Induction, Experimentation and Causation in the Social Sciences
Round 1
Reviewer 1 Report
Referee report for:
Type of manuscript: Article
Title: Induction, Experimentation and Causation in the Social Sciences
For Special Issue: The Problem of Induction throughout the Philosophy of Science
There seems to be little that is not common knowledge in this paper. The paper details no interesting advance over what everyone already knows.
Moreover, the author does not, in my opinion, take seriously the vast difference in complexity between, say, an atom and, say, the economy of Brazil. Given the shocking complexity of the social, it is a wonder we have any science of the social at all. In distinction, the author says “I am not convinced by this argument [the argument that the complexity of the social makes it ‘in practice impossible to make successful inductions in these [social] disciplines’ (line 129ff)] because the natural world is equally complex [as the social world” (line 132ff).
By “natural world,” the author means that which is studied by the sciences of physics, chemistry, geology, biology, etc.
Yes, the natural world is very complicated. But it is simply and obviously false that the natural world is as complex as the social one. Consider that we have the Large Hadron Collider: the world’s most sophisticated machine. This accelerator was used to find the Higgs Boson. A type of particle predicted by the Standard Theory back in the 1960s. Finding this particle in 2012 was a great advance in physics. No social science can even come close to such audaciously precise predictions and to have such predictions verified decades later. Or consider the Hubble Space Telescope. This mighty machine has looked back almost to the beginning of time. The HST has revolutionized our thinking about the cosmos. Nothing in any social science can some close to such success. Clearly complexity is directly implicated.
Consider the terrorist attacks on Sept. 11. The United States had at the time an army of intelligence researchers looking for anything like a terrorist attack. So it isn’t true that the US was not looking for an attack of some sort. Yet one slipped through — and a big one, at that. Why? Because all the supercomputers in the world could not model Bin Laden’s terrorist organization. The Sept. 11 attacks are routinely cited as an example of a black swan — an event so out of the norm that it is not even predictable. The presidency of Donald Trump is likewise labeled by black swan by many types of social theorists. Black swans abound in the social realm. And, yes, the natural sciences are surprised by black swans from time to time (the discoveries of both dark matter and, especially, dark energy are relevant examples), but black swans show up in the social sciences considerably more often than they do in the natural sciences. Why? Because the natural sciences are not that complicated. Compared to the social sciences, the natural sciences are easy.
What the social sciences really need to do is to take their own complexity seriously.
In the conclusion of the paper under review, the author does not really have much to recommend to us. The reason for this is that the author has vastly underestimated how complex the social world is.
Author Response
Please, see th attachment
Author Response File: 
Author Response.pdf
Reviewer 2 Report
This paper argues that the distinction between natural the social sciences lies in the fact that for the former, but not for the later, we have succeeded in finding causal relations that allow manipulation. This is because experimentation is very limited in the social sciences, while the same is not true for the natural sciences. The paper ends by stating that we should aim for controlled experiments in the social sciences.
The paper is clear and the topic interesting, but I feel it does not add much to the current scholarship, plus it seems to overstate the main claim beyond the limits of what the paper argues:
1) The paper argues for a distinction between the natural and the social sciences, but it is surprising that the paper only talks about physics and economics. This leads to a very strong conclusion which is clearly unsopported by the evidence. Additionally, the paper fails to pay attention to other authors who have discussed precisely the analogy between physics and economics in detail. I recommend that the authors read Cartwright's "The Dappled World".
2) The paper contains some very dubious statements, some of which are false. At some point, we are told that laws in physics are different from regularities in the social sciences because the former are not ceteris paribus. This is plainly false, as a century of research in the philosophy of science has demonstrated. In fact, more recent views suggest that lawhood comes by degrees, even though all are ceteris paribus. I suggest that the authors read Mitchell's work on pragmatic laws, as well as some of the work of Cartwright ("How the laws of physics lie") and Giere ("Science without laws").
3) The paper ignores that most research in the biological sciences is based on the use of regression analysis, and that no serious laws have been established in biology, despite the existence of controlled experiements (e.g., in molecular biology). It is ok to do so, but then the conclusions of the papers must be limited to physics and economics, rather than to the distinction between natural and social sciences.
4) The paper is overall very poorly referenced.
- For example, we are told that social sciences are new disciplines whether natural sciences are historically older, but no references are added to support such a strong claim. I say strong on purpose, because I am not certain Geology was established as a discipline until the 2nd half of the 20th century, whereas it seems clear that economics was established in the 19th century.
- Another example, in p. 2 we are told that not everyone agrees that the social sciences should provide guidance for future actions, but only one paper in econometrics is cited. I am sure Popper said that in "The Poverty of HIstoricism", and possibly other did as well. But some authors should be credited for such strong statement.
- In p. 3 we are told "Many in the debate about methodology in the social sciences have argued that social phenomena are much more complex than physical ones, hence it is in practice impossible to make successful inductions in these disciplines.", but we are not told who are these many.
5) The paper puts forward some definitions that seems inadequate or unjustified.
- In p. 2, the authors define "understanding" as "knowing the meaning of certain actions". But this is a peculiar definition, which is at odds with the debate on understanding in contemporary philosophy of science. This is ok, but the authors should justify that they are defining the concept as it is used in Weber's thought, which is what I suspect the authors are doing. I am surprised also that if the authors are making the distinction between social and natural sciences as made by Weber, they fail to cite Windelband and Rickett, which are known to have made more or less the same principled distinction before Weber.
- In pp. 6-10 the concept of "cause" and the manipulation criterion is discussed, but Woodward is not cited, him being one of the main actors in this debate in contemporary philosophy of science.
- At some point, the authors make a distinction between laws and causation (p. 11, and I think before too). But this is false. If laws characterize counterfactual dependencies and causes do as well, this distinction that the authors make is incorrect, for both laws and causal claims would have exactly the same truth conditions.
Overall, I think the paper needs very serious revisions.
Author Response
Please see the attachment
Author Response File: Author Response.pdf
Reviewer 3 Report
Author(s),
The paper helpfully lays out how common social science research designs map onto the philosophical literature about causation and probability. I imagine that introductory and advanced philosophy students as well as scientists could glean a good deal from this paper.
One opportunity to improve the paper would be to elaborate on the notion of manipulation and its relationship to causation. Footnote 5 states, "I don’t think we should state manipulability as a necessary condition for being a cause." The literature on manipulation and causation is vast, but readers would benefit from knowing that this has point has been made in that literature. For a general discussion of manipulation's relationship to causation see Woodward (2016) and for a discussion of manipulation's relationship to causation in social science specifically see Byrd (2019). In general the paper would benefit from pointing out some of the potential (and underdetermined) interpretations of successful manipulation (e.g., Byrd 2019, Figure 3) and/or relaying the distinction between affirmative and negative manipulation principles.
References
Byrd, N. (2019). What we can (and can’t) infer about implicit bias from debiasing experiments. Synthese, 198(2), 1427–1455. https://doi.org/10.1007/s11229-019-02128-6
Woodward, J. (2016). Causation and Manipulability. In E. N. Zalta (Ed.), The Stanford Encyclopedia of Philosophy (Winter 2016). Metaphysics Research Lab, Stanford University. https://plato.stanford.edu/archives/win2016/entries/causation-mani/
Author Response
Please see the attachment
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
The new version of the paper is better than the last version I reviewed a while ago. There is a systematic typo across the paper, though: In many places, it seems we are gonna be added references but instead of references we see question marks (??). I guess the author will include the references later, although I wonder why has submitted the paper without doing so now? Is maybe a typo of the editor manager used by MDPI?
In any case, I only have two sources of concern, which I expressed also in my previous review.
1) Concerning Cartwright, I didn't say that the author of this paper disagrees with her. I said I do not see how what the author of this study presents is novel with respect to her work. And then, my suggestion, was to at least acknowledge in the paper that her 1999 work, and the work afterwards, does something extremely similar to this author's work.
2) I am still unclear about the generality of the claim, for eeven in the same science there are substantial differences depending on the research question. So everything being said about specific sciences "in general" is based on choosing the specific case studies correctly. Note that this limits, in my view, the appeal of the hypothesis.
Author Response
I realise that I need to spell out in somewhat more detail how my views relate to those of Cartwright, which I have done in lines 336-367 in the revised version of the paper. In contrast to Cartwright I hold that laws are true and needed for inferring causal relations. Moreover, Cartwright holds that capacities and causes play a central role in science and she writes as if these things are kinds of entitities. But a cause is always a cause of an effect, so it is the causal relation which is of interest. And knowledge of causal relations require knowledge about relations between quantities, i.e. about laws, although laws themselves don't state causal relations, which I have discussed in the section 'laws and causation'.
Regarding the second point, about the generality of my conclusion, I'm not really sure how to interpret it. In any case I have added subjunctive clause stating a condition. Now the conclusion reads:
'Since experiments often are impossible, the way to move forward is to look for natural experiments, in so far as one wants causal knowledge useful for future actions.'
