Applying Science Discussion

Based on discussion originally under GodLanguage:

{Your view of the role of science is naive and, frankly, incorrect. You think the role of science and logic is 'discovery' of concepts - things that can be demonstrated. What science does is 'kill' models and concepts. Science is a rather negative discipline that forever says "you're wrong" and never, ever says "you're right". The best you can hope for science to say is: "I haven't figured out why your hypothesis is wrong. Yet.", and for it to say the same thing for long enough - under enough new observations - that the hypothesis becomes a theory. Further, from GoedelsIncompletenessTheorem and RicesTheorem, much the same can be said of mathematics and computation: math can prove some true things true, and some false things false, but there will forever be a gap where one cannot prove anything at all. That you imagine logical models derived of science would find a "Single Right Way" (or establish any "Right Way") mostly tells me that you really, truly, don't grok science.}

You are wrong about what science is. Either way, just be clear why your pet is better and show specifically where, specifically why, and specifically how it is better, and not just some minor variable in lonely isolation. Whether that's part of "science" or not, we'll save that def fight until AFTER you do the basics. -t

If your correspondent is "wrong about what science is", which I dispute, would you be so kind as to present what you believe science to be?

See "Science is Testing Models" under WhatIsScience. The "real world" in software engineering usually relates to matching requirements (solving problems) and the economic issues of software production, maintenance, and reliability. -t

Given "Science is Testing Models" (from WhatIsScience) I don't see why you disagree with the above definition. The two definitions are the same. The result of testing a model is that it is either preserved (for future tests!) or discarded, or as the author above put it, killed.

Testing can produce both pro's and con's for models. It doesn't inherently lean one way or the other. I will agree that some models are hard to test to gain either pro's or con's and some domains or models may have rough spots that may lean in the pro or con side. -t

That is neither the Scientific Method or a scientific method. Science is the process of rejecting hypotheses (a model is a hypothesis); science does not produce evidence in favour of models, though it is true that science does not "inherently lean one way or the other" because it "leans" against any and all models. As soon as repeatable evidence is found to invalidate a model or hypothesis, it must be rejected. Evidence in favour of models is provided by imagination, intuition and/or observation.

If model A has proven via testing to have predictive ability (matches real world) and model B has yet to be tested, then most rational people would agree that A is the better model of the two so far. And "rejected" is usually a matter of degree in practice. -t

Further, consider model C which has 8 matching observations and 2 non-matching observations compared with model D which has 1 match (confirmation). You seem to be suggesting that D is the automatic "winner" because falsifications exist for C, while I'd argue that C is the best so far because it has been tested more even though it has known imperfections. -t

  Hypoth..Matches..Non-Matches
  ----------------------------
  C..........8.........2
  D..........1.........0

I suppose one could argue that the "score" depends on one's needs (oh, oh, ItDepends again). If you want a "perfect" hypothesis, then C has "failed" because it has known falsifications. However, it may still be the more useful model of the two even with its imperfections. If one is looking for a theory to explain the underpinnings of something, then perhaps both hypotheses should be rejected (or at least kept at arms length). However, if you merely want to predict nature, then C is arguably at least as strong as D. This issue was also raised in AddingEpicycles in relation to regression formulas. As a predictive model, epicycles proved useful. As an explanation model, it proved poor. An "explanation model" is one that attempts to mirror a mechanism to explain something. Contrast this with something like polynomial regression which is a tool only meant to predict, not explain. -t

All very interesting, but what you're actually describing is social behaviour in relation to belief systems. It is not science.

I'm not sure these are fully distinct. Science has a "goal", and that implies a relationship with social behavior because defining the "goal" is not as simple is it might appear at first. For example, is science's goal to merely predict nature, to explain nature, or are these interrelated? I agree that explaining nature is the "better" goal, but not necessarily the only goal. In practice, most people care more whether the weather forecast is accurate rather than whether the model is "right". "I don't want a lesson on fluid dynamics, I just want to know if it will rain on my wedding day." Ideally we'd want both, but having just forecasting is not an outright "failure" of science to most people. The Newtonian physics model has been proven "false" by what appear to be your standards. Yet most agree it is very useful and we still rely heavily on it for things at human scale.

The goal of science is discovery, explanation, and accurate prediction. Newtonian physics is indubitably false in every respect. I agree that it is useful, but that does not diminish the fact that science has proven it invalid. Its inaccuracy, however, may be acceptable for practical purposes. Such applications are not science per se. They are technology or methodology.

This issue may be the crux of the problem here. Almost no practical software abstraction is "perfect". Thus, our abstractions fail in a similar manner to how Newtonian physics "fails". However, it's still a very useful tool because its accuracy is "good enough" for most situations, yet is a far simpler model to work with for most human beings than the alternatives. The level of usefulness of a software abstraction is often the main question and the sore point of debates. Except for a handful of "purity zealots", the utility of a UsefulLie is a far more important issue than the fact it may be a "lie" (less than perfect). Related: WorseIsBetter. It may not be economical to go from 99% accurate to 99.999% if our model/tool becomes 10 times more complicated. The practice of software engineering is about weighing trade-offs, and NOT about "right theory" versus "wrong theory". Owners and customers will often accept imperfect models if it keeps their costs down. I'm not promoting sloth here, only saying that science and societal decision-makers often have different goals. I see science's role in S.E. that of providing better tools to understand and forecast the trade-offs, not necessarily producing the best model of reality. -t
- {Please try to avoid using the phrase "purity zealot" to identify people who consistently disagree with you. It's inaccurate, when what you really mean is "people who disagree with you." I've no seen any "purity zealots" here. I'm not even sure what they'd be -- maybe people who insist on using SKI combinators to write videogames?}
- Can you suggest an alternative? There are indeed "zealots" here from my perspective, but I will strive to use a more neutral phrase if it still conveys what I want to say.
- {I did. I suggested referring to them as "people who disagree with you", because that's the context in which you invariably invoke the "purity zealot" appellation.}
- No, a lot of people disagree with me for difference reasons on different issues. Not specific enough.
- {Be that as it may, "purity zealot" conveys no useful information and is intentionally insulting. However, aren't your insistences on TableOrientedProgramming or maintaining bags in database systems with no benefit, forms of zealotry?}
- How about "purity proponents". Is that acceptable?
- {No, because it implies that purity is sought for its own sake. Outside of art and pure mathematics, that is almost never the case.}
- Their weak practical justification makes it appear it is for "its own sake". Anyhow, this is growing into a sticky mess here. I'll see if I can re-word the original statement...

Perhaps another way to say the same thing is that science does not by itself "rank" competing hypotheses. It does not "fail" a hypothesis; it only offers information about it. The ranking techniques applied by humans will generally depend on what they are trying to achieve. In other words, "Is hypothesis X useful for need Y" is outside the "strict" interpretation of the scientific process. -t

Science does not rank hypotheses, but it certainly does "fail" hypotheses. It can never perfectly confirm a hypothesis, because there is always the possibility that a repeatable experiment might yet be discovered that disproves, or fails it. Ranking of hypotheses is a matter for the human application of models and/or scientific results, e.g., engineering. In itself, such ranking is not science per se.

I'm not sure we can completely separate humanity from "science". Science is in the mind, and only humans have a "mind" of the kind that can comprehend science as we know it. For example, we explore Mars more than Venus because Mars is more Earth-like. The universe doesn't "care" more about either planet. How we conduct science is influenced by human issues. (The Soviet Union was more interested in Venus than the US, but some suggest it's merely because they had more early successes there, and built upon them.)

If there are two competing theories, to conserve resources we may choose what we see as the "best so far". Thus, human economics comes into play. The universe doesn't otherwise "care" that we have limited resources to explore competing theories. Science is a tool, and humans use it not much differently then they have any other tool going back eons.

If you can envision "science" outside of human bias, I'd like to hear it. In practice, any conceivable "machine" that explores and learns is going to have limited time and/or resources and thus must choose what to explore and thus in some way rank theories. There is a "science" concerned with making decisions based on limited resources: it's called "economics". And, economics involves politics and WetWare. We can't hide from those just because they are "messy and hard". Many are of the mistaken notion that one can sit in a quiet corner and derive the One True Formula for software engineering without going outside and doing field-testing. This was the "Greek fallacy" (EvidenceEras). -t

Pure science exists, by definition, outside of human bias. Unfortunately, current technology means Science can only ultimately be conducted and its findings disseminated by humans. Therefore, the process of Science encourages repetition, duplication and competition in order to reduce the effect of human bias. Certainly the interaction between humans and an idealised Science is worthy of study, and it is studied in everything from branches of statistics to the philosophy of science.

In practice, we are stuck with the science that is filtered through human eyes.

CategoryScience, CategoryMetrics