As there seems to be some innocent (or perhaps not so innocent) confusion over what prediction means in science, I will yet again attempt to explain it. (Keeping in mind that, as Sinclair said: “It is difficult to get a man to understand something when his salary depends upon his not understanding it.”
Science is concerned with falsifiability, consilience, causality, parsimony and, among other things, prediction. Thomas H Huxley stated: “Science is simply common sense at its best—that is, rigidly accurate in observation, and merciless to fallacy in logic . . . Science is organized common sense where many a beautiful theory was killed by an ugly fact.” (One reason for my harping on fallacious reasoning.)
When I discuss prediction in science, this topic falls under the heading of philosophy of science. This is why, at least in part, there is controversy. Most people who perform research are not trained in philosophy of science. There is nothing inherently wrong with that; except when such an individual decides he is an expert on the topic and refuses to learn from his blatant mistakes. For example, I was not trained in philosophy of science but Niall Shanks holds a PhD in it and through his tutelage, and my many, many hours of study, I have become at least conversant in the fundamentals.
Philosophy of science is important. Mayr:
To replace their former chief concern, philosophers of science have specialized in elucidating the principles whereby theories or concepts are formed. They search for the rules that specify the operations by which scientists answer the "What?" "How?" and "Why?" questions they encounter. The major domain of philosophy relating to science is now the testing of "the logic of justification" and the methodology of explanation (see Chapter 3). At its worst, this type of philosophy tends to degenerate into logic-chopping and semantic quibbling. At its best, it has forced scientists into responsibility and precision. (Mayr 1998) p36 (Emphasis added.)
Keller and Lloyd continue this theme:
ke poets, and even unlike most speakers of ordinary prose, scientists expect and indeed generally assume that their language is (or at least ought to be) both precise and clear. Scientific terms are intended to mean neither more nor less than what they say, and to say neither more nor less than what they mean. In the traditional model for scientific language, at least since Leibniz, Condillac, and Pascal, terminological ambiguity, uncertainty, and double entendre are generally seen as evidence of scientific inadequacy—-as impediments simultaneously to progress and to truth and, accordingly, as impurities requiring removal . . . They [words] work to help make arguments persuasive, even to turn arguments into “proofs.” It is words that take us from the logic of predicate calculus to the logos of scientific reasoning. (Keller and Lloyd 1992)
We apply the above to the concept of prediction in Animal Models in Light of Evolution (p251-2)
In the case of predictive animal modeling, what we are typically interested in is prediction of human outcomes. The animal model systems are stimulated (perhaps by some toxicological insult of interest) and animal data is gathered. This data derived from the animal model, in and of itself, settles nothing about the actual course of human phenomena. The animal data enables the investigator to form hypotheses—expectations—about what he or she thinks is likely to happen when humans are similarly stimulated (with all the due allowances for differences in dose and so on). At this point all the investigator has is a hypothesis about human responses. The business of science is the very business of the testing of hypotheses. In the present case this requires careful studies of humans so that the human data can be compared with the expectations rooted in animal model data, thereby confirming or falsifying the animal-based hypotheses (it is also possible that the evidence gathered does not settle the issue one way or another, and hence that there is a need for more detailed studies). We again point out—to forestall a species of fatuous criticism—that not all tests and studies involving animals are done with prediction in mind. Nevertheless, those tests promoted as being predictive must be judged by how well they actually predict human response. It also makes sense to ask whether a particular method has a track record of success, and if so, how this was determined.
Keeping in mind the above, note what Quine says:
Science, for all its refinement, does not lose the common touch. The observation categorical is still the touchstone. It says that if the experimental condition is set up, observable by the scientists concerned, then the predicted observation will ensue. If the prediction fails, then the theory, which implied the observation categorical, is refuted . . . The empirical meaning or content of the theory, we might say, is the set of all implied observation categoricals . . . But prediction is always the bottom line. It is what gives science its empirical content, its link with nature. It is what makes the difference between science, however high-flown and imaginative, on the one hand, and sheer fancy on the other. (Quine 2005) (Emphasis added.)
Biology is a statistics-based science, to a large degree; hence prediction becomes a statistic-based concept. The father of vivisection, Claude Bernard, was a strict causal determinist, meaning that if X caused Y in a monkey it was also cause Y in a human.
Physiologists . . . deal with just one thing, the properties of living matter and the mechanism of life, in whatever form it shows itself. For them genus, species and class no longer exist. There are only living beings; and if they choose one of them for study, that is usually for convenience in experimentation. [(Bernard 1957 (1865)) p 111]
He was also offended by the statistical nature of medicine [(Bernard 1957 (1865)) p 9, 125, 138-9].
This false idea leads certain physicians to believe that medicine cannot but be conjectural; and from this, they infer that physicians are artists who must make up for the indeterminism of particular cases by medical tact. Against these anti-scientific ideas we must protest with all our power, because they help to hold medicine back in the lowly state in which it has been so long . . . if based on statistics, medicine can never be anything but a conjectural science; only by basing itself on experimental determinism can it become a true science . . . I think of this idea as the pivot of experimental medicine, and in this respect experimental physicians take a wholly different point of view from so-called observing physicians. [(Bernard 1957 (1865)) p 138-9].
History has proven Bernard wrong on the unimportance of statistics. The fields of evolutionary biology and complexity theory have proven his deterministic view naïve. But this is where the reluctance on the part of many vivisection activists to discuss statistics comes from. Be that as it may, one cannot avoid statistic when discussing predictability in the context of living systems.
I have gone over the basic statistics important to prediction many times hence will skip positive predictive value and so forth for the time being. Suffice it to say animal models fail to predict human response to drugs and disease (See Animal Models in Light of Evolution or “Are animal models predictive for humans?”) For example, Matthews:
When calculated correctly (see Appendix), the LRs [likelihood ratios] of the animal models examined by Schein et al. have 95% confidence intervals that fail to exclude unity for all ten of the organ system toxicities considered. In other words, the data provide no statistically credible evidence that these animal models contribute any predictive value, either separately or in combination. (Matthews 2008) (Emphasis added.)
Littman and Williams:
Given these statistics and examples [of the poor track record of animal models] it makes sense to move the rationale for efficacy of novel drug targets to one based on evidence in humans. The key questions are whether this can be done efficiently and how experimental medicine can contribute to this to provide benefit during clinical development. The hypothesis put forward here is that advances in molecular diagnostics and biomarker technologies will underpin the success of this new paradigm . . . Improved preclinical [animal] models have not materialized, and so human experimentation is still the ultimate model, although we would hope that preclinical models will in time improve. (Littman and Williams 2005)
Many studies have falsified the concept that animal models are predictive of human response to drugs and disease, yet vivisection activists continue to falsely claim that animal models are predictive. They base this, at times on a false definition of prediction. The following are not examples of what prediction means:
- Gives us a general idea.
- Points us in the right direction.
- Makes us think.
- Lets us know if we are on the right track.
- Occasionally getting it right.
- We tested 25 species and 1 ended up giving the same results as humans therefore it was predictive.
- For any given drug an animal can be found in retrospect that mimicked human response therefore the animal model is predictive.
Animal models are not predictive for human response to drugs and disease, hence the vivisection activist must either deny the obvious and treat the reader like she is uneducated or stupid, or change the subject in hopes of continuing the prediction claim with new data. Not relevant data, just new data. When vivisection activists change the subject from prediction to using animals in other ways, such as heuristic devices or as bioreactors (to grow viruses or antibodies), they are doing so either because they fail to understand what prediction means (not an intellectually flattering state of affairs) or because they fully understand it and cannot deal with the consequences and hence misrepresent reality (not an ethically flattering state of affairs).
Why is the above, and indeed all the aspects of science, including prediction, important? Thagard said:
. . . society faces the twin problems of lack of public concern with the important advancement of science, and the lack of public concern with the important ethical issues now arising in science and technology, for example we around the topic of genetic engineering. One reason for the dual lack of concern is the wide popularity of pseudoscience and the occult among the general public. Elucidation of how science differs from pseudoscience is the philosophical side of an attempt to overcome public neglect of genuine science. (Thagard 1998)
Society today faces real problems. The answers to many of these problems, at least in part, involve science and an understanding of science. When vivisection activists put salary before everything else they confirm that they do not care about the problems society faces, including illness, and are more than willing to sabotage the one mechanism society must embrace in order to deal with the very real challenges it faces. When science is corrupted by the very people who claim to be experts, society as a whole loses.
But I must give credit where credit is due. Vivisection activists in almost every university say essentially the same things that have been said in this forum and they are getting away with it. Essentially no one with any clout in the scientific, regulatory, or skeptical communities is calling them out on this.
Upton Sinclair was right.
Bernard, Claude. 1957 (1865). An Introduction to the Study of Experimental Medicine. Translated by H. C. Greene. New York: Dover.
Littman, B. H., and S. A. Williams. 2005. The ultimate model organism: progress in experimental medicine. Nat Rev Drug Discov 4 (8):631-8.
Matthews, R. A. 2008. Medical progress depends on animal models - doesn't it? J R Soc Med 101 (2):95-8.
Mayr, Ernst. 1998. This Is Biology: The Science of the Living World: Belknap Press.
Quine, WV. 2005. Quiddities" An Intermittently Philosophical Dictionary. Cambridge: The Belknap Press of Harvard University Press.
Thagard, Paul. 1998. Why astrology is not science. In Philosophy of Science: The Central Issues, edited by M. Curd and J. A. Cover: Norton.