Objective Direct and Indirect Evidence, and Subjective Inferences

I think you are referring to the philosophical doctrine of Logical Positivism/Empiricism. You are correct to say that few support it today. (See phil of science texts for why)

But epistemological doctrines in philosophy have no bearing on how science is done. The status of scientific theories in purely a matter for the relevant community of scientists.

Upthread, you mention deduction and reasoning in science. As T_a points out, deduction is not the only type of inference in science. I believe it is mainly used to deduce consequences of theories or models in order to the theories or models empirically. When in comes to creating or assessing explanation and theories, Inference to the Best Explanation is prevalent in science. I can post a bit more if you are interested.

You have your wires crossed here. If it is an inference then it isn’t deductive reasoning. Hypotheses are always an inference and are not derived from deductive reasoning.

Exactly. Every baby that is born has a genome unique to it, except in the case of identical twins. Every conception is a one-off.

I don’t think this example works too well. We can observe that the physical world exists, so that is not assumed. Also, we have tons of evidence supporting the conclusion that the universe had a beginning, so that isn’t assumed. I don’t see how it is an assumption to ask what caused our universe to begin to exist as it does now.

I would disagree. If a claim is not backed by evidence in keeping with the positivistic epistemology then the claim is considered to be non-scientific. Science doesn’t make any claims of ultimate knowledge or claim that the scientific method is the only way of being rationally justified in your beliefs. Christian scientists are a perfect example of this, being able to do science and also hold faith based beliefs.

Actually I was thinking more specifically about verificationism. Just got mixed up a bit in my terms, although positivism and verificationism if not the same, seem to be close to the same thing.

Verificationism , also known as the verification principle or the verifiability criterion of meaning , is the philosophical doctrine that only statements that are empirically verifiable (i.e. verifiable through the senses) are cognitively meaningful, or else they are truths of logic (tautologies).

Well, I somewhat agree. And although it’s not really the main focus of what we’re discussing, nonetheless, I think for a moment it’s worth addressing the effects that assumptions have on science in general.

As far as I know, assumptions inevitably are involved in formulating theories. For example, it’s been said that both Einstein and Heisenberg applied assumptions based on verificationism to certain aspects of their theories.

And even though it seems reasonable to conclude the descriptive aspects in theories, e.g., mathematical formulas, aren’t affected by such assumptions, it seems reasonable and quite evident to conclude the interpretive aspects are.

And it seems to me since the direction of scientific research and development are both built on established theories, they can be indirectly affected in a negative way if there are built in assumptions in established theories that are wrong.

So I would think there should be a way for those built in assumptions to be identified in order to be aware of their presence and be mindful if any discrepancies arise further down the road that calls them into question.

Sure. Please do. In the meantime I’ll jot down how I think it works from what I’ve read and discussed with others. From what I can gather, deductive reasoning goes from more general to more specific.

All other types of reasoning, e.g., inductive, abductive, Bayesian, etc., go from specific to general. With deductive it’s either, or, while the rest are based on degrees of probability, which I believe would also include inferences to the best explanation.

So from trying to figure it out, it seems like from a hypotheses (the general) a testable predictions (the specific) is made, and a deductive inference is made based on whether or not the evidence shows the prediction to be true. If h, then e; e, therefore h; not e, therefore not h.

Now I’m wondering if all the hypotheses involved in a specific theory are confirmed deductively if that would entail the theory also being confirmed deductively, and if so if there are any such cases where this has happened? Maybe you know something about that?

So for the specific to general inferences, I would imagine that in the process of making an observation (the specific), and from that formulating a hypothesis (the general), that any type of inference other than deduction can be used.

And I get the impression that some form of either Bayesian inference, or inference to the best explanation are the most commonly used in science today. But that’s just a guess on my part based on a few discussions I’ve had. So as far as you know, am I generally getting the drift of things so far?

From what I’ve read inferences include any kind of reasoning, including deduction. But maybe there are different ways of defining it? However, I agree that hypothesis are not arrived at from deductive reasoning. Nonetheless, I’m not talking about arriving at an hypothesis, but testing an hypothesis. You said here:

That seems like a deductive case where it’s either, or. So in testing an hypothesis would you agree that deductive reasoning is employed?

So I guess by commonalities you mean events with common characteristics? When I refer to reoccurring events the distinction I’m trying to make isn’t about an exact event that is reoccurring, but a categorically identical event that is reoccurring.

So human life beginning and developing is a daily reoccurring categorically identical event, i.e., human conception, development, etc., that science can study and acquire information about much more easily than a past event like a universe beginning and it’s initial developing stages, for which there are no categorically identical events reoccurring in the present, and in fact no other categorically identical events that we know of having occurred at all.

Oh, I totally agree with you that there is plenty of evidence for a beginning. However, it’s not unusual when discussing the issue for someone with an opposing view, like materialism or physicalism, to push back and argue that, “it can’t be proven,” i.e., it’s just an assumption. But as you pointed out, that’s arguably a well supported assumption.

However, even though I can see what you mean when you say that it doesn’t work too well, I think the issue is that we’re defining assumption differently. I see an assumption as any position held that isn’t directly stated in a claim but underlies a claim, and that can’t be proven regardless of how well supported it is. I get the impression that you are defining it as simply a position that isn’t well supported.

Now I’m not saying we should question every assumption whether or not it’s well supported. But my concern is that there are times that assumptions are made that are not well supported, or are widely thought of at one point in time as well justified that are later found to be untenable. And it’s these types of assumptions that I’m concerned with that can, at times, go undetected.

Oh, I agree with you that science doesn’t make any claim to ultimate knowledge or that it’s the only way at all to rationally justify any claim. But it’s those, including scientists, who held–and possibly hold today–to positivism, and other similar views who did or still do make those claims. And unfortunately, as I understand it, this, along with other similar views, were majority views among scientists in the not so distant past.

A particularly strong type of verificationism is part of LE: namely, statements which cannot be verified empirically are simply meaningless (except for so-called analytic cases like math theorems). Unfortunately, that restriction makes verificationism itself meaningless! So so much for that idea.

You may hear the word used today, but often it just means we need some way to test scientific claims; it is not usually meant to give a standard for judging the worth or meaningfulness of ideas.

That may be. But the key is to separate ideas that helped create new theories by firing the imagination and assumptions used to test and theories in science. In particular, verificationism did not matter for accepting GR over Newtonian gravity; it involved the orbit of Mercury and the bending of starlight in eclipses instead.

You are right that new theories often assume the truth of existing, better confirmed scientific theories. That was part of the measuring discussion as well. The key here is to distinguish philosophy like verificationism from assumed scientific theories.

But then if a test of a new theory seems to falsify it, then some may claim the issue is really with the underlying scientific theory that was assumed. To adjudicate such claims, it’s up to the scientific community and the objective process used to assess such competing claims. There is no completely deductive decision process. Instead, Inference to the Best Explanation comes into play. I’ll put that in a separate note.

You are generally right about deduction. But I would sharpen the above. Deduction produces consequences but it does not test them. Formulating and conducting tests by real world experimentation and observation takes practical skill, developed by the apprenticeshop processes of science (eg by doing post-doc work). The quality of that process and the deductions that led to it are then assessed as part of peer review in scientific practice.

IBE is to my mind the only reasoning approach used to go from the specific to the general in science (I think enumerative induction is rare in science). It can be used to create new models or theories (eg Darwin’s evolution by NS); it can also be used to select from competing explanations (eg a specific mechanism to explain the a persisting change in population genetics, like drift, NS, etc).

What does best explanation mean? First it has to be a scientific explanation. That means methodological naturalism, which has been thrashed out elsewhere, so I’ll mention that it covers practices like consistency with fundamental physics, use of efficient causation only, and complete specification of explanatory mechanisms to the extent possible (eg being specific about the nature of a proposed designer). These practices are used because they have proven successful in producing theories meeting science’s goals.

What makes an explanation best? The relevant scientific community will use some combination of these factors, again through the scientific process.

  • should be empirically testable;
  • explains the phenomena observed in existing empirical work;
  • is consistent with other related sciences;
  • has wide scope and precision;
  • appeals to plausible mechanisms;
  • simple, smooth, elegant, and non-ad hoc;
  • makes novel predictions for a wide variety of circumstances;
  • and underwrites contrasts (why this rather than that)

Why these? Because they have proven to be successful in the scientific goals of prediction, explanation, control (some might add truth, but I don’t; I see it as a result of science).

Who judges success: the concerned scientific community first, but also scientists in nearby domains, private industry through applying science to technology, and finally societies, by deciding where to spend their money. Yes, the last process involves politics. Societies that have poor politics fall behind others (eg Stalin and Lysenko). One can recognize successful science partly by whether societies adopt it, regardless of their dominant culture.

Bayesian inference can refer to two things.
First, there is Bayesian statistics. This is a particular form of statistical analysis which includes a mathematical representation of prior knowledge, in the form of a probability distribution. The other possibility for statistical work is Fisherian inference, which avoids priors and has other technical differences. It is the home of p values and null hypotheses and confidence intervals. (You may hear Neyman-Pearson in this context as well). Statistical inference processes are part of experimenting and observing. Statistical inference results go into the judgement of best for IBE.

There is also an overall philosophy of assessing via the Bayesian mindset. This is not so much a specific idea as a view that the role of evidence is to update our knowledge. It is about confirming theories. In contrast, Fisherian approaches are generally associated with evaluating theories by falsifying them. So hypothesis testing rejects null hypotheses in the standard example. The idea here is not so about knowledge update; it is about have practices which commit the lower average error over the long term.

In practice, both approaches are used.

Last comment before I take a break.
I read you as thinking that there must be some standard criteria, some fixed rules, that anyone can review and apply as long as they have some basic knowledge. I think such an idea is wrong. Instead, science is an ongoing practice by a community of trained practitioners that not only assesses explanations, theories, and models on a case by case basis, but also adjusts the standards used to do that as experience is gained in what works .

Yes. But as I pointed out, underlying assumptions based on things like verificationism don’t affect the descriptive nor mechanical aspects. But they do affect the interpretive aspects which does matter in ways that are not always so obvious.


My understanding is that IBE is a way of reasoning, so in that sense I don’t see that it would be restricted to science only, and in fact is applied in other fields like history. So I assume you’re talking in the context of doing science that it be required to be a scientific explanation. (That contextual distinction would be the same for the other criteria mentioned as well.) And I don’t agree with “assuming naturalism”. I’ll go along with the idea behind MN though, that science is only concerned with the study of the physical aspects of reality.

OK. So it is a factor that goes into the judgment.

I’m not sure that’s exactly what I’m aiming for. I think what I’m aiming for is a way for a relatively intelligent layperson to get a better understanding of what’s going on “under the hood”, so to speak, in science enough to be able to make sufficiently informed judgments about what degree of confidence to put in any particular scientific claim.

So far I’ve been gaining helpful insight from my discussions on this forum into how science is done and how much and where the areas of subjectivity are that I think have been useful for that purpose. I’m not saying how much I will succeed, but so far it seems like I’m making steady progress.

Not sure what you’re saying here. Are you saying that truth is a result of science, or the success of the criteria are a result of science?

After reviewing some different webpages, I believe you are right.

This whole time I was thinking of “inductive reasoning”, but it kept coming out as “inference”. However, I was incorrect to think that scientific methodologies are strictly inductive. It appears that there is both deductive and inductive reasoning within science.

If that is the case, then every single theory in science is an assumption. I don’t think this usage of the word has much use. Is it just an assumption that microorganisms can cause disease since we can’t absolutely prove that they do cause disease? Is it just an assumption that matter is made of atoms since we can’t absolutely prove it is true?

In these cases, it is best to directly address the weakness of a conclusion.

What we often see in some conversations is someone trying to make inconvenient evidence go away by calling it an assumption. It is this behavior that should be avoided.

The beauty of science is that people of many different belief systems can come together and find agreement. Whatever personal beliefs a scientist might hold, be it positivism or some other, doesn’t affect the science.

There are a number here at PS that are at least in effect positivists who reject the existence of the immaterial because it cannot be scientifically demonstrated. So they are making an ultimate knowledge or ultimate truth claim, even in saying that they merely disbelieve in God.

I was trying to make a noticeable distinction in my comment. As far as I can tell all theories have built in assumptions. But the way it looks to me, like a postulate, a theory itself is a distinct type of assumption that is an attempt to explain phenomenon. And just because you can’t prove it to be true doesn’t mean it can’t be true. It’s acceptance is dependent on how well it can be shown to be the best explanation of whatever phenomenon it explains.

Generally that may be the case. I think it just depends on the circumstances. Sometimes addressing the underlying assumptions is necessary to bring to the surface an issue that can’t be demonstrated otherwise.

I think much would depend on how evidence and assumption are being defined.

I would say that’s true with the empirical aspects. But I would say it’s a difference case when it comes to the theoretical aspects.

Do you think there is a difference between a conclusion and an assumption? If a prosecutor presents fingerprint, DNA, fiber, eye witness evidence that clearly links the defendant to a murder, would the jury be assuming the defendant is guilty?

There are many, many theories in science that find consensus across many different political and religious worldviews.

Here’s an explanatation I found useful:

Inference is a statement which must be true, if the given information is true. Assumption is a statement which must be true, for the given information to be true; in other words, for the conclusion to hold true. If we focus on the second point of difference above, we can see that an inference can never be same as an assumption .
Thus, from our discussion above, we can ‘infer’ that “Inference” and “Assumptions” are two mutually exclusive sets i.e. a statement cannot be both an inference as well as an assumption for the same given [premises]. Why? Because while an Inference will always be logically deductible from the given information, an assumption will never be.

So going by that I would have to adjust my previous comment to say that an assumption would not be the same as a theory, but that a theory would, it seems, have to entail the presence of assumptions whether explicit or implicit.

No. But I suspect there would have to be some underlying assumptions that they would be drawing on to reach their conclusion.

My point is that the empirical/mathematical/mechanical aspects of a theory are usually not where people differ, because those aspects are objective in nature. It’s the theoretical aspects of a theory that are usually at the core of any differences, because they are subjective in nature and more susceptible to misguided judgments, biases, etc.

It’s not uncommon for there to be more than one way to interpret the same evidence, e.g., the evidence regarding quantum mechanics has many interpretations. So although a consensus may sometimes be a good indication, it’s not necessarily the most reliable indication that theoretical aspects of a theory are the correct interpretation.

If we limit ourselves to science, the best interpretation is the one that explains the most data, makes testable predictions, and has testable mechanisms. If we move outside of science, then we enter into a wide open space where supernatural explanations can abound. So when we say that there are many interpretations, we have to be mindful to include the epistemology we are using.


I’m not so convinced about testable predictions. In the first place, as far as I can tell, predictions are subjective in nature. Then there’s the fact that there are past theories which are no longer accepted which made plenty of new testable predictions during their day that turned out to be correct.

So I’m dubious of using that as a criteria for judging theoretical aspects of a theory. Inference to the best explanation, I would say, is certainly the main criteria. I’m not sure how testable mechanisms fits into the picture here?

I’m not so sure that’s the case. Can you explain how that would work?

Agreed. But in the case of QM, the only interpretations I’ve seen are, I believe, scientific.

You can objectively model the mechanisms under question and see if observations match the model. For example, you can make predictions about the orbits of planets according to Newton’s and Einstein’s equations and see which of those predictions is supported by observation. We could also say the same for the mechanisms in Newton’s and Einstein’s models since they involve instantaneous and non-instantaneous propagation of gravity, respectively. We can design experiments to see which of these mechanisms best fits observations. The LIGO detection of gravitational waves propagating at the speed of light is a perfect example.

If you don’t have to have empirical measurements, testable mechanisms, or falsifiable hypotheses then nearly all claims can have the same truth value. DNA at a crime scene could have been planted there by a leprechaun, as an example.

You may want to check out quantum mysticism.

1 Like

Yes, but a prediction is still a subjective act as far as I can tell. At least I don’t know of any objective way to show that a theory would deductively entail a particular prediction. But even if that were the case, as I pointed out already, accurate predictions are features of past theories that are no longer accepted. So how can a prediction be a reliable indicator?

Oh, how would you make a case for that in a court of law? Supernatural doesn’t mean you can abandon logic and reason.

Equations and statistics seem objective to me. If we calculate what an orbit should be using equations, wouldn’t that be an objective prediction?

They are no longer accepted because their predictions were not accurate.

You would make the case in the same way that creationists make their case: it’s a different interpretation of the facts.

What the prediction is about may be objective, but it seems to me that doesn’t necessarily mean that coming up with the prediction is an objective endeavor. The word prediction itself is associated with the word guess.

In the stronger case, not just a new instance of an old empirical generalization, but an entirely new empirical generalization follows from some theory, and turns out to be experimentally confirmed. Instances of this are the prediction of the existence and orbit of a hitherto unknown planet by Newton’s theory; and the prediction of the white spot at the center of the shadow of an opaque disc and of the hitherto entirely unsuspected phenomenon of conical refraction by Fresnel’s wave theory of light.
Structural Realism: The Best of Both Worlds? pg. 114

Seems the predictions made were accurate in these cases, doesn’t it?

Not sure I follow what you’re saying. No matter how the facts are interpreted, it still needs to be done in a reasoned and logical manner. Seems to me it would not be possible to make a case in court for leprechauns doing anything, at least not that would still be considered within reason.

Newton’s equations did not predict the precession seen in orbits while Einstein’s equations did.

Then we would have to ask why supernatural events would not be reasoned or logical in this particular case.

1 Like

Yes, I’m aware of that. However, that’s not really addressing the issue I’m raising. The issue is that, as the examples I provided seem to indicate, there are cases of accurate predictions that still stand to this day having been made by theories that are no longer considered correct.

So that means that there can be a theory that can be incorrect, and still possibly make accurate predictions. And if such is the case, then it seems predictions are not a very reliable means of confirming a theory.

Well, for one thing, I would imagine the first line of reasoning would be to look for an explanation from a source within the natural world since that would be the most obvious place to start. So unless that has been ruled out, leprechauns wouldn’t even be a consideration.

However, if for some reason it became significantly obvious that a natural explanation wasn’t in the cards, then it would be reasonable to look around for alternative explanations. But concerning leprechauns, there’s not much of a case for their existence to start with. So that would seem to rule out that explanation pretty much from the beginning.