The Revival of Instrumentalism

The scientific realism debate seems to take a neo-instrumentalist turn. After Kyle Stanford's Exceeding our Grasp (2007), in which the key argument for instrumentalism of a short was based on the claim that scientists qua theorists have certain limitations in conceiving alternatives to current scientific theories, the realism debate was dominated mostly by arguments as to how best to block the pessimistic induction and to defend selective realism. In the last two years, however, instrumentalism (or scientific anti-realism, in general) has staged an impressive comeback, mostly by means of two books: Brad Wray's Resisting Scientific Realism (2018) and Darrell Rowbottom's The Instrument of Science (2019). Though there are significant differences in the kind of anti-realism advanced by the two authors (as is evinced, for instance, by the reviews of each other's books in Notre Dame Philosophical Reviews), there is a common theme and a common strategy: both philosophers present alternatives to realism based on the claim that science is not in the business of discovering truths about the world (especially of entities and properties that lie beyond the limits of the visible); and both philosophers mainly aim to rebut key arguments in favour of scientific realism and/or to defend key arguments against scientific realism. What seems to be missing is a positive argument for instrumentalism, viz., an argument which is not based on making scientific realism unattractive but on making instrumentalism attractive and philosophically compelling. 


In this paper, I am going to argue that there is no positive argument for instrumentalism in these two books and that this shows that there are certain inherent limitations in adopting instrumentalism as a philosophical theory of science. The case for instrumentalism, in other words, will depend on the strength of the arguments against scientific realism. But these arguments are vulnerable to significant criticism. 


The key reason why there is no positive argument for instrumentalism is that in Wray's case anti-realism is motivated by scepticism; but the weaker position that Wray intends to defend (which, as Wray admits, is Constructive Empiricism) is equally vulnerable to sceptical challenges. While, in the case of Rowbottom, the key reason is that his "cognitive instrumentalism" requires an epistemic dichotomy between unobservables and observables, which simply cannot be sustained.


Finally, I will rebut three arguments. Rowbottom's 'methodological argument against scientific realism; Rowbottom's argument for semantic instrumentalism and Wray's argument from the radical theory change.

Instrumentalism is the view that scientific concepts and theories are merely useful tools that help scientists in their efforts to predict, understand, and control phenomena of interest, rather than more or less accurate representations of such phenomena (cf. Rowbottom's characterisation of cognitive instrumentalism in Rowbottom 2019). Hence the term: scientific concepts and theories are effective instruments that further scientists' epistemic and practical interests and do not, or not necessarily, or not for the most part, provide an adequate depiction of reality. Another way to cash out the difference between instrumentalism and its main opponent, realism, is thus in terms of the aims of science. While instrumentalists and realists agree that predictive success, understanding, and the successful control of phenomena of interest are among the aims of science, the instrumentalist denies, and the realist accepts, that truth (concerning both observable and unobservable aspects of reality) is also among these aims. 
This way of defining the view creates a potential problem for instrumentalists. There is no doubt that the scientific understanding phenomena of interest is one of scientists' main goals, as evidenced both by scientists' own pronouncements as well as by philosophical discussions of the aims of science that are independent of the realism-instrumentalism debate (see for instance Baumberger et al. 2017). However, if, as can plausibly be maintained, truth is a prerequisite for adequate understanding, i.e., if for an account of phenomenon of interest p to yield understanding of p, the account must be (approximately) true, truth enters through the back door. Thus, if understanding is maintained to be factive a dilemma ensues for the instrumentalist: she either has to give up understanding as a central aim of science or else accept that her position collapses into realism. 
The main goal of this paper is to resolve that dilemma by developing a theory of scientific understanding that gives up the factivity requirement. Its core idea is that scientists regard accounts as providing understanding if the accounts make essential use of concepts and ideas that achieve predictive success and success at controlling phenomena of interest in the long run. A good account, on this view, is not one that merely 'saves the phenomena', but one that employs concepts and ideas that are effective in realising scientists' practical goals over time. I will defend my account against both factive (such as Strevens 2008 and Kvanvig 2009) and non-factive alternatives (such as Elgin 2007 and De Regt 2015).

Semantic instrumentalism of the form defended by the logical positivists – according to which discourse about unobservable things should be construed as elliptical discourse about observable things – is widely agreed to be untenable. But as I argue in The Instrument of Science, one might instead adopt a more moderate empiricist position on scientific discourse, which I call property instrumentalism. The basic idea behind property instrumentalism is as follows: talk of an unobservable thing may be construed literally only to the extent that said thing is ascribed observable properties (or described via analogies with observable entities). In this paper, I will explore to what extent property instrumentalism is compatible with the alethic, metaphysical, and epistemic components of scientific realism. In doing so, I will also explore exactly how property instrumentalism relates to 'instrumentalism' construed more broadly. I will contend that scientific realists might reasonably 'water down' their positions on scientific discourse, and accept property instrumentalism while retaining the core elements of scientific realism. I will show how the 'no miracles argument', or an argument approximating this, might still be endorsed by such realists.

Lessons from the Strawman Instrumentalist


Recently, anti-realist positions, including instrumentalism, are enjoying a renaissance of sorts (see Stanford 2006; Rowbottom 2019). But what instrumentalism entails still remains unclear. Current varieties of instrumentalism are plagued by the instrumentalism of the past. I aim to examine the history of instrumentalism with the ultimate aim of clarifying what a plausible instrumentalism would look like.


Instrumentalism became the foil for scientific realists in the 1950s and 1960s (see Smart 1963; and Popper 1956). Karl Popper and other argued that instrumentalism was widespread among physicists. Niels Bohr and other physicists associated with the Copenhagen Interpretation of Quantum Mechanics were often cited as typical instrumentalists. Popper was especially concerned about the Copenhagen Interpretation's appeal to the Principle of Complementarity. It "amounted to a 'renunciation' of the attempt to interpret atomic theory as a description of anything" (see Popper 1956/1963, 135).


I aim to show that the instrumentalism of the 1950s and 1960s was a strawman position, created by scientific realists, for their own rhetorical purposes. First, I argue that Bohr is not obviously an instrumentalist in the sense that Popper and others suggest. He did not appeal to the Principle of Complementarity for the purposes for which Popper alleges. Nor did Bohr's appeal to the Principle have the consequences that Popper suggests. Second, I argue that it is unclear whether any philosopher of science identified as an instrumentalist in this period. I then explore what a plausible instrumentalism might look like.