Evidence and the Assessment of Causal Relations in the Health Sciences

Abstract

This contribution claims that the two fundamental notions of causation at work in the health sciences are manipulative and mechanistic, and investigates what kinds of evidence matter for the assessment of causal relations. This article is a development of our 2007 article, ‘Plurality of Causality’, where we argue for a pluralistic account of causation with an eye to econometrics and a single medical example. The present contribution has a wider focus, and considers the notion of evidence within a whole range of disciplines belonging to the health sciences. Section 1 addresses the relations between kinds of evidence and causal accounts, and it is shown how different notions of causation can be employed in various medical cases. Section 2 calls attention to issues crucial for any adequate epistemological theory of causation, such as the distinctions between types and tokens, observational and experimental regimes, explanation and prediction. Lastly, the notion of context is articulated, highlighted in its role in the assessment of causal links. All these issues are tackled in the framework of what we label a ‘bottom–up’ epistemology.

Acknowledgements

The authors thank two anonymous referees of this journal, whose suggestions have been very helpful in preparing the final version of the article.

Notes

Section 1 is by Raffaella Campaner; Section 2 is by Maria Carla Galavotti.

The preventive measure was discovered respectively by Lind in 1753; Pott in 1775; Abbe in 1915. The causal agent was identified in 1928 (ascorbic acid deficiency); 1933 (benzopyrene); 1974 (N-nitrosonornicotine).

A distinction between Hill's guidelines that primarily relate to a ‘probabilistic regularity’ view, and those that relate to a ‘generative’ view of causation is also suggested in Thygesen, Andersen, and Andersen (2005).

‘Evidence for a mechanism of action’, Howick, Glasziou, and Aronson (2009, 186).

On this point we partly agree and partly disagree with Gillies (2011), who stresses the role of what he calls ‘interventional evidence’ in establishing a causal link, but admits of only actual, and not counterfactual, manipulations.

The notion of ‘mechanism sketch’ is used by Tabery in Machamer, Darden, and Craver's sense, and will be recalled in Section 2.

See Campbell (2008, 206–209). Campbell actually claims that interventionism contrasts with mechanical approaches. We do not believe such a contrast holds.

The most detailed and perspicuous version of Salmon's theory of explanation is contained in Salmon (1984). See also the collection of articles in Salmon (1998).

This move was prompted by criticisms by Hitchcock (1995).

For a more detailed outline of Salmon's theory, see Galavotti (1999).

More on the context dependence of causality to be found in Galavotti (2008, 2010).