Do Scientific Theories Give Us the Literal Truth, or Are They Merely Useful Instruments for Making Predictions?

Introduction

Science works. When one makes a prediction using a well established scientific theory it gets confirmed by observation. A scientific theory can be labelled successful “if it makes substantially correct predictions, if it leads to efficacious interventions in the natural order, if it passes a battery of standard tests”1. So why are theories so successful? Realists explain the success of theories by

their truth. They believe science is actually describing the way the world actually is, the unobservable entities and structures that a theory posits are truly describing the actual state of affairs. Hilary Putnam argues that the success of science would be a miracle if it was not describing the way the world actually is2. Against the view that Science is giving us a true picture

of the unobservable world is that of selectionism3. Selectionists believe science doesn’t gift us

truth, rather theories are useful instruments for making predictions. They don’t agree that the success of science compels us to become realists. Rather they see the scientific environment as one in which theories are involved in an evolutionary struggle, where predictively successful theories survive while others are eliminated. Selectionists accept what science says about observables but remain agnostic about the unobservable structure, entities and processes. Bas Van Fraassen popularised this view and Bray Wray has developed and defended it4. Wray argues

1 Laudan, 1981, p23

2 Putnam, 1975

3 Wray, 2007, p81-89

4 Wray, 2010

that selectionism explains the phenomena of competitions between scientific theories and past failed theories, while realism does not, therefore selectionism is the better position to hold.

Seungbae Park responds to Wray by arguing that realists have subscribed to approximate truth

not literal truth in their charter and this can accommodate failed theories and competitions5. In

this essay I will argue that selectionism is the correct position because it explains more phenomena in science than realism. I will show that Park’s realist explanation doesn’t work because it overplays approximate truth and he thereby fails to undermine Wray’s argument for the superiority of selectionism. Additionally I will attempt to strengthen the case for selectionism by highlighting another phenomenon present in the scientific environment, constraints.

Constraints in science like those operative in the biological sphere restrict which options selection can work on, as such they can lead science away from true theories. Three such constraints I will highlight are coherence with other theories, unconceived theories and entities, and conservatism. These restrict the theory options that are available to scientists, so a theory that may be approximately true may never come to surface. If this is right then this means a predictively successful theory may be approximately true or may just be the best of a bad lot.

Constraints are a serious challenge to realists, and they seem to have no good answer available to them. I conclude that realism can’t shelter competitions, failures and the tapestry of constraints present in the scientific environment. Selectionism better accounts for these phenomena and therefore it is the correct position to hold.

5 Park, 2014

Failed Theories and Competitions

Wray argues for the superiority of selectionism’s explanatory power over realism for competitions between scientific theories and past failed theories. He considers the widely accepted theories that are no longer thought to be true to be a thorn for realists. Before the time that a successful theory was discarded, Wray says the realist would have had no choice but to call it true. Since it was discarded it puts the realist in a precarious situation. They will have to strip it of a realist truth stamp and will have to say that a theory that once described the unobservable structure and entities of the world no longer does. He points to studies of Larry Lauden6 and

Martin Carrier on the predictive success of failed theories. Predictive power is not sufficient to entail truth as some past theories such as phlogiston theory, caloric theory of heat, and Newtonian mechanics were predictively successful yet turned out to have false ontologies. Even in cases where novel accurate predictions were confirmed, the underlying theory that these were derived from turned out to have misrepresented the structure of the world. Carrier explains that there are some theories ‘in which wrong aspects of wrong theories are responsible for [novel]

predictive success7’’. Wray thinks the history of science gifts anti-realists permission to unlink

the predictive success of a theory and its supposed truth, and this is a major problem for realism. On the other hand, selectionists aren’t bothered by this phenomenon, as they can explain failed theories as a consequence of new developments within the scientific fields. As a field develops, pressures increase on theories to produce better results, make more predictions, and explain novel phenomena. Some theories can withstand these new challenges while others fail to succeed

6 Lauden, 1981

7 Carrier, 1991

and are eventually rejected. Selectionsts have an answer that is not available to realists. Wray also highlights scientific competitions as a problem for realism. Two predictably successful theories can be present at the same time. Wray cites the 16th century competition between Copernican and Ptolemaic theory of planetary motion. Both were at the time on par in terms of their predictive power yet made incompatible and radically different claims about the actual picture of the world. Wray argues that realists can’t claim that two prevailing theories are giving us true descriptions of the world, as they make quite different and conflicting claims about the actual underlying structure, entities, and processes. Again selectionism doesn’t have a problem with this. Indeed it’s compatible with selectionism to have two predictively successful which scientists would subscribe to. Selectionists are agnostic about the unobservables so they don’t have to accept that any one of the successful competing theories is actually telling us the actual state of the world. Wray concludes that selectionism better accounts for both competitions and failed theories than realism.

Park challenges Wray’s argument that realism fails to explain competitions and failed theories. He concedes Wray’s point that predictive success alone does not entail truth. He argues that the realist only has to make a link between predictive success and approximate truth, not the literal truth of a theory. He sets out to show that increasing predictive success is a reliable indicator of being closer to the truth. To see this he asks us to consider an analogy: “… suppose that one hundred people are located in Los Angeles. All of them take a step toward the White House, but none of them arrives in the White House. Thus, taking a step toward the White House is not a reliable indicator of being in the White House. Even so, taking a step toward the White House is

a reliable indicator of being closer to the White House.’’8 Park thinks the realist can admit that

the formerly successful theories which are now rejected are not literally true. Even so, he claims this does not rule out the possibility that the rejected theories were more approximately true than their forerunners. Turning Wray’s example against his position, Park says the Copernican and Ptolemaic theories are a case in point, the latter was less approximately true than the former.

Wray wouldn’t disagree that Copernican theory is more successful and more truer than the Ptolemaic theory. So Park thinks Wray then also has to concede that increasing predictive success is a reliable indicator of being closer to a true theory. This he believes is enough to show that realism can account for failed theories. On the point about competing theories Park makes a very brief response. He thinks Wray has mischaracterized what a realist would say about scientific competitions and misses the leverage approximate truth grants:“No realist would say that the Ptolemaic theory and the Copernican theory coexisted because these theories accurately represent the underlying structure of the world. Rather a realist would say that they coexisted because both theories approximately represent the underlying structure of the world, and neither theory was revealed to be closer to the truth than the other.”9 Park believes a realist can accept

two competing theories as approximately true and over time the one that is closer to the truth will

reveal itself to be more predictively successful than the other. So competitions for Park aren’t an issue for realism as Wray wants to portray. Park concludes that selectionism doesn’t have the upper hand in explanations, rather Wray simply misreads the capacity of the realist responses. He concludes that realism explains both phenomena adequately.

8 Park, 2014, p.7

9 Park, 2014, p.7

Misusing Approximate Truth

Park has overplayed the term approximate truth. He is right when he says taking one step towards the White House justifies one’s inference that they are closer to the White House than before. However, this does not justify the belief that they are in the White House as approximately true. Even if someone has left Los Angeles and has covered 40% to 60% of the distance to the White House, we can’t say that they are approximately in the White House. Even if a more predictively successful theory is a reliable indicator of moving closer to truth, we can’t know if our current theories are approximately true as we have no idea how far off the mark our theories are from the actual state of affairs. It is necessary to know the true structure of the world before we know the perimeter which allows us to start using the label approximate truth for theories appropriately. In the case of the White House we know its exact location, and we can set a precise distance at which we can start saying a person is approximately in or near the White House. With scientific theories, however, we don’t have that privilege. We may have successful scientific theories that actually are approximately true, however, we wouldn’t be able to differentiate these from other successful theories that are truer than their predecessors, but very far from being justifiably labeled as approximately true. Park may object that his analogy was not meant to justify us labelling our current best theories as approximately true, only that increasing predictive success is a reliable indicator of being closer, albeit incrementally, towards truth. If that is the case then his analogy can only support the view that a failed theory which was superseded was less true than the one that replaced it, nothing more. Thus he hasn’t really achieved much by way of this analogy in explaining failed successful theories.

On the issue of scientific competitions Park doesn’t say much. He thinks that theories coexist because “both theories approximately represent the underlying structure of the world”10. I don’t

think this answer puts the realist in a good position, because if Park was around in the 16th century when Copernican and Ptolemaic theories were pretty much on par in terms of their predictive success he would have to label both of them as approximately true. This is quite absurd, in what sense can Copernican heliocentrism be on on par in terms of approximate truth with Ptolemaic geocentrism? What does approximate even mean if it can be stamped on one claim about the underlying structure and another totally distinct one? Park’s use of approximate truth for competing successful theories doesn’t work either. His use of approximate truth is imprecise and does not motivate his account for explaining the phenomena of competitions and failed theories. Additionally there is a third phenomenon present in the scientific environment which is hard to explain for the realist, constraints.

Tapestry of Constraints

Selection in nature operates on live options that are available, these options are limited by constraints. When studying a biological trait we may be tempted to think that it is perfectly fit for the environment, however there are constraints that are tangential to its ideal adaptive form such as drift, embryonic restraints, unbreakable genetic linkages, phyletic history, bauplan, and developmental factors. Likewise even a successful theory is chosen amongst options which are

10 Park, 2014, p.7

limited by constraints such as coherence with other theories, unconceived theories and entities, and conservatism. These constraints are interwoven and amplify each other so it’s fit to label them as a tapestry of constraints. The realist has to assume that theory options are unrestricted, that we can come up with a full range of theories ranging from ones that can hardly be called approximately true to others that are very close to truth. Since the theories that are close to the truth will be more predictively successful than others, these will become accepted while the others are rejected. The existence of this tapestry challenges the assumption that scientists have conceptual access to a complete spectrum of live theory options. Constraints may well keep the predictively supreme theories unavailable, unconvinced and buried deep within the web of theories. Therefore scientists are less likely to choose theories that may be the most predictively successful because these may never come within their purview.

Web of Theories

Firstly theories are constrained by other theories. A theory is not tested on its own; rather it is expected to fit within a coherent holistic web of theories. This interlinked web is hard to untangle and theories rely on each other for direct input of variables. Such relationships can sometimes be tangential to the true state of affairs that the theory is trying to unveil. In the early part of the 20th century such a tension was revealed. For Darwin’s theory to work it needed hundreds of millions of years. To his dismay the renowned physicist William Thomson had worked out the age of the earth to be ten times lower than what was sufficient. Obviously this was a cause of Darwin’s ‘sorest troubles’, since the best physics of the day conflicted with evolution by natural selection there was good reason to doubt it. Eventually the standard view was revised and sufficient time

for Darwinian selectionism to work was granted11. If the standard view hadn’t been revised

natural selection may not have become the dominant theory to explain biodiversity. Even

hard-nosed realists would admit every individual theory, no matter its success, is prone to some error. This is why they prefer approximately true rather than literally true when they seek to describe the unobservable posits of a theory. The issue is that since each theory is sitting within a web of other error prone theories, the risk of error is compounded. Takeaway lesson from this is that theories are constrained by the developments of other theories and this can be an important factor in their acceptance. The formulation of new theories and success of existing theories is restricted by the development of other theories. Realists don’t have a way of mitigating this constraint. They can’t deny theories are interdependent or that each theory is error prone. They will simply have to accept this limitation. The realist may argue that although a theory may be limited within a web of theories, as a whole this web moves closer to the truth. This is a fair point, however there are limitations which can make progress more interdependent and sluggish. Scientists from different fields would have to sit together in interdisciplinary committees and resolve conflicts in theories. Even if such rare opportunities arose, this is hardly going to be straightforward as scientists don’t have in-depth knowledge of the nuances from other fields. A web of theories that is supposedly moving towards truth is further hindered by unconceived theories and entities.

11 Lindley, 2004, pp 164-214.

Unconceived Theories and Entities

Kyle Stanford points out that in the history of science there were alternatives to our best theories which were unconceived at the time12. Once these conceptually elusive theories became available

they were more predictively successful than previous ones. Einsteinian relativity is a prime example of this. The Newtonian paradigm, despite its enormous success, couldn’t explain some anomalous data. This data could have easily been explained by relativity but it wasn’t as the theory was unconceived of at the time. Stanford thinks this gives us reason to believe that our current best theories might have unconceived alternatives. Darrell Rowbottom has developed this argument significantly to include unconceived “entities of different types: observations, models, predictions, explanations, methods, instruments, experiments, and values”13. Since each theory

sits in a web of interdependent theories, and there may be unconceived theories and entities within the web, we have a powerful recipe for constraining the potential alternative theories that can be live options for scientists to select from. Since our theory options are constrained, theories that may be approximately true, and may be the most predictively successful are less likely to see the light of day. Therefore our most successful theories today may be the best of a bad lot.

Successful theories are selected from amongst a set of theories that have been conceived. However we don’t know the full potential set of theories as many may be unconceived. How would the realists know that from the set of conceived theories there are some that are approximately true? They wouldn’t have any way of figuring that out. If the realist is right and predictive success is linked to being closer to the truth then unconceived alternatives can take us off track from finding more predictively successful theories and hence further away from truth.

12 Stanford, 2006

13 Rowbottom, 2016, p.1

At best the realist can claim our most successful theories are closer to truth than their alternatives and as science progresses we can keep uncovering these unconceived theories and can therefore move closer to truth. This however is tacitly accepting defeat as they can’t say now that our best theories are approximately true and they also can’t say how many unconceived alternatives are out there. So they wouldn’t be able to label theories as approximately true even if there were long periods of progress and paradigm shifts. What makes this concoction of unconceived theories and entities, and the interdependence of theories worse is the increasing intellectual and theortical conservatism in science.

Conservatism in Science

For some time science commentators have expressed worry about conservatism brought about by ‘Big Science’. Historically science was done by ‘gentlemanly specialists’ who had free time, were self-funded, free of formal peer review and pressures of other scientists. They could research what they wanted, the way they wanted, without restriction. These free spirits produced paradigm shifting novel ideas and theories. From the mid 19th century onwards they were becoming increasingly rare and science started to be populated with trained specialists. States, private companies, and universities came forward to fund scientists for their own interests.

Scientists started to take wages for their scientific research, and had to conform to peer review and take into consideration the views of a large numbers of other scientists. All these changes “have served to reduce not only the incentives but also the freedom scientists have to pursue research that challenges existing theoretical orthodoxy or seeks to develop fundamental

theoretical innovation.”14 Finding novel unorthodox unconceived answers risks doing useless

research in unchartered territories. There is more incentive for scientists to stay safe and fit the parameters of their research in a way that is conducive to securing grants and the approval of the institutions that employ them. Professionalization of modern science has made it more conservative, less innovative, and more prone to the problem of unconceived alternatives than the science practised by the early gentlemen.

The amalgamation of the three intertwined constraints above means that realism is less likely to be true. This is due to that fact that we have constraints that restrict the live theory options that scientists can work with. Realists rely on predictive success to invoke truth yet predictively superior theories may be buried by the tapestry of constraints. So our most predicely successful theories may be significantly inferior to other theories that we have no idea exist. Even if increasing success is a reliable indicator of being closer to the truth we may be well off the mark even with our best theories. With these constraints in mind realists would have to hope that our current best theories just happen to be those that are approximately true. Indeed it would be miracle if this was the case! On the other hand the selectionist would have no problem in explaining these constraints. Selectionism holds that from a range of theories that are available, those that are most predictively successful survive while others are eliminated. A tapestry of constraints simply restricts the number of live theory options that available to scientists. So our most successful theories are only selected amongst those that have been conceived, there may be much more predictably superior theories buried within the pile of unconceived theories. So we

14 Stanford, 2015, p.3

have good reason to be agnostic about unobservables because we don’t know if a theory that may be correctly labeled as ‘approximately true’ has been conceived. Selectionism is hence better suited to accommodate a tapestry of constraints.

Concluding Remarks

Selectionists claim the success of science needn’t be linked to truth, rather it can be explained as a Darwinian struggle between scientific theories. Wray has defended this view by arguing that the selectionist view is better than realism as it explains competitions and failed theories in science. Park challenges this account; he argues that realism can account for both of these adequately. I have shown that Park’s use of approximate truth is not sufficient to explain failed theories and competitions. Moreover, I have supplemented Wray’s account by highlighting a third phenomenon, a tapestry of constraints. Theories sitting with a holistic web coupled with unconceived theories and entities, and theoretical conservatism blend well to reduce the number of live theory options that are available to scientists. This causes the probability of having access to a full range of theories of which some can be correctly described as approximately true to be lowered. Selectionism sits better with competitions, failed theories, and the tapestry of constraints than realism, therefore it is the correct position to hold. We are justified to be agnostic about unobservables, scientific theories are useful instruments for making predictions, nothing more. Theories work but they do not give us truth.

Bibliography

  1. Laudan, Larry, A Confutation of Convergent Realism, 1981, Philosophy of Science 48(1): 19-49.
  2. Putnam, Hilary, Realism in Mathematics and Elsewhere, 1975, Philosophical Papers vol. 1., Cambridge University Press.
  3. Wray, K. Brad, A Selectionist Explanation for the Success and Failures of Science, 2007, Erkenntnis 67 (1): 81-89.
  4. Wray, K. Brad, Selection and Predictive Success, 2010, Erkenntnis 72 (3): 365-377.
  5. Seungbae Park, On the Evolutionary Defense of Scientific Antirealism, 2014, Axiomathes.
  6. Carrier, Martin, What is wrong with the miracle argument?, 1991, Studies in the History and Philosophy of Science, 22(1), 23–36.
  7. David Lindley, Degrees Kelvin, 2004, Joseph Henry Press, pp 164-214.
  8. Kyle Stanford, Science, History, and the Problem of Unconceived Alternatives, 2006
  9. Kyle Stanford, Unconceived alternatives and conservatism in science: the impact of professionalization, peer-review, and Big Science, 2015, Synthese.
  10. Darrell Patrick Rowbottom, Extending the Argument from Unconceived Alternatives: Observations, Models, Predictions, Explanations, Methods, Instruments, Experiments, and Values, 2016, Synthese.
Share this article:

Leave a Comment