I am mostly a coherentist and get constantly tripped up by the correspondencist attitudes the sequences here take. So it may be a job for me. But beware, I am a sloppy arguer, I suck at being precise and exact, as I think in pictures which may be a good thing but the result is often “sorta-kinda y’know what I mean?” and useless for people who have a mathemethicians precise mind.
A) My main issue with correspondence theory is over-valuing the accuracy of observation, sensory experience etc. There is a hidden assumption that hypothesis-building or theories are far, far more inaccurate than observation. Eliezer frequently talks about just opening the box and looking, just opening your eyes and seeing, just checking etc. in short he has high confidence in observation being accurate.
B) I think observation is nothing but a lower-order theory/model/hypothesis. It can be just as inaccurate as theories. Quite literally: not only the conscious mind is affected by biases, but even the visual cortex.
C) The proress of science was retarded mostly by not having access to good enough observation instruments. You cannot really be a Galilei without a telescope. Bare-eye observation fails us all kind of ways, gives us a universe that is likely to be Ptolemaian.
D) But observation with instruments is just as problematic, as instruments can go wrong, can get miscalibrated, and ultimately they themselves rely on theory. You cannot build a LHC without already having lots of theoretical physics. Observation is fallible.
E) From this follows that you cannot simply match unreliable theories to unreliable observations and call it a day. You must also match observations to observations, theories to theories, and sometimes even observations to theories.
F) In other words, truth is whatever is coherent with the whole body of science, all observations and all theories cross-validating each other. One potentially faulty observation or ten does not a theory validate.
G) Quine: Two Dogmas, demonstrating how can only experimentally verify the whole body of science only, not any individual statement.
H) Data is Latin for “given”. Its etymology sounds like we are getting our data on fax from Heaven. In reality, data is anything but given. Data is gathered, mined through hard and fallible work.
I) The useful data is that gives diffs. I.e. when I am debugging a software I need not only data that says it fails in this case but also that it does not fail in that case. Mining these kinds of data is not easy, fallible, and relies on theory. I.e. I hunt for a diff only when I already have a hypothesis of what may be the cause of failure. As data is not given, you often need to form a hypothesis first and mine and gather data specifically to test it.
J) Objection: but in instrumental rationality, we want to change our sensory experience, so even if my observation of something being painful is wrong, if I made the pain go away I solved the problem, right? No. You are a doctor. A patient complains of stomach pain. You give painkillers. Pain goes away. A year later he dies of cancer. We cannot simply reduce instrumental rationality to felt, observed needs, and just assume the correctness of such assumption does not matter. Theory plays a role here. Knowing medical theory which gives a guess of cancer helps the doctor and the patient more than a very, very accurate observation of the pain.
Finally, let me quote from this Quine summary “Our beliefs form a web, with the outer fringes connecting to experience. Revision at the edges leads to revisions elsewhere in the web, but the decisions of where the revisions will occurs are underdetermined by the logical relations among the beliefs. In the revision, “any statement can be held true, come what may, if we make drastic enough changes elsewhere in the system”.
Of course, “Darwin” from “The Simple Truth” would point out that some of those drastic enough changes involve dying.
To simplify the whole thing, if truth is correspondence to reality, our data / observation of reality should be highly reliable meaning it should be “given” (Lat. datum) instead of mined and gathered through hard and fallible work, the direction of that work often determined by fallible hypotheses, relying on potentially faulty instruments, even worse bare-eye observation, bias in the visual cortex and symptoms of pragmatic problems often being quite far from their actual causes.
And since it is not the case, we are better of matching everything with everything, not just theory with observation, and that grand system of cross-matching is called the body of science.
G) Quine: Two Dogmas, demonstrating how can only experimentally verify the whole body of science only, not any individual statement.
Scientists conduct experiments on individual statements, or small sets of them, all the time. Quine’s nihilistic conclusion that “any statement can be held true, come what may, if we make drastic enough changes elsewhere in the system” fails to grapple with the question of how knowledge can be achieved in the face of the fact that it is achieved. The words “drastic” and “come what may” should have been a clue to him, a small sensation of confusion, for these words admit there is something wrong with going to such lengths, while the sentence denies it.
I spent most of this morning making arrangements for a trip abroad. To this end I looked up hotels and railway timetables on various web sites, and at last made choices, decisions, bookings, and payments.
I expect the trains to depart and arrive at the times stated, my tickets to be accepted, and the hotel I have booked to exist and to be expecting me. What is the coherentist analysis of this situation? Why am I rightly confident in these arrangements, subject only to the realistic but fairly unlikely possibilities of engineering faults, the hotel burning down, and the like?
Correspondence (matching theories to observations) is a subset of coherence (matching everything with everything)
It is a very useful subset as long as observations are reliable and easy to procure, which is in your case, and indeed in most, but not all cases it is so.
A counter-example would be Many-Worlds: you cannot match it with observations, but you can match it with other theories and see it follows the pattern.
Your observations rest on definitions which come from other parts of your knowledge. Trains depart on time? Down to the nanosecond level or second level will be okay? Is 10 secs late still okay? Based on the starter timezone, even if it goes through multiple ones? If they base departure time on starter time zone and arrival on destination time zone, won’t it upset your expectation of the trip length? If not, can you miss a connection? What does ticket accepted mean, would presenting a false ticket and bribing the conductor would count as accepted? Would a well made false ticket that tricks all the conductors do? This is not nitpicking, it just means your observations are obvious because they rest on all kinds of non-conscious, tacit, consensual knowledge beyond that. And that is roughly what Quine meant: I can prove hardly any time train ever departs on time if I just make another change in the system, such as saying on time means nanosecond exactness. This is a change not worth making, of course.
Correspondence (matching theories to observations) is a subset of coherence (matching everything with everything)
Correspondence is not just matching theories to observation. It is matching theories to reality. Since we don’t have pure transcendent access to reality, this involves a lot of matching theories to observation and to each other, and rejecting the occasional observation as erroneous; however, the ultimate goal is different from that of coherence, since perfectly coherent sets of statements can still be wrong.
If your point is that “reality” is not a meaningful concept and we should write off the philosophizing of correspondence theorists and just focus on what they actually do, then what they actually do is identical to what coherentists actually do, not a subset.
As far as I can tell, most coherentists want to match theories with reality too, because truth doesn’t really have any other useful definition. The goal is not to be coherent within a random and reality-detached set of sentences: the goal is to be coherent with the whole of science. When a scientists rejects (assigns very low probability to) the observation of a perpetuum mobile on the basis that it contradicts the laws of physics, that is a standard coherentist move. This is another one. The goal is to avoid having to waste time and costs on non-fruitful data gathering. Ultimately the only thing that is rejected is that blind data-only approach that may be considered the straw-manning of the correspondenceist position, except that one is actually unfortunately used too much. A coherentist will simply not spend money buying an airplane ticket to check if someone’s garage has a dragon, the proposition contradicts so much we already know that the very low prior probability does not worth the cost. You may as well call this a wiser version of correspondencism, the barriers are not exactly black and white here. This is unfortunately philosophy, so fairly muddy :)
You can have holism without coherence where you require that the whole of science is true by correspondence, but the parts aren’t.. Inasmuch as it is correspondence, it isnt coherence.
The correspondence theory of truth is a theory of truth, not a theory of justification. Correspondentists don’t match theories to reality, since they don’t have direct ways of detecting a mismatch, they use proxies like observation sentences and predictions. Having justified the a theory as being true, they then use correspondence to explain what it’s truth consists of.
This is not nitpicking, it just means your observations are obvious because they rest on all kinds of non-conscious, tacit, consensual knowledge beyond that.
That’s what nitpicking is!
And that is roughly what Quine meant: I can prove hardly any time train ever departs on time if I just make another change in the system, such as saying on time means nanosecond exactness.
Changing the definition of the words in a sentence does not change the proposition that was originally expressed by the sentence. It just creates a different proposition expressed with the same words, and is irrelevant to understanding the original one.
I am mostly a coherentist and get constantly tripped up by the correspondencist attitudes the sequences here take. So it may be a job for me. But beware, I am a sloppy arguer, I suck at being precise and exact, as I think in pictures which may be a good thing but the result is often “sorta-kinda y’know what I mean?” and useless for people who have a mathemethicians precise mind.
A) My main issue with correspondence theory is over-valuing the accuracy of observation, sensory experience etc. There is a hidden assumption that hypothesis-building or theories are far, far more inaccurate than observation. Eliezer frequently talks about just opening the box and looking, just opening your eyes and seeing, just checking etc. in short he has high confidence in observation being accurate.
B) I think observation is nothing but a lower-order theory/model/hypothesis. It can be just as inaccurate as theories. Quite literally: not only the conscious mind is affected by biases, but even the visual cortex.
C) The proress of science was retarded mostly by not having access to good enough observation instruments. You cannot really be a Galilei without a telescope. Bare-eye observation fails us all kind of ways, gives us a universe that is likely to be Ptolemaian.
D) But observation with instruments is just as problematic, as instruments can go wrong, can get miscalibrated, and ultimately they themselves rely on theory. You cannot build a LHC without already having lots of theoretical physics. Observation is fallible.
E) From this follows that you cannot simply match unreliable theories to unreliable observations and call it a day. You must also match observations to observations, theories to theories, and sometimes even observations to theories.
F) In other words, truth is whatever is coherent with the whole body of science, all observations and all theories cross-validating each other. One potentially faulty observation or ten does not a theory validate.
G) Quine: Two Dogmas, demonstrating how can only experimentally verify the whole body of science only, not any individual statement.
H) Data is Latin for “given”. Its etymology sounds like we are getting our data on fax from Heaven. In reality, data is anything but given. Data is gathered, mined through hard and fallible work.
I) The useful data is that gives diffs. I.e. when I am debugging a software I need not only data that says it fails in this case but also that it does not fail in that case. Mining these kinds of data is not easy, fallible, and relies on theory. I.e. I hunt for a diff only when I already have a hypothesis of what may be the cause of failure. As data is not given, you often need to form a hypothesis first and mine and gather data specifically to test it.
J) Objection: but in instrumental rationality, we want to change our sensory experience, so even if my observation of something being painful is wrong, if I made the pain go away I solved the problem, right? No. You are a doctor. A patient complains of stomach pain. You give painkillers. Pain goes away. A year later he dies of cancer. We cannot simply reduce instrumental rationality to felt, observed needs, and just assume the correctness of such assumption does not matter. Theory plays a role here. Knowing medical theory which gives a guess of cancer helps the doctor and the patient more than a very, very accurate observation of the pain.
Finally, let me quote from this Quine summary “Our beliefs form a web, with the outer fringes connecting to experience. Revision at the edges leads to revisions elsewhere in the web, but the decisions of where the revisions will occurs are underdetermined by the logical relations among the beliefs. In the revision, “any statement can be held true, come what may, if we make drastic enough changes elsewhere in the system”.
Of course, “Darwin” from “The Simple Truth” would point out that some of those drastic enough changes involve dying.
To simplify the whole thing, if truth is correspondence to reality, our data / observation of reality should be highly reliable meaning it should be “given” (Lat. datum) instead of mined and gathered through hard and fallible work, the direction of that work often determined by fallible hypotheses, relying on potentially faulty instruments, even worse bare-eye observation, bias in the visual cortex and symptoms of pragmatic problems often being quite far from their actual causes. And since it is not the case, we are better of matching everything with everything, not just theory with observation, and that grand system of cross-matching is called the body of science.
Scientists conduct experiments on individual statements, or small sets of them, all the time. Quine’s nihilistic conclusion that “any statement can be held true, come what may, if we make drastic enough changes elsewhere in the system” fails to grapple with the question of how knowledge can be achieved in the face of the fact that it is achieved. The words “drastic” and “come what may” should have been a clue to him, a small sensation of confusion, for these words admit there is something wrong with going to such lengths, while the sentence denies it.
I spent most of this morning making arrangements for a trip abroad. To this end I looked up hotels and railway timetables on various web sites, and at last made choices, decisions, bookings, and payments.
I expect the trains to depart and arrive at the times stated, my tickets to be accepted, and the hotel I have booked to exist and to be expecting me. What is the coherentist analysis of this situation? Why am I rightly confident in these arrangements, subject only to the realistic but fairly unlikely possibilities of engineering faults, the hotel burning down, and the like?
Correspondence (matching theories to observations) is a subset of coherence (matching everything with everything)
It is a very useful subset as long as observations are reliable and easy to procure, which is in your case, and indeed in most, but not all cases it is so.
A counter-example would be Many-Worlds: you cannot match it with observations, but you can match it with other theories and see it follows the pattern.
Your observations rest on definitions which come from other parts of your knowledge. Trains depart on time? Down to the nanosecond level or second level will be okay? Is 10 secs late still okay? Based on the starter timezone, even if it goes through multiple ones? If they base departure time on starter time zone and arrival on destination time zone, won’t it upset your expectation of the trip length? If not, can you miss a connection? What does ticket accepted mean, would presenting a false ticket and bribing the conductor would count as accepted? Would a well made false ticket that tricks all the conductors do? This is not nitpicking, it just means your observations are obvious because they rest on all kinds of non-conscious, tacit, consensual knowledge beyond that. And that is roughly what Quine meant: I can prove hardly any time train ever departs on time if I just make another change in the system, such as saying on time means nanosecond exactness. This is a change not worth making, of course.
Correspondence is not just matching theories to observation. It is matching theories to reality. Since we don’t have pure transcendent access to reality, this involves a lot of matching theories to observation and to each other, and rejecting the occasional observation as erroneous; however, the ultimate goal is different from that of coherence, since perfectly coherent sets of statements can still be wrong.
If your point is that “reality” is not a meaningful concept and we should write off the philosophizing of correspondence theorists and just focus on what they actually do, then what they actually do is identical to what coherentists actually do, not a subset.
As far as I can tell, most coherentists want to match theories with reality too, because truth doesn’t really have any other useful definition. The goal is not to be coherent within a random and reality-detached set of sentences: the goal is to be coherent with the whole of science. When a scientists rejects (assigns very low probability to) the observation of a perpetuum mobile on the basis that it contradicts the laws of physics, that is a standard coherentist move. This is another one. The goal is to avoid having to waste time and costs on non-fruitful data gathering. Ultimately the only thing that is rejected is that blind data-only approach that may be considered the straw-manning of the correspondenceist position, except that one is actually unfortunately used too much. A coherentist will simply not spend money buying an airplane ticket to check if someone’s garage has a dragon, the proposition contradicts so much we already know that the very low prior probability does not worth the cost. You may as well call this a wiser version of correspondencism, the barriers are not exactly black and white here. This is unfortunately philosophy, so fairly muddy :)
You’ve got coherentism confused with holism.
Is holism even a thing?
Yes. So is Google.
You can have holism without coherence where you require that the whole of science is true by correspondence, but the parts aren’t.. Inasmuch as it is correspondence, it isnt coherence.
The correspondence theory of truth is a theory of truth, not a theory of justification. Correspondentists don’t match theories to reality, since they don’t have direct ways of detecting a mismatch, they use proxies like observation sentences and predictions. Having justified the a theory as being true, they then use correspondence to explain what it’s truth consists of.
That’s what nitpicking is!
Changing the definition of the words in a sentence does not change the proposition that was originally expressed by the sentence. It just creates a different proposition expressed with the same words, and is irrelevant to understanding the original one.