calhais,

I can assume my senses can give me reliable data, but it would be stupid of me to assume that they always do. More important is how I figure out when they give me reliable data, and when not.

No assumption is actually necessary. When I say that we can trust our senses I am saying that the data they provide normally allows us to drive about the city and do the other things of routine life. We trust our senses for the same reason that we trust a good doctor, namely on the basis of past experience.

Of course, I'm not claiming that sensory data always produces reliable results, and I agree with your tangential statement that it is important to identify the types of sensory data that support reliable models. (I assume that is what you meant.) Note that in order to identify sensory data that yield reliable models, we work from past experience. We see the office door and make use of our model of that part of the real world to confidently conclude that it is our exit point, that we turn the knob and open the door. We don't try to walk through the wall. Our initial exploration of the real world is what gives meaning to sensory data that might, at first, be confusing. And there is probably a genetic component as well in that some ancestral species may have initially hard-wired in the meanings of some sensory data.

Moreover, I wonder how you could know when and when not to place faith in your senses without relying on your senses in some way.

Since we are not dealing with certainty, but rather with credibility based on experience, there is no problem in using models based on sensory data to evaluate other models based on sensory data! (We lose faith in our senses if the models of reality they support fail. Our basic assumption is that our collective senses--thus excusing malfunctioning individuals--do report something about the real world. Trusting our collective senses, then, would not be about whether they fetch data from the real world--that being assumed. Such trust would reflect the reliability of those models the data gives rise to.)

In short, we would not place our trust in sensory data that contradicts sensory data having a much higher credibility. If we saw T-Rex crossing a street, perhaps after an all-night party bash, we might reasonably conclude that our senses were unreliable in producing a good model.

The first pathway you gave for building credibility was this: a model is credible [gains credibility would be better--Greensnake] when its predictions match later observations. Wonderful. That isn't the end, though. One of the central questions I'm trying to put forward here is how you know that predictions match observations. This decomposes into two questions: how do you know what the predictions are, and how do you know what the observations are?

We begin with the basic assumption that our collective senses do meaningfully relate to the real world. Predictions are deductions from a model (of some aspect of the real world) that describe the empirical data we should find if the model is right. Observations are statistically reliable reports/data that describe certain conditions in the real world. In this case those observations would be relevant to testing the model.

One scientific model held that a certain species of ant, that inhabited a trackless waste in the Sahara, found its way back to the nest by "counting" its steps. A key prediction of that model is that if those legs were artificially lengthened (by attaching little stilts) then returning ants would overshoot the nest. The scientist performed the experiment and observed that those ants really did overshoot their nest! Your concern about defining predictions and observations (as related to science) is well and good, but for our purposes it does seem like a waste of time.

Another thing that you seem not to have addressed in depth is that despite the uncertainty of the world, we must still act and make choices--we must somehow draw rather absolute decisions from uncertain conditions; it isn't always acceptable to go with what will most likely work, and yet we must do something.

A bad gambler can sometimes win by bucking the odds, but the rational strategy is always to make choices that maximize your odds at winning. Why would any rational mind bet against odds that can be improved by another choice?

[The chief strength of the scientific method is that it has refined the intellectual (and physical!) tools to maximize a credible investigation of the real world. And, we don't have to take someone's word for it.--Greensnake]

We often do have to take someone's word for it, and the appeal to authority is part of what credibility is about.

We can observe airplanes flying and indirectly partake in other wonders that would scarcely be possible if the sciences did not work spectacularly well. Which authorities should we appeal to in order to certify that experience? In science, the appeal is always to the evidence. Nobody, however famous, writes a paper and ends it by saying it's correct because he, the world's expert, said so!

The proper use of scientific authority is as a convenient handle for referencing a body of work, a reference that carries considerable credibility. Thus, we say that a solid scientific consensus wholly justifies our acceptance of the global warming hypothesis. We accept global warming not because those authorities said so but because of the highly credible work they did and the fact that they are the experts in an area that we may be unfamiliar with.

I wasn't sure what you meant when you wrote `maximize a credible investigation of the real world,' …

I meant that a great deal of research and experience has gone into the intellectual tools and methods that scientists use. As a result, they give us the best available insight into the workings of the real world. I.e., the scientific approach is significantly superior to other known methods. As to determinism, it seems that quantum mechanics has nixed that idea. There is also a paradox that I've never seen resolved. I might go with a probabilistic version of determinism.

[All of these findings fit together into a tapestry of interlocking, known facts even as they open vistas to new frontiers. That adds a large degree of credibility to them.--Greensnake]

This suggests that there is some measure of `connectivity' (a better word than `interlocking-ness') of a system of propositions that is related to their credibility. Would you explain that relationship more fully? …

Philosophers who have explored the epistemology behind the "scientific method" have spoken of such matters and you would do well to go there. I just don't see the point in digging into tedious details better covered elsewhere.

And you equated credibility with a statistical notion of a confidence interval. That is at odds with the derivation of Fisherian statistics, at least, since the confidence interval itself is founded on the assumption of a probability distribution, which isn't entailed by the colloquial use of the word `credibility.'

As indicated throughout much of my posts, I am talking about credibility. Sorry if I confused matters by using "confidence interval" here.

This is an important distinction, but you stated it vaguely:
[The currency of the real world is the degree of credibility, not certainty.--Greensnake]
You never defined certainty. Do so now.

From the point of probability "certainty" would equate to a probability of 1. Deductive logic correctly applied, in a system where the postulates are givens, yields conclusions that are certain. The area under the curve: y = -(x-1)^2 + 1, from 0 to 1, is exactly 2/3. Of that we can collectively be 100% certain.

In the real world, certainty about an hypothesis would require the elimination of all logically possible alternative hypotheses. The ability to eliminate an infinite number of hypotheses does not seem attainable, so certainty would not apply to statements about the nature of the real world. Are we going to split more philosophical hairs here? Certainty does not seem to be that difficult a concept.

Empirical evidence is that portion of the real world that serves as the ultimate starting points for a model of some aspect of the real world. A scientist observes a collection of stones whose distribution serves as empirical evidence for his model that the area was once inhabited by pre-historic humans.

PROBLEM 1 is that you haven't made a way to gather empirical evidence without relying on models; thus, you haven't made a way to gather initial empirical evidence about a natural system in a way that would justify you calling it empirical evidence rather than reasoning.

Often it is the random accumulation of facts about the real world that give rise to tentative models, so the model need not serve as an initial guide. Whether you choose to call the founding observations "empirical evidence" or not is largely a matter of semantics. The model then makes predictions that will lead to a search for the empirical evidence that determines its credibility.

PROBLEM 2 is that you didn't make it clear whether you think that inductive reasoning and empirical evidence can be the same thing.

Reasoning, viewed as an abstract inductive or deductive process, cannot be a physical part of the real world. Logic has no atoms and energy and occupies no space! Empirical evidence, as I defined it, is very much a part of the physical world. Therefore, the two can never be the same thing! The problem with philosophy is that it is absurdly easy to get lost in a cloud of details and miss the forest.

Revising a model is not difficult conceptually. You simply make changes in it to accommodate new data. It's still the current (best) model. The difficult part is in deciding whether too many of these bandages have been applied, whether the model should be scrapped in favor of a new model. That requires scientific judgment.

Now, I point out the problem: notice that model MA is the inductive reasoning used to produce and update model A.

All models about the real world are the conclusions of an inductive reasoning process. They are not the process itself. I think that you are getting buried in the cracks and crannies of a tree's bark and missing the forest. Keep in mind that in deciding how to update a model we are dealing with an algorithm. Models, as defined here, are simplified pictures of the real world--not a recipe for how to patch up a model.

[A system of pure logic without benefit of empirical data, such as the rules for chess, is self-contained. There is no logical reference to anything outside of it! The rules for chess know nothing about the rules for checkers or the rules for the game of go, much less about mountains, rivers, and valleys!--Greensnake]

I'm pointing out the adjective `self-contained' here. Define it, define `system of pure logic,' explain how something can wholly lack any connection to empirical data, and prove that all systems of pure logic are self-contained;

I have no idea why this should confuse you. "Self-contained" simply means that the system is built from deductions of postulates that are "givens." Such postulates are neither true nor false (think of a branch of mathematics or chess) and, therefore, can't say anything about the real world. Truth in such a system does not go beyond what is in the postulates. That's what I mean by "self-contained."

By "system of pure logic" I am referring to the total collection of conclusions drawn deductively from the postulates that define the system, postulates that are "givens." That is, empirical data plays no role either as inputs or as conclusions. In such systems you cannot logically arrive at any conclusions that are not already packed into the postulates. Hence, if you begin with postulates that say nothing about the real world, you cannot reach valid conclusions about the real world or those portions of it that qualify as empirical evidence.

I think that we can skip the rest of your tedious requests for definitions.

There is no such thing as logic that is inherently unrelated to empirical evidence because we are the sole bearers of logic and we ourselves are natural systems, and as natural systems we can be sources of empirical evidence about natural systems.

In what way is the logic of chess related to empirical evidence? Does it say anything about rocks and rivers? Moreover, in discussing inductive and deductive logic I am focusing on their abstract qualities, their logical properties. That we are natural systems is irrelevant to that purpose. You are conflating the logic of the reasoning process with the reasoner.

As for the intimate workings of the real world, I am more interested in what someone like Sean Carroll (famous physicist/cosmologist at Caltech) has to say than I am in what a philosopher unfamiliar with physics has to say. Sean did write a fascinating book ("The Big Picture") that might be worth your time to read.

Here's a summary in a fat nutshell of those miles of shifting sand dunes:

1. We begin with the basic assumption that our collective senses (thus excluding malfunctioning individuals) do fetch consistent data from a reality outside of ourselves. (We can identify rocks and trees, rivers and stars with reasonable confidence even if we have false ideas about them.) We have nothing to lose in making this assumption because, without it, we can't meaningfully talk to each other about the real world "out there." Sensory data is the only demonstrable input from that reality.

2. By "real world" I mean the universe of atoms and energy described in part by physicists, astronomers and cosmologists, by chemists and botanists, and by scientists in general. I include anything that can physically affect elements of that reality.

3. If a theist believes in a god who created, and can affect, the real world, then that god is part of the real world by definition.

4. We recognize and use two basic forms of reasoning: Deductive and Inductive. 100% certainty can only be had if we operate within a deductive system that is defined by its postulates. Such postulates define the system, so it makes no sense whatsoever to ask if they are true or false. You might as well ask whether the rules for the venerable Japanese game of go are true or false! A meaningless question! Branches of mathematics, chess, checkers, and go are examples of such deductive systems. Everything that you can do on the chessboard, or the go board, or everything you can discover in a branch of math, is teased out of the postulates. You can't go beyond what is in the postulates. You can't use the rules of checkers to say anything about mountains or trees or chess.

Deductive systems can be set up with postulates that are chosen to reflect the real world, but that begs the question as to whether they do, in fact, reflect the real world with 100% certainty. Thus, the real argument becomes a case of inductive reasoning over the veracity of the postulates.

Inductive reasoning about the real world begins with those recognized elements (basic facts) in the real world that will be used to reach the conclusion. Call them "empirical evidence." Our conclusion is a model (our best mental picture to explain the empirical evidence). Thus, inductive reasoning about the real world is a path that connects parts of the real world to a model about the real world. In science models are almost always mathematical and capture only the essential core, the important ideas, the main framework that allows a reasonable test of them. (The math becomes impossible if you include all the messy--and usually irrelevant--details in the model.)

5. Inductive reasoning, then, is the only form of reasoning that we know of that can say anything about the real world, and every inductive argument about the real world must be founded on empirical evidence. Thus, any valid argument about God's existence in the real world must be anchored in empirical evidence, "God" being the model (conclusion) reached. Note that God can't be reasoned into existence without a reliance on empirical evidence. Pure reason (a deductive system defined by its postulates) is a self-contained system in that none of its valid conclusions can go beyond what is packed into its postulates. And, if a set of postulates define a system, such as checkers, then they are "givens" and not something that might prove true or false. Such systems of pure reason can't say anything about the real world. A statement about the real world would have to be true in order for us to learn anything about that world!

6. An atheist, therefore, has every right to ask the theist for empirical evidence for God's existence. There is nothing logically improper or prejudiced about such a request. Attempts by theists to argue that God's existence follow from pure reason, without a peek at the real world, is nonsense unless by "God" they are referring to something that cannot affect the real world.