knowable independently of experience/evidence (pedantry: ...aside from the experience of the language to communicate it).
- say, anything that follows from logic alone.
- e.g. regardless of observation, we can say "all bachelors are unmarried", whereas for other things we need observation.
A posteriori knowledge are things that can only be knowable, or verifiable, from empirical evidence.
- that which is (or must necessarily be) deduced from epirical evidence, from experience, observation, or personal decision.
The distinction is related to objective versus subjective observation.(verify)
Law
In law, a priori refers to being based on hypothesis or deduction, rather than experimentation.
It can still refer to subjective, semantic details:
testimonials are automatically subject to a priori plausability - personal back knowledge. (verify)