Ceteris Paribus Laws
Why does Nancy Cartwright say that “truth doesn’t explain much”? Is she right? Why or why not?
Nancy Cartwright claims that the “truth doesn’t explain much” and in her reasoning points to the inherent conflation of what is true in nature and how we explain it. The core of her argument relies on the notion that covering laws are scarce, and that most of the accurate scientific explanations are born of ceteris paribus generalizations—but these are not true laws. She opposes Hempel, Suppes, Salmon, and Hanson, arguing that because their respective inductive, probabilistic, statistical, and contextualistic models rely on the laws of nature, which are only useful in ceteris paribus scenarios, that they are not helpful in actually explaining phenomena.
Covering law theorists, and many other naïve observers, suppose that nature is well-regimented, and that there are covering laws for everything. Cartwright, on the other hand, posits that nature, much like society, is constrained by some specific laws and principles, but is not determined by them (Cartwright 161). Most of the time, for Cartwright, there is no law at all—and I agree. As Cartwright’s understanding of nature works in much the same way as society, the lack of true laws in social sciences can expose the frailties of the laws in natural sciences.
In history, for example, you could begin to explain the American Civil War by pointing to the schism between the abolitionists and anti-abolitionists of slavery, or the economic debate between industrial labor and manual labor, and so on, and so forth. Then, with these explanations, you could seek to derive a law that describes the causality of the historical explanation. However, much like predicting inflation and recessions in economics, there are no laws that can definitively predict the future any more than there are laws that definitively explain the past. Hence, if there were to be another country with the same or similar conditions as the US in the 1860s, the law derived from the American Civil War would not be able to accurately predict a civil war in that country—as there are too many unknown or unmeasured factors that differ from the US experience—and thus there can be no law at all. Instead, this “law” would merely be a ceteris paribus generalization, which can only help to explain what happens in certain, “right”, conditions where all other variables are the same (Cartwright 159). Nevertheless, and the inability to generalize an American Civil War law notwithstanding, we can still derive a descriptive explanation for why the American Civil War took place. But, as this explanation cannot be linked to any governing covering law, Cartwright’s desired divorce between describing and explaining nature takes shape.
Cartwright uses examples of Snell’s law, camelias dying and potatoes cooking faster or slower in certain conditions to shift this argument to natural sciences, but here I prefer to use an even more concrete mathematical example laid out by Bromberger; in attempting to determine why a certain flagpole’s shadow is x meters long, one can deduce and thereby explain the length of the shadow from (i) the length of the flagpole, (ii) the position of the sun, (iii) the laws of optics, and (iv) basic trigonometry (Salmon, 47). Likewise, if these factors can thereby be deemed as a law, the same law should be capable of finding the height of a flagpole given the length of its shadow. However, the law does not work in reverse. This application of classical deductive explanations underlines why covering laws are usually ceteris paribus laws and, further, why they do not help to explain the many phenomena that present themselves in conditions where ceteris paribus does not hold.
With covering laws, an explanation is just a prediction in reverse. In that sense, if a scientist knows the relevant laws, they can predict or explain an outcome. In this model, when the law is found to be true and accountable for a phenomenon, it suffices as the explanation. However, as these laws are innately constructed to be true only when conditioned by ceteris paribus, they are rarely truly laws at all. The fundamental laws themselves do not objectively describe regular occurrences; hence, these ceteris paribus laws, which do explain, are not actually laws—but they are treated as such in practice nonetheless.
I believe that Cartwright’s argument is correct, but not necessarily novel or profound. According to Cartwright, the problem arises as covering laws only actually apply in certain or ideal conditions, whereas in scientific practice they are used to explain phenomena in non-ideal conditions as well. However, laws are necessary for the natural progression of methodical science where, as Kuhn so clearly argued, science presses on by testing and retesting the established laws until such a time that they can be undermined, opening the way for a new law and paradigm. All robust and discipline science must employ the ceteris paribus rule in establishing the rules of its empirical experiments.
Indeed, a disciplined releasing of the requirements of ceteris paribus in experimental science is often the source of progress, as it allows for the testing of variables that are otherwise assumed in the course of regular law-based science.
Furthermore, Cartwright’s argument is constructed, in essence, by using counter-examples of the deductive-nomological model—which has been done before. She opposes Hempel, Popper, Oppenheim, and other covering law theorists. However, she goes a step further by also disagreeing with Suppes and Salmon, who themselves were arguing cases against the DN model. Her discord arises in her differing resolve: while Suppes proposed a probabilistic model of causation, and Salmon a statistical relevance model, Cartwright merely sought a separation between describing nature and explaining it all-together. While I do agree with her premise, it is one that science has debated since long before the DN model or even logical positivism. Hence, while science should perhaps separate description and explanation based on Cartwright’s findings, philosophers of science will inevitably continue to seek alternative ways to merge description and explanation.
Seven Sexes
What does Collins mean in “The Seven Sexes” when he says that scientists “decide the character of gravity waves” (p. 220)? Is he right?
Collins introduces his investigation into the science of gravity waves with an analogy of ships (knowledge) in bottles (validity), claiming that our understanding of the world (our understanding of ships) can only be measured by our knowledge of what is already established (the enclosed bottles). In the case of gravity waves, science cannot not work like it does with ships in bottles, because scientists are not living in a world where all the ships are already in bottles—or in an environment filled with the adequate prior knowledge with which to gain further knowledge, for that matter. Instead, scientists studying the validity of gravity waves must construe what little information there is, and build apparatuses based off that information to prove or disprove a phenomenon that cannot be decidedly proven or disproven with any given experiment. This dilemma, argues the sociologist Collins, is conceived from a lack of agreement between scientists in the field, as any apparatus incapable of measuring a gravity wave can be deemed as a poor apparatus, instead of disproving gravity waves; and, likewise, any apparatus that does successfully measure a gravity wave can be disputed as having too miniscule or too inconclusive results, instead of proving gravity waves. I will argue in this paper that Collins is right in arguing that scientists “decide the character of gravity waves”.
In building his case, Collins likens scientific knowledge to a cultural artefact. Through Collins’ sociological perspective, the gravity wave scientists are still developing the terms and objects of their culture. Without these terms and objects, the scientists have no way of agreeing on what constitutes a gravity wave measurement, what apparatus is best at measuring that undetermined measurement, and therefore what can be seen as proving or disproving the evidence.
In other scientific inquires, scientists construct carbon copies of successful experiments to prove results and add further knowledge to their field. In the Gravitational Radiation field, however, scientists are reluctant to build such carbon copies. Firstly, finding results consistent with those of the originator’s apparatus would award credit to the originator, and not the facsimile (Collins 210). In this case, if the scientists were able to produce even more convincing positive results with a new apparatus, then the scientists could be seen as pioneers progressing the field; on the other hand, if they found convincing negative results with a new apparatus, they could then be seen as better experimenters than the originators and therefore, for the time being, be credited as disproving the existence of gravity waves.
Secondly, the scientists are reluctant to build carbon copies because they feel the originator’s apparatus can and, perhaps, should be improved. One of Collins' interviewees argued that carbon copies are not only boring, or lack creativity, but that scientists should use the parts of the originator’s apparatus that they are not experts on, and add or improve parts that they are experts on (Collins 211). The problem here, admit the scientists, is that it may or may not be true that the new apparatus is better that the last. This doubt arises because the scientists are, in the first place, not even sure that the originator himself has built a “good gravity wave detector”. Therefore, the scientists should have no reason to build an apparatus isomorphic to the original. Moreover, as I have stated, if their new devices do not “detect gravity waves”, then their failure does not necessarily deem them as ineffective scientists, as they may have legitimately found evidence to disprove gravity waves and Einstein’s general theory of relativity (Collins 211). The point is, there is no way of knowing.
For Collins, these reasons open the gates to a multitude of further dilemmas. As the scientists themselves cannot agree on what constitutes a “good gravity wave detector”, then any results produced by those detectors cannot be agreed upon as “good” or “bad”. Here, some scientists point to possible discrepancies in computations, one even stating that "you don't know whether those girls [the computers] were talking to their boyfriends at the time” (Collins 212). Others disagree about the importance of results coinciding with sidereal versus solar days, and still others disagree over the importance of having detectors thousands of miles apart. Because the scientists cannot even agree on prioritizing which results should be more or less convincing than others, or if the results are even valid in the first place, Collins finds that no consensus can be drawn on what results are “good”, on what apparatus is “good”, or, therefore, what experiment is “good”.
After pooling his discoveries, Collins posits that, because scientists cannot agree on what constitutes a good experiment at this stage, the scientists' actions should be seen as negotiations over what constitutes a competent experiment, as opposed to the replicating or testing of results. Therefore, in deciding what constitutes a good experiment, Collins claims that the scientists are “deciding the character of gravity waves” (Collins 220).
Given Collins’ expansive report, detailing the uncertainty in the field of gravitational radiation, I agree that the scientists were, in many ways, deciding the character of gravity waves. As the devices were unproven, the results highly debated, and the scientists unsure of the field’s direction, all of their attempts at replicating and testing experiments were futile. Nevertheless, as their many fruitless experiments and apparatuses led to communication, and their communication led to negotiations over what actually constituted a viable experiment, the scientists were inadvertently contributing to the nature and character of gravity waves. By this, I do not mean that the scientists were deciding whether or not gravity waves exist; rather, they were conceiving the terms and objects necessary to understand and determine how a gravitational wave could or could not be measured. With these terms and objects, the scientists would confirm their bottle (validity), in which to design their ship (knowledge), even if the implications of these conclusions are a quietism which seems tragic and contrary to the progressivism and optimism that science most often likes to identify with.
Diversifying Research Portfolios
Sketch the argument (presented by Strevens) that it is good for society to “diversify its portfolio” in deciding what research programs to support, that is, that it is usually better to spread resources among several competing research programs than to concentrate them in one program, even the most promising program. Now suppose that one research program R is doing much better than the others. (That is, scientists are becoming more and more convinced that R’s central tenets are true, and those of its competitors false. For a few relevant comments on this possibility, see Strevens, “The Role of the Priority Rule”, p. 67). On any of the reward schemes discussed by Strevens, there will tend to be a convergence of resources on R. But this will result in the research portfolio rapidly becoming less diverse. Is this always a good thing? Can you think of a case where it would work out badly? In such a situation, what aspects of science’s social organization might slow (intentionally or otherwise) the convergence of resources on R.
Strevens argues that while science aims to provide collective recourses to all research programs, it struggles to do so because science is innately as competitive as it is cooperative (Strevens 55). Scientists seek, first and foremost, prestige; in deciding between any two research programs, whether they share the same goal or not, the scientist will choose the one with the highest payout in terms of such prestige. In making this decision, however, the scientist must debate the probability and size of the potential payout, which leads to further complications, calculations and competition. Nevertheless, as Strevens argues and I will layout in this paper, reward schemes allow for an efficient distribution of recourses, both in terms of rewarding scientists for their achievements and for the greater good of society. In particular, I will focus this paper on the distribution of recourses for competing research programs, arguing Strevens’ case that it is good for society to spread recourses among several competing programs, rather than just the most promising one.
While common sense perhaps urges a devotion of recourses to the research program with the highest potential of realization, Strevens makes the case that those recourses should be spread among the several competitors as well, essentially allowing for a more “diversified portfolio”. Strevens likens this scenario, in which several research programs pursue the same scientific goal, as so often happens, to a winner-confers-all benefits race (Strevens 68). In this race, only the research program that is successful, and first at being successful, is rewarded. This means that, in the aforementioned debate between a program’s potential and the size of its payout, a scientist will place more importance on the program’s potential. However, as Strevens also emphasizes the economic notion of “decreasing marginal returns”, each additional labor input will provide less and less of a impact on a program’s potential (Strevens 63). With this in mind, Strevens imagines an ideal “central planner” who would allocate each additional recourse to different programs so that the marginal returns from each program would be about equal (Strevens 64). Even without a “central planner”, Strevens argues that this allocation occurs on its own, as scientists are rewarded based on their contribution to their program's probability of success, and so scientists will begin to choose programs that give them the higher probability of prestige not only in terms of a program’s intrinsic potential (Strevens 70). This model, however, is most ideal for Strevens’ additive case, in which programs have independent objectives. For the winner-confers-all benefits race, the effect of decreasing marginal returns (Marge) is somewhat different.
At this point, Strevens transforms Marge into science’s priority system. Here, society aims to extract the greatest return from a winner-confers-all race; and, this happens when the allocation of resources among programs maximizes the probability that at least one program succeeds (Strevens 68). This probability increases, in a winner-confers-all race, in such a way that the distribution of recourses favors the higher-potential program more heavily than it does the distribution for the previously discussed additive case. Here, while recourses will still be spread out, more recourses will be allocated to the program with the highest potential. Therefore, in a winner-confers-all race, as a program R begins to have an increasingly high potential, more and more recourses will be devoted to its efforts and, thus, lead to a less and less diverse research portfolio.
This transition is not always negative, but it can be. To see how the convergence of recourses towards R could have a negative impact, I turn to Harry Collins’ investigation into gravity waves. In Collins’ research he found that different groups of scientists, all working to achieve the same goal, could not agree on anything as a whole (Collins 212). They all built slightly different apparatuses to measure the gravity waves, had slightly different opinions on what actually constituted a gravity wave, and so on. Therefore, if the scientists were to agree on which apparatus was most capable of detecting a gravity wave, they would then find their conclusion as to whether gravity waves exist or not, at least temporarily. If R represented the program that had the most advanced apparatus with the most trustworthy scientists, and the other programs agreed on this, then they would build their own apparatuses to resemble that of R, and recourses would converge to R. Then, suppose that R found no measurements of gravity waves. This would lead the field of Gravitational Radiation to come to an agreement that Einstein was wrong. On the other hand, if the scientists continued to advance their own experiments and not converge on R, then they would come to a completely different conclusion.
In this example and others, however, science’s social organization would slow or even prevent entirely the convergence on R. In Collins’ own research, he found that scientists would not agree on which apparatus was best, and therefore not converge on R, because each scientist wanted, first and foremost, prestige. Relating this back to Strevens, this winner-confers-all scenario goes hand in hand with the notion that science not only should diversify its portfolio, but also that it often does—and does so naturally.
This distinction between a winner and a loser in scientific race, where only the winner is rewarded, can benefit scientists and society as a whole. Because of the nature of competition, notwithstanding scientists’ risk aversion, science will naturally find an efficient balance of recourses so that scientific objectives are realized. This balance, I think, relies on the initial reluctance of scientists to converge as to avoid a loss of prestige. Nevertheless, as a program becomes gains increasingly high probability, scientists will converge for the very same reason, but at this point only to salvage prestige.