Uhl, Alfred (2019): Correlations are not causation - remarks from epidemiology. Epistemology of Big Data Health and Health Care. Bertalanffy Center for the Study of Systems Science (BCSSS), 4. April 2019, Wien.

Full text not available from this repository.


It has long been established based on theoretical considerations that causal relationships cannot be derived from merely observed associations. Bernhard Russell clearly demonstrated this with an example of two clocks striking every day one after the other. Deriving causality from association is commonly called “Cum Hoc Fallacy” in the case of simultaneity and “Post Hoc Fallacy” in the case of a sequence. Experiments are generally regarded as ideal way to test causal relationships, but commonly experiments cannot be conducted due to insurmountable research limits. The undeniable insight that causality cannot be inferred from mere observation is contradicted by the immediate experience that many causal relationships can be successfully deduced from pure observation in everyday life. Intuitive inference from observations to causality is a genetically pre-programmed Gestalt phenomenon usually arising instantly like a spontaneous feeling and without any conscious thinking. Humans are ingenious in intuitively recognizing patterns and causal relationships and often turn out far superior to current developments in AI, and there are bad in logic and statistics, but the patterns and causal relationships they detect do not always correspond with reality. This situation induces a dilemma. On the one hand it is impossible to do without intuition, particularly where experimental research is not feasible, and at the same time it is necessary to critically examine the results of intuition rationally and empirically, i.e. one must not blindly rely on one’s intuition. In my presentation some examples are provided which are interpreted causally by most observers, whereby is simple to show that the examples are not qualified as empirical evidence for these conclusions. Additionally some further relevant examples of everyday illusions are presented – inducing illusions which should definitely be considered. There is no simple escape out of this dilemma. Where decisions must be taken even if we are not able to rely on well-established causal relationships, where only hypotheses derived from intuition are existing. A sensible approach lies somewhere between associative intuition, which Gigerenzer (2007) described in "Gut Feelings", and the analytical-logical approach, described in "Helping Doctors and Patients Make Sense of Health Statistics" by Gigerenzer et al. (2008). Particularly in the latter context, many other authors are worth mentioning who have systematically dealt with mental errors and provided vivid examples of how unquestioned intuition can lead to utterly wrong conclusions. I am thinking of "The Invisible Gorilla" by Chabris & Simons (2011), "Thinking Fast and Slow" by Kahneman (2011) or "The Logic Of Failure" by Dörner (1996), just to give a few examples.

Item Type: Conference or Workshop Item (Lecture)
Subjects: OEBIG > Kompetenzzentrum Sucht
Date Deposited: 23 Mar 2020 20:22
Last Modified: 23 Mar 2020 20:22
URI: https://jasmin.goeg.at/id/eprint/1248