If free will does not exist, can we distinguish between experimental and observational sciences?

[Boojatta]

[Goblinmonger]

1. No n/t

[cosmik debris]

2. Maybe n/t

[Goblinmonger]

5. OK

:shrug:

[cyborg_jim]

3. I don't see what free-will has got to do with that.

[cosmik debris]

4. Or not.

Suppose two chess experts watch a game of chess and rate each move as follows: inadequate, satisfactory, good, or excellent. Suppose that a given move gets rated as inadequate by some chess experts and good by other chess experts. Alternatively, suppose that a given move gets rated as satisfactory by some chess experts and excellent by other chess experts. Would you be able to conclude that the quality of moves in chess is subjective? If you reach that conclusion, then what do you think determines whether the game is a win for white, a win for black, or a draw?

[Boojatta]

30. You are right that this thread asks a "yes" or "no" question.

I can see that one might wish to respond to such a question by saying that one is unable to give a definite answer, but I would hope for some kind of explanation as to why one is unable to give a definite answer.

Also, in any thread that I start, if any contributor posts a block of text that he or she did not write, then it will be much appreciated if such contributor marks it as an excerpt or quotation and credits the author of the text.

[Name removed]

31. Deleted message

Message removed by moderator. Click here to review the message board rules.

[Dhalgren]

6. Yes. n/t

[Boojatta]

9. How? e.o.m.

[Dhalgren]

10. Simply because we have no free will does not mean that we can make

determinations with our tiny brains in any way that would even hint at any but the most rudimentary predictive abilities. Since the Universe is big and we are small, we must, perforce, distinguish between observation and experimentation. Experimentation may indeed tell us what is already knowable, but that nevertheless we are incapable of discovering otherwise, due to our tiny brains. Just because we may arrive logically at the determination that there is no free will, it does not follow that "experimentation" and "observation" are therefore synonymous.

[Boojatta]

12. Three Questions

Simply because we have no free will does not mean that we can make determinations with our tiny brains in any way that would even hint at any but the most rudimentary predictive abilities.

Assuming that we can make only the most rudimentary predictions, what conclusion do you arrive at?

Since the Universe is big and we are small, we must, perforce, distinguish between observation and experimentation.

Does a need to distinguish between them automatically provide us with a method for distinguishing between them?

Experimentation may indeed tell us what is already knowable, but that nevertheless we are incapable of discovering otherwise, due to our tiny brains.

Would it be possible to observe the people who are allegedly performing "experiments" and, by means of such observations, record all information that we need to know about the "experiments"?

[damntexdem]

7. It was predetermined that I wouldn't care.

Can I now choose to care?

[Dhalgren]

11. Yes, but your choice was predetermined...

:evilgrin:

[valerief]

8. The best things in life are free. nt

[More Than A Feeling]

13. What is "free will"? n/t

[Boojatta]

21. This thread is a response to those who say that ...

there is no free will. Thus, the relevant meaning of the term "free will" is the meaning assigned by those people.

Unless I misunderstood Dhalgren's contribution to this thread, Dhalgren is an example of someone who claims that there is no free will.

Why not send a PM to Dhalgren to request an explanation of the meaning of the term "free will"?

[More Than A Feeling]

23. You already said that (in a PM)!

Do you even have a definition of free will? Or (since you keep pointing me towards someone else) do you just start threads around terms, and hope that someone will assume a definition, and then you can go with that, rather than actually know the meaning of what you are asking about?

[Dhalgren]

26. You don't have to PM me. My idea of "free will" is basically this:

It is the idea that an entity in existence can initiate a causal chain with no antecedent other than the desire of said entity. "Free will" would have each extant "thing" capable of initiating causal chains - and not just capable of doing so, but unable not to do so. That, at least, is my take on it.

:)

[Random_Australian]

14. Of course. Free will has nothing to do with it, providing that the

universe is approximately the same in both cases. But I don't see you saying "if there was no free will and we couldn't see the pink elephants that are everywhere".

[Boojatta]

15. Okay, let's forget about free will for the moment.

How do you distinguish between experimental and observational sciences?

[Random_Australian]

16. The distinction is small.

The only real difference between epxeriment and observation is that in observational analysis correlation is meaningless in terms of causation, as there are many uncontrolled variables.

In experimental work, there are few, and the extent of their effect is controlled such that you can infer causation.

[Boojatta]

17. What does it mean for a variable to be "controlled"? e.o.m.

[Random_Australian]

19. The extent to which it affects each observation changes by a small

amount between each observation.

(Close enough for argument's sake - in real life there is a bit more to it)

[Boojatta]

20. Suppose for example that I observe a sundial.

Then, fifteen minutes later, I make another observation. The change in the position of the Sun affects the observations, but the effect is small because it's not a particularly big sundial. Comparing the two observations, I see only a small difference. Can I conclude that the position of the Sun is a controlled variable?

[Random_Australian]

22. What you do is look at how much error the change causes in your measurement.

For instance, if you wanted to see if two angles were almost the same (on platforms about 15 minutes walk apart), by eye (which has an accuracy of about +/- 4 degrees for a reasonable sundial) then you could wait until the angle of the first was made on the sundial, walk over to the other platform and compare.

The sun will have moved a smaller amount than the accuracy of your initial measurement, so it would be acceptable and in that experiment the position of the sun is basically constant, so it would be a "controlled" variable.

Note that we only use the word "controlled" because "independent" was already taken, and for the vast majority of experiments these things are controlled. This doesn't imply literal control though!

[Boojatta]

24. Can you use your criteria to classify a different kind of study?

Is detecting radio waves from extraterrestrial sources and analyzing them in an effort to find patterns indicating they may have come from extraterrestrial civilizations part of an experimental or observational study?

[Random_Australian]

25. It's not a study at all; it's a comparison.

While observing radio sources per se is a study, an observational one, in which you note the correlation between sources and their representation and physical properties using already known experimental studies.

However, looking for life is simply going through the data and trying to match patterns to what might be expected. There is no temporal component, therefore there is no cause and effect.

[Boojatta]

27. Radio waves are produced by some processes, no?

However, looking for life is simply going through the data and trying to match patterns to what might be expected. There is no temporal component; therefore there is no cause and effect.

If we want to decide what caused some radio waves to be transmitted and, in particular, whether they were produced by an extraterrestrial civilization, then are we not trying to infer a cause?

I have a couple of questions related to other scientific investigations.

1. Apparently at least one eruption of Vesuvius occurred in August of AD 79. If archaeologists and medical historians infer that this eruption killed some people at Pompeii, then are they necessarily performing experiments and not merely conducting observational studies?

2. When some ancient astronomers predicted eclipses, were they simply making lucky guesses? Were they unable to infer the cause of eclipses and limited to identifying phenomena that are to some extent correlated with eclipses?

[Random_Australian]

28. Aha, subtle distinction about the radio waves.

There are two seperate experiments, one observational and one known as a 'quasi-experiment' where you have control over many but not all variables.

The obervational one is comparing the radio waves, the quasi-experiment is finding the patterns that humans make. (The uncontrolled variale is the limits of human thinking)

Anyway, to the other questions:

1) No, the people working on historical things are ALL observational studies. However, when presented with something that can only have a few causes, it's often practical to find which is the most likely. For Pompeii, for instance, it is considered very likely that it was the volcano.

2) They were neither making lucky guesses, nor performing exact experiments.

The methods they used are uncertain to me, but they could either work these things out mathematically from very many simple models, or work it out very many mathematical ways from a more accurate model. And yes, there is a fair chance that they could correctly identify underlying phenomena; there is nothing in observational experimentation that prevents one from identifying what's actually going on; it's just more difficult to build a case. (And from a single experiment, pretty much impossible as I think I said above)

[Boojatta]

29. Referring to "the only real difference", how do you arrive at those answers?

The distinction is small.
The only real difference between experiment and observation is that in observational analysis correlation is meaningless in terms of causation, as there are many uncontrolled variables.

In experimental work, there are few, and the extent of their effect is controlled such that you can infer causation.

[Random_Australian]

32. Not difficult. In observational experiments there are many variables at work,

thus many possible causes for any given 'effect', and in strictly experimental there is not.

I'm sorry, but I don't know what bit you aren't getting; I think you'll have to put it clearer :rofl:

Or in other words, if this reasoning is false you should be able to create a contradiction; is there any situation in which you think the above is an incorrect description?

[Boojatta]

33. It seems that you are defining experimental and observational thus:

Experimental:
"few" uncontrolled variables
causation can be inferred

Observational:
"many" uncontrolled variables
causation cannot be inferred

Do you now wish to delete mention of causation from your definition?

How many uncontrolled variables are to be considered "few" and how many are to be considered "many"?

[Random_Australian]

36. Almost.

Experimental:
(you do you best to make sure there are) no variables which affect the outcome which are dependent on the trial (that is, whether or not you do a control run or a real one)

Causation can be inferred. (That is, it can be shown there is a very high probability that a certain cause has a certain effect.

Observational:
Gather data of some variable versus another variable, correlate with other variables against the same reference one.

Causation cannot be inferred.

There are other types of experiment and analysis, but I suspect the one you want to know about is

Quaso-experiment:

There is one variable which cannot be controlled by the experimenter, and then they do their experiment within that framework.

So to answer the rest of your question, as soon as there is one other known variable, your inference of causation drops to about one-half. That is, you think one of two variables is behind it all.

:)

[Boojatta]

37. Quoting from the post that I am replying to:

Experimental:
Causation can be inferred. (That is, it can be shown there is a very high probability that a certain cause has a certain effect.

Observational:
Causation cannot be inferred.

Based on the above, it is clear that the answers are as follows:

Apparently at least one eruption of Vesuvius occurred in August of AD 79. If archaeologists and medical historians infer that this eruption killed some people at Pompeii, then are they necessarily performing experiments and not merely conducting observational studies?

Yes, they are necessarily performing experiments rather than observational studies.

2. When some ancient astronomers predicted eclipses, were they simply making lucky guesses? Were they unable to infer the cause of eclipses and limited to identifying phenomena that are to some extent correlated with eclipses?

They were able to infer, with high probability, the cause of eclipses. Thus, they were performing experiments rather than observational studies.

[Random_Australian]

40. No, the probability that the inferrence of causation is correct is not a

fundamental property of the experiment, it is a desired outcome.

The defining property still comes down to the control of the other variables, so while sometimes causation can be inferred with strength in observational analysis, the fact that you can do so is a property of the universe rather than the study. (That is, it is a property of the universe that there are very few ways in which you can cause eclipses)

Furthermore, you'll note that in both pompeii and eclipses, the people went through the various possible causes, showing some to be likely wrong until only one was left.

[Boojatta]

43. "The defining property"

Suppose a space probe is designed to be autonomous: it will go to Venus and transmit data to the Earth, but it will not receive any signals from the Earth.

Its mission is relatively simple. It will enter the upper level of Venus' atmosphere, record the distance to the surface of Venus, record the temperature, and transmit that information to the Earth. Then it will slowly descend to the surface and, as it descends, it will record the distance to the surface and temperature at that distance and transmit that information to the Earth. When it reaches the surface, its mission is over.

Given the small number of uncontrolled variables, is that an experimental study rather than an observational study?

[Random_Australian]

44. Depends what the experimenter was after. If the temp changes in the mars

atmosphere, on that day and in that place, are what they wanted, then it would be an experiment to find those.

Mind you, I'm not sure exactly how many variables are uncontrolled, so I can't be sure.

[Boojatta]

46. How do you decide what the variables are?

(A detail: the probe goes to Venus, not Mars.)

[Random_Australian]

47. Technically, anything that can change is a variable

In practical terms, only things whos change could significantly alter the outcome of the experiment is a variable, though that is often shorthand for "uncontrolled variable plus independent variable plus dependent variable"

It is not necessary to find every variable, so how many there are I do not know. What happens in science is that sufficient control trials are run such that these variables are taken into account, without ever knowing what they are.

[Boojatta]

48. If we don't know what the variables are, then how can we determine

how many there are? If we can't determine how many variables there are, then how can we conclude that the study is experimental or conclude that the study is observational?

[Random_Australian]

49. Like I said, there are two ways.

For practical purposes in things like physics, we use a number of control runs so that the variables, whatever they are, are shown to be reasonably constant over the experiment, or for things like medicine we usually use some completely uncorrelated variable, as in a random selection, so whatever those variables are, they still make for equal means.

And that's how you do it.

If you, for some reason, can't do either of those, then it is an observational study.

[Boojatta]

50. It's still not clear what "the variables" means in general.

Suppose we want to check that Kepler's third law involves the same constant of proportionality for Venus, Mars, and Jupiter. If we measure the period of orbit of a planet and then wait for it to do another complete orbit and confirm that the two times are equal within the limits of observational error, then have we confirmed that "the variables are reasonably constant"?

We recognize that checking Kepler's third law involves manipulating measurement devices, but not actually manipulating the planets. That suggests that it is an observational study. However, to make your "definition" give that classification, it would seem necessary for you to show that, when we use a number of control runs, the variables are not reasonably constant over the study.

If you want to apply a principle to both physics studies and medical studies, then you should first state the principle in general terms.

You spoke of using "some completely uncorrelated variable", but you didn't specify what the variable is uncorrelated with.

When you use the words "If you, for some reason, can't do either of those, then it is an observational study", it sounds as though you are saying that lack of an idea for manipulating data that has already been obtained might make a study observational, but that if and when an idea for manipulating already-obtained data is thought of, the study might become experimental. Is that possible?

[Random_Australian]

51. What? What? I've got your general terms, right here!

In any experiment, observation, quasi-experiment or whathaveyou we are trying to see what causes what, for any temporal science.

In any attempted experiment, we look at the data and see if there is some correlation between one variable and another.

We wish to know if there is indeed a relationship, or if there was some other reason for the correlation.

That is all the general principle that is required. As for what I said about physics and medicine, those were not absolutes. They were examples of things that were generally practicable.

Now, onto your question.

Two orbits would be somewhat tenative, though the exact equation escapes me for now. If we either A) saw some more orbits, or B) Took each datum after a small arc rather than a complete orbit, or C) could show, from already known principles, that if there were variables which have a large enough on the experiment would also do something else (which hasn't happened), then we would have a much stronger case for invariance over each one.

Mind you, the actual experiment isn't "measuring the constant" but rather measuring physical variables from which we can calculate the constant.

If we really want to know the constant, we just repeat it a heck of a lot of times. If it is the same on each trial, then we can conclude that there are no other significant variables, because this is what invariance over many samples is.

"That suggests that it is an observational study"

Actually, not only is it not about changing instruments, but the suggestion was incorrect too - if there are only two things that change across the sampling, then it is an experiment.

Furthermore, changing the instruments changes only the statements we can make about our conclusions, that is, we always say (for an observed constant) "this property is constant within the limits of accuracy of our measurements", and then the change the limits of accuracy.

"You spoke of using "some completely uncorrelated variable", but you didn't specify what the variable is uncorrelated with."

Everything. Every single variable in the entire world. (That is a very slightly idealised case, but possibly the closest idealisation to reality that you'll ever see)

"When you use the words "If you, for some reason, can't do either of those, then it is an observational study", it sounds as though you are saying that lack of an idea for manipulating data that has already been obtained might make a study observational"

Well, you've sounded it out incorrectly then; this refers only to "are you able to determine consistency over the experiment?" which is incorporated into the data taken.... so no matter what you do to the data, you've still got the original limits. You can't change observation to experiment like that.

(Of course, there is the moderately rare case in which you can show that no other known variable could account for such a thing, and that there is unlikely to be an unknown variable of sufficient magnitude to change the experiment, eg. Pompeii, one of the few observational studies to have established a very likely causal relationship)

[Boojatta]

34. A different scenario

Suppose 400 people, who were all diagnosed with the same ailment, agree to participate in a study.

Each of the 400 people is identified by means of a unique code. A roulette wheel is used to assign to each code either the color red or the color black.

Suppose that, as a result of the spinning of the roulette wheel, 192 people are assigned the color black; 208 people are assigned the color red. The 192 people assigned the color black are given placeboes. The 208 people assigned the color red are given an experimental drug.

Suppose that 206 of the 208 who were given the experimental drug become asymptomatic, but all 192 who were given placeboes continue to display symptoms of the ailment. That would seem to be an impressive result, but there is an even more impressive result that requires some explanation.

QUESTION:

How did it come to be that 100% of those who were assigned the color black took placeboes and that 100% of those who were assigned the color red took the experimental drug?


Also, if we are to explain why those 206 people who became asymptomatic were the ones who became asymptomatic, then there are some other questions that we also need to answer:

1. Why did those particular 400 people agree to participate in the study?

2. Why at the given moment in world history was that particular substance being tested as an experimental drug for the treatment of that particular ailment?

3. Why did the color red get assigned to those particular 208 people?

4. Why did 2 people who took the experimental drug not become asymptomatic?

Note: use of the term "experimental drug" does not suffice to establish that we are dealing with an experimental rather than observational study. We could just as easily have called it an "observational drug."

[Random_Australian]

35. There is one more thing too - things happen in a probabalistic sense, not a deterministic.

Before I start, I would like to point out that we obtain results not as certainty that some effect is going on, but as a probability that some systematic effect has occured.

In any experiment, we control very many variables, none to 100% accuracy, but simply to within a certain probability. For instance, in the case of the boolean variable "which tablet did they take" the chances of them taking the wrong one are made sufficiently small, say one in ten million, and then the chance that enough took the wrong tablets without anyone realising for enough people to affect the outcome of the experiment is very small.

Then, we look at the data, and see there is some very high probability that some systematic effect has occured. Of course, there is the chance that nothing had any effect, each group had the same chance of getting better (say 50%) and that we just got a fluke result, like throwing twenty heads in a row.

Of course, in the example you gave this would be very small indeed.

Now, to really focus on your question:

We specifically choose the roulette thing to be independent of the variables which affect the people, in other words while all those things and many, many more will affect the people in question, there is an equal chance that any effect that helps them get through the disease will go to each side, same with the ones that will hinder their recovery.

So when we have enough persons, this evens out (convergence of probabilities) and the effect from the last systematic difference between them (the medicine) is said to be the one causing the difference in observed probability of the person getting better.

Of course, in real life this comes under close scrutiny.... almost all of the time.

[Boojatta]

38. I don't see where you answered the questions.

For example, suppose that we are observing those who are conducting the study of the experimental drug.

The following is a fact about their behavior: 100% of those who were assigned the color black were given placeboes and that 100% of those who were assigned the color red were given the experimental drug.

What caused them to behave in that manner? For example, did a Zeus-like being who beat them whenever they tried to give the experimental drug to someone who had been assigned the color black? Did they receive food whenever they tried to give the experimental drug to someone who had been assigned the color red?

[Random_Australian]

39. Except, like I said, there is not a 100% chance that they will get the

right pill.

However, the answer to your question is essentially pragmatic:

As long as the deciding factor was independent of variables that systematically affect the experiment, simply that it is observed that almost 100% of the people received the drug is enough

Now, the other bit:

"What caused them to behave in that manner?"

Simple cognition really - once the various tasks were given (find the medicine, administer as learned, do the various checks) they were performed simply because the person desired that outcome. For whatever reason (and the exact reason must vary from person to person, however the ones you mentioned are not a deciding factor).

[Boojatta]

41. "Simple cognition really"

Don't you distinguish between behavior and cognition?

When we study human behavior, how many variables are there? In a typical observational study of human behavior, how many uncontrolled variables are there? In a typical experimental study of human behavior, how many controlled variables are there?

[Random_Australian]

42. Controlled only means that they don't affect the experiment enough to

make the outcome invalid, remember. (answering the second and second last question first)

So, all we need is a selection based on a variable independent of all the other ones, so the amount, while technically nearly infinite, is inconsequential.

Now, we come to the last question: How many controlled variables in a typical experimental study?

They all are, by random allocation. Here's the thing, though, the likelyhood of a random imbalance is inversely (and non-linearly) proportional to the amount of subjects tested, thus for a very small amount of people, the fact that all variables are "controlled" still does not mean the result is reliable. Because of this, we use p-values.

As for the first question:
Behaviour is what they do.
Cognition is what has been inferred from experiment.

A very strong link between cognition and behaviour has also been established, such that for any given behaviour it is not illogical to say that it is very likely that there was a cognitive causal antecedent.

However, it should be noted this is a distinction of little more than nomenclature; while we draw a boundary betwixt brain and body, nature makes no such distinction.

[More Than A Feeling]

18. But you still won't tell me what it is! n/t

[scavenger]

45. Could be, we only have a delusion of fee will./nt