Skip to main content
SpringerLink
Log in

Inference to the best explanation and mechanisms in medicine

  • Published:
Theoretical Medicine and Bioethics Aims and scope Submit manuscript

Abstract

This article considers the prospects of inference to the best explanation (IBE) as a method of confirming causal claims vis-à-vis the medical evidence of mechanisms. I show that IBE is actually descriptive of how scientists reason when choosing among hypotheses, that it is amenable to the balance/weight distinction, a pivotal pair of concepts in the philosophy of evidence, and that it can do justice to interesting features of the interplay between mechanistic and population level assessments.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Log in via an institution

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Similar content being viewed by others

Notes

  1. Also, a related problem is how evidence is expressed, most of the time in qualitative, rather than quantitative, terms; I thank one of the reviewers for drawing my attention to this. For a full overview of the problems faced by the Bayesian approach, all the more insightful since it is provided by a Bayesian theorist, see Earman [2].

  2. Semmelweis’s discovery of the causes of puerperal fever has been a favorite example for both Lipton [4] and Bird [5]. Bird has also provided an analysis of Bradford-Hill’s criteria in terms of his own brand of IBE, in [6].

  3. In Lipton’s words, ‘The sensible modesty consists in making no claim that Inference to the Best Explanation is the foundation of every aspect of non-demonstrative inference…. It is glory enough to show that explanatory considerations are an important guide to inference. Consequently, there is no need to argue heroically for a perfect match between the explanatory and the inferential virtues. Similarly, in the third stage there is no need to argue that explanatory considerations are our only guide to inference, just that they are a significant guide, an important heuristic’ [4, p. 121].

  4. More precisely, the research done in laboratories that issue in physiopathological knowledge—immunological, biochemical, biophysical, etc.

  5. Otherwise, one falls prey to a misguided causal dualism, or causal pluralism, about causation; see [10]. The interested reader is warmly encouraged to peruse that paper as well, since it offers the metaphysical counterpart for the ontology of mechanisms that could not be enclosed, for lack of space, in the present article, which is mainly focused on IBE and the epistemological side of the discussion.

  6. As Psillos stresses, ‘…which way the reasoning should go depends on explanatory considerations. Insofar as the conclusion “All As are B” is accepted, it is accepted on the basis it offers a better explanation of the observed frequencies of As which are B in the sample, in contrast to the (alternative potential) explanation that someone (or something) has biased the sample’ [12, p. 620].

  7. Indeed, it might well be that how one judges various sources of evidence as to their reliability, involves, at least in part, inferences to the best explanation, where such inferences point to the truth of the evidence being furnished or provided; see Harman [11, pp. 93–94].

  8. For a presentation of Galenic and Hippocratic views on the pulmonary pathology (and physiology), as well as the background humoral theory, see Debru [13], Nutton [14], and Jouanna [15].

  9. Since I will provide examples that make use of Mill’s methods, it will be useful to quote at least three of them here (out of a total of five), as they were laid out by Mill in his System of Logic—the method of agreement, of difference, and of concomitant variations: ‘FIRST CANON. If two or more instances of the phenomenon under investigation have only one circumstance in common, the circumstance in which alone all the instances agree is the cause (or effect) of the given phenomenon. This is sometimes known as the method of agreement. SECOND CANON. If an instance in which the phenomenon under investigation occurs, and an instance in which it does not occur, have every circumstance in common save one, and that one occurring only in the former, the circumstance in which alone the two instances differ is the effect or the cause, or a necessary part of the cause, of the phenomenon. This is sometimes known as the method of difference…. To these four canons may be added a fifth, the method of concomitant variations. Whatever phenomenon varies in any manner whenever another phenomenon varies in some particular manner is either a cause or an effect of that phenomenon, or is connected with it through some fact of causation. The difficulty of discovering causation is greatly increased by the fact that in many cases there are plurality of causes and intermixture of effects’ [17, p. 455] (italics added).

  10. See, for instance, Lipton [18, pp. 39–40, 42–43]; see also Bird [5, 6, 19] and Psillos [12, 16, 20]. Aside from Mill’s methods, Psillos and Bird also discuss the nomological side of the explanation in question, and Bird has emphasized the eliminative aspect of IBE.

  11. A comprehensive overview of Mill’s methods and their significance can be found in Cartwright [21].

  12. See Bird [22, 23], especially his statements on the difference between the methodology of discovering causes and the metaphysics behind causation. There is no room here to develop the subject, but note, as to some prima facie worries, that many counter-examples to the counterfactual methodology of discovering causes (example cases of pre-emption) could be dealt with by eliminating the non-circularity condition and by other means (see Broadbent [24] on backtracking counterfactuals and also the extraordinary recent work of Michael Strevens [25, 26] on explanation and the kairetic criterion). Note also that the cases of pre-emption involving mutually disjunctive causes advanced by proponents of causal processes should not pose problems since such causal processes count themselves as difference making.

  13. Lipton [4, p. 115]; see also the exchange of articles and replies between Lipton and Salmon, in particular Salmon [27] and Lipton [28].

  14. I defend the stronger view of mechanisms as difference making in [10].

  15. Charles Pierce, who was the first to describe IBE (under the name of ‘abduction’) in the 19th century, characterized it as ‘reasoning from effects to cause’; cf. Niiniluoto [30, p. S436]. If there is a science that is most interested in this, surely that would be medicine.

  16. However, as mentioned in the beginning of this section, taking into account the reliability of evidence when inferring causes or best explanations, on the one hand, and providing reasons why a certain source of evidence is reliable or not, on the other hand, are distinct issues.

  17. On a Bayesian view, one of the formulas proposed for the weight of the evidence is the logarithm of the likelihood ratio: W(H:E) = log[P(E/H)/P(E/~H), where H would be the hypothesis, say, that the accused committed the crime, and ~H the hypothesis that it did not commit the crime. An alternative formula would be [P(E/H) + P(E/~H)]/[P(E/H) − P(E/~H)] which represents sinh {W(H:E)/2} [41, pp. 251–253]. The likelihood ratio above is a form of the Bayes factor, which is used in selecting among competing models. However, the issue of weight is still very much in dispute, with many other measures of weight being proposed; for an overview, see for instance, Glass [42]. That is to say, Bayesians still have a problem with the weight of evidence, or more specifically, the numerical expression of it. My purpose in this section is to show that IBE does not have a conceptual problem with the balance/weight distinction—in other words, it can coherently deal with and incorporate these aspects of the evidence. Since, as mentioned in the Introduction above, I do not take IBE to be a rival of Bayesianism but rather its companion in those situations where the numerical data is not sufficient or the evidence is rather qualitative than quantitative, I will not go into the proposals that IBE itself can offer numerical expressions of the weight of evidence, advanced by Glass himself in [42, 43], and also more recently by Douven and Wenmackers [44] and Douven and Schupbach [45, 46], interesting as these proposals may certainly be.

  18. Psillos [16, pp. 441–447]; Psillos identifies one facet of the holy grail temptation with ‘the new Dogma, Bayesianism’ and rejects the compatibility between the latter and IBE. Psillos is, however, too harsh in this matter.

  19. For a general presentation, see http://ebmplus.org/ [48].

  20. That such population studies alone are sufficient for confirming causal claims has been contested by proponents of the Russo-Williamson thesis; see [8, 9]. Even more boldly, that such population studies alone are necessary for confirming causal claims has been contested, most vocally by John Worall; see, for instance, Worall [49].

  21. In the first section above, I noted that ‘individuation’ or ‘precision’, as Psillos calls it, is one of the main explanatory virtues employed in IBE.

  22. I presented Mill’s methods by directly quoting Mill in the first section, footnote 9. Recall that the method of continuous variations says that if one causal factor is found to vary whenever another causal factor varies in some way, then the two are related as cause and effect. The method is at work in most interventionist accounts of causation (e.g., Woodward’s) and can be identified in interventionist assessments such as RCTs.

  23. ‘Even where a mechanism linking A to B is well established and known in some detail, it can be hard to infer whether A has a positive effect on B, or A prevents B, or indeed whether A has any net effect on B at all. This is particularly true in cases where the mechanism is complicated: where there are several links on a pathway from A to B or where there are several pathways from A to B. It is also a problem where a mechanism is known to be non-robust over time or over other changes in situation. It is typically evidence of correlation that is crucial for determining whether any causation is positive or negative and what the net effect is. Thus evidence of mechanisms should be used in conjunction with evidence of correlation, not on its own, to infer causal claims.… The human body is a complex system, and the more we discover about it the more it seems that it is very common to have multiple mechanisms operating. If there are multiple mechanisms operating, they may impact on each other, and one or more may mask the effects of the mechanism you have discovered’ [1, pp. 349–351].

  24. For instance, consider the lack of patient compliance in the Finnish Mental Hospitals Study, 1968, which reached levels of significance for women in just one of the hospitals. Similar lack of compliance was reported for the Minnesota Coronary Survey, 1968 [31, pp. 36–37].

  25. Even in 1992, Uffe Ravnskov was arguing in the British Medical Journal that ‘apart from trials discontinued because of alleged side effects of treatment, unsupportive trials were not cited after 1970, although their number almost equaled the number considered supportive…. [A]uthors of papers on preventing coronary heart disease by lowering blood cholesterol values tend to cite only trials with positive results’ [50, pp. 15–19].

  26. See Worall [49, 51] for discussion.

  27. For a list of practitioners supporting this view see http://www.thincs.org/members.php [52]; for a list of publications arguing in favour of this view, see http://www.ravnskov.nu/references/ [53].

  28. See Ravnskov [54], Sutter [55], and DuBroff and de Lorgeril [56].

  29. For a recent example, see Chowdhury et al. [57], analysed in DuBroff and de Lorgeril [56]. I should underline that my purpose here is not to argue that theorists like Ravnskov are right in denouncing the so-called ‘cholesterol myths’, but to show how it is that such theorists were able in the first place to formulate their hypotheses, by starting from population studies with inconclusive results. My suggestion, as I will detail below, is that such inconclusive population studies arise because the problems of masking and extrapolation show not only at the mechanistic, microstructural level, as Clarke et al. claim, but can also manifest at the level of population correlations. Accordingly, we should strive to make our mechanistic knowledge more precise, in order to target and refine subsequently the population studies.

References

  1. Clarke, B., D. Gillies, P. Illari, F. Russo, and J. Williamson. 2014. Mechanisms and the evidence hierarchy. Topoi 33(2): 339–360.

    Article  Google Scholar 

  2. Earman, J. 1992. Bayes or bust? A critical examination of Bayesian confirmation theory. Cambridge: MIT Press.

    Google Scholar 

  3. Clarke, B., B. Leuridan, and J. Williamson. 2014. Modelling mechanisms with causal cycles. Synthese 191(8): 1651–1681.

    Article  Google Scholar 

  4. Lipton, P. 2004. Inference to the best explanation, 2nd ed. London: Routledge.

    Google Scholar 

  5. Bird, A. 2010. Eliminative abduction—Examples from medicine. Studies in History and Philosophy of Science 4: 345–352.

    Article  Google Scholar 

  6. Bird, A. 2011. The epistemological function of Hill’s criteria. Preventive Medicine 53: 85–96.

    Article  Google Scholar 

  7. Feduzi, A. 2010. On Keynes’s conception of the weight of evidence. Journal of Economic Behavior & Organization 76(2): 338–351.

    Article  Google Scholar 

  8. Russo, F., and J. Williamson. 2007. Interpreting causality in the health sciences. International Studies in the Philosophy of Science 21(2): 157–170.

    Article  Google Scholar 

  9. Russo, F., and J. Williamson. 2011. Epistemic causality and evidence-based medicine. History and Philosophy of the Life Sciences 33(4): 563–582.

    Google Scholar 

  10. Dragulinescu, S. 2016. Mechanisms and difference making. Acta Analytica. doi:10.1007/s12136-016-0292-1.

    Google Scholar 

  11. Harman, G. 1965. The inference to the best explanation. The Philosophical Review 74: 88–95.

    Article  Google Scholar 

  12. Psillos, S. 2002. Simply the best: A case for abduction. In Computational logic: From logic programming into the future, ed. A.C. Kakas and F. Sadri, 605–625. Dordrecht: Springer.

    Chapter  Google Scholar 

  13. Debru, A. 1996. Le corps respirant: La pensée physiologique chez Galien. Leiden: Brill.

    Google Scholar 

  14. Nutton, V. 2013. Ancient medicine. London: Routledge.

    Google Scholar 

  15. Jouanna, J. 1992. Hippocrate. Paris: Fayard.

    Google Scholar 

  16. Psillos, S. 2007. The fine structure of inference to the best explanation. Philosophy and Phenomenological Research 74: 441–448.

    Article  Google Scholar 

  17. Mill, J. S. 2002 [1843]. A system of logic. Honolulu: University Press of the Pacific.

  18. Lipton, P. 1993. Making a difference. Philosophica 51: 39–54.

    Google Scholar 

  19. Bird, A. 2007. Inference to the only explanation. Philosophy and Phenomenological Research 74: 424–432.

    Article  Google Scholar 

  20. Psillos, S. 2000. Abduction: Between conceptual richness and computational complexity. In Abduction and induction: Essays in their relation and integration, ed. A.C. Kakas and P. Flach, 59–74. Dordrecht: Kluwer.

    Chapter  Google Scholar 

  21. Cartwright, N. 1989. Nature’s capacities and their measurement. Oxford: Oxford University Press.

    Google Scholar 

  22. Bird, A. 2005. Laws and essences. Ratio 18(4): 437–461.

    Article  Google Scholar 

  23. Bird, A. 2007. Nature’s metaphysics: Laws and properties. Oxford: Oxford University Press.

    Book  Google Scholar 

  24. Broadbent, A. 2007. Reversing the counterfactual analysis of causation. International Journal of Philosophical Studies 15(2): 169–189.

    Article  Google Scholar 

  25. Strevens, M. 2011. Depth: An account of scientific explanation. Cambridge: Harvard University Press.

    Google Scholar 

  26. Strevens, M. 2012. Replies to Weatherson, Hall, and Lange. Philosophy and Phenomenological Research 84: 447–505.

    Article  Google Scholar 

  27. Salmon, W. 2001. Reflections of a bashful Bayesian: A reply to Peter Lipton. In Explanation, theoretical approaches and applications, ed. G. Hon and S. Rakover, 121–136. Dordrecht: Springer.

    Google Scholar 

  28. Lipton, P. 2001. Is explanation a guide to inference? A reply to Wesley C. Salmon. In Explanation, theoretical approaches and applications, ed. G. Hon and S. Rakover, 93–120. Dordrecht: Springer.

    Google Scholar 

  29. Illari, P., and J. Williamson. 2012. What is a mechanism: Thinking about mechanisms across the sciences. European Journal for Philosophy of Science 2: 119–135.

    Article  Google Scholar 

  30. Niiniluoto, I. 1999. Defending abduction. Philosophy of Science 66: S436–S451.

    Article  Google Scholar 

  31. Steinberg, D. 2007. The cholesterol wars. New York: Academic Press.

    Google Scholar 

  32. Kritchevsky, D. 1995. Dietary protein, cholesterol and atherosclerosis: A review of the early history. Journal of Nutrition 125: 589S–593S.

    Google Scholar 

  33. Steinberg, D. 2005. An interpretive history of the cholesterol controversy, part III: Mechanistically defining the role of hyperlipidemia. Journal of Lipid Research 46: 2037–2051.

    Article  Google Scholar 

  34. Brown, S., and J.L. Goldstein. 2009. History of discovery: The LDL receptor. Arteriosclerosis, Thrombosis, and Vascular Biology 29(4): 431–438.

    Article  Google Scholar 

  35. Kelly, T. 2008. Evidence: Fundamental concepts and the phenomenal conception. Philosophy Compass 3(5): 933–955.

    Article  Google Scholar 

  36. Kelly, T. 2014. Evidence. In The stanford encyclopedia of philosophy, ed. Edward N. Zalta. http://plato.stanford.edu/entries/evidence/. Accessed March 10, 2016.

  37. Keynes, M. 1921. A treatise on probability. London: Macmillan & Co.

    Google Scholar 

  38. Joyce, J. 2005. How probabilities reflect evidence. Philosophical Perspectives 19(1): 153–178.

    Article  Google Scholar 

  39. Cohen, J. 1986. Twelve questions about Keynes’s concept of weight. British Journal for the Philosophy of Science 37(3): 263–278.

    Article  Google Scholar 

  40. O’Donnell, R. 1992. Keynes’s weight of argument and Popper’s paradox of ideal evidence. Philosophy of Science 59(1): 44–52.

    Article  Google Scholar 

  41. Good, I. 1985. Weight of evidence: A brief survey. http://www.swrcb.ca.gov/water_issues/programs/tmdl/docs/303d_policydocs/207.pdf. Accessed March 10, 2016.

  42. Glass, D. 2012. Inference to the best explanation: Does it track truth? Synthese 185: 411–427.

    Article  Google Scholar 

  43. Glass, D. 2007. Coherence measures and inference to the best explanation. Synthese 157: 275–296.

    Article  Google Scholar 

  44. Douven, I., and S. Wenmackers. 2015. Inference to the best explanation versus Bayes’s rule in a social setting. British Journal for the Philosophy of Science. doi:10.1093/bjps/axv025.

    Google Scholar 

  45. Douven, I., and J. Schupbach. 2015. Probabilistic alternatives to Bayesianism: The case of explanationism. Frontiers in Psychology 6: 459.

    Article  Google Scholar 

  46. Douven, I., and J. Schupbach. 2015. The role of explanatory considerations in updating. Cognition 142: 299–311.

    Article  Google Scholar 

  47. Furie, M., and R. Mitchell. 2012. Plaque attack: One hundred years of atherosclerosis. American Journal of Pathology 180(6): 2184–2187.

    Article  Google Scholar 

  48. University of Kent. n.d. EBM Plus. http://ebmplus.org/. Accessed May 5, 2016.

  49. Worall, J. 2007. Evidence in medicine and evidence-based medicine. Philosophy Compass 2: 981–1022.

    Article  Google Scholar 

  50. Ravnskov, U. 1992. Cholesterol lowering trials in coronary heart disease: Frequency of citation and outcome. BMJ 305(6844): 15–19.

    Article  Google Scholar 

  51. Worall, J. 2010. Evidence: Philosophy of science meets medicine. Journal of Evaluation in Clinical Practice 16(2): 356–362.

    Article  Google Scholar 

  52. The International Network of Cholesterol Skeptics (THINCS). 2014. Members. http://www.thincs.org/members.php. Accessed May 5, 2016.

  53. Uffe Ravnskov. 2006. The cholesterol myths: References. http://www.ravnskov.nu/references/. Accessed May 5, 2016.

  54. Ravnskov, U. 2013. High cholesterol may protect against infections and atherosclerosis. QJM 96(12): 927–934.

    Article  Google Scholar 

  55. Sutter, M. 1994. Blood cholesterol is not causally related to atherosclerosis. Cardiovascular Research 28: 575.

    Article  Google Scholar 

  56. DuBroff, R., and M. de Lorgeril. 2015. Cholesterol confusion and statin controversy. World Journal of Cardiology 26(7): 404–409.

    Article  Google Scholar 

  57. Chowdhury, R., S. Warnakula, S. Kunutsor, F. Crowe, H. Ward, and L. Johnson. 2014. Association of dietary, circulating, and supplement fatty acids with coronary risk: A systematic review and meta-analysis. Annals of Internal Medicine 160: 398–406.

    Article  Google Scholar 

  58. Steinberg, D. 2005. An interpretive history of the cholesterol controversy, part II: The early evidence linking hypercholesterolemia to coronary disease in humans. Journal of Lipid Research 46: 179–190.

    Article  Google Scholar 

  59. Witztum, J., and D. Steinberg. 2010. History of discovery oxidized low-density lipoprotein and atherosclerosis. Arteriosclerosis, Thrombosis, and Vascular Biology 30: 2311–2316.

    Article  Google Scholar 

Download references

Acknowledgments

This article was written as part of the project ‘Grading Evidence of Mechanisms in Physics and Biology’, funded by the Leverhulme Trust (http://blogs.kent.ac.uk/jonw/projects/grading-evidence-of-mechanisms-in-physics-and-biology/). I am grateful for useful comments from Kristoffer Ahlstrom-Vij, Veli-Pekka Parkkinen, Jon Williamson, and two anonymous referees for this journal. The main and managing editors of this journal have been extraordinarily helpful with editing and content-wise suggestions for clarification. Finally, I would like to thank Mike Kelly for encouraging me to continue work when this article was at an early stage.

Author information

Authors and Affiliations

  1. Philosophy Department, School of European Culture and Languages, University of Kent, Canterbury, UK

    Stefan Dragulinescu

Authors
  1. Stefan Dragulinescu
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Stefan Dragulinescu.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Dragulinescu, S. Inference to the best explanation and mechanisms in medicine. Theor Med Bioeth 37, 211–232 (2016). https://doi.org/10.1007/s11017-016-9365-9

Download citation

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s11017-016-9365-9

Keywords

Search

Navigation