Last November 21st, a paper was published in Nature(available for free), co-authored from three scientists, two of them from Cambridge, UK, and one from Melbourne, Australia, suggesting twenty tips for interpreting scientific claims. Their goal was tohelp improving policy makers’ understanding vis-à-vis scientific results. Politicians are used to take advise from experts and consultants, and they constantly have to assess quality, limitations and biases of studies on which are based their arguments and policies.
Authors of this work suggest their 20 lessons should be part of any training programs dedicated to civil servants, politicians, policy advisers and journalists who have to interact with scientists. When I was Dean of the French School of Public Health (EHESP, Rennes), a school who is in charge, by monopoly, of training all civil servants in charge of the public health care system in France, we introduced a mandatory core curriculum where such kinds of basics were taught. However, I think, when I stepped down, last year, that we were still far from having achieved fascinating goals presented in this paper (I invite interested readers to directly read the full article in Nature).
Here, I will only list the 20 tips as selected by William J. Sutherland, David Spiegelhalter and Mark Burgman, since I believe such a list may inspire all those who are intending to set up a training program dedicated to high profiles in charge of public affairs in their own country:
1. Differences and chance cause variation
2. No measurement is exact
3. Bias is rife
4. Bigger is usually better for sample size
5. Correlation does not imply causation
6. Regression to the mean can mislead
7. Extrapolating beyond the data is risky
8. Beware the base-rate fallacy
9. Controls are important
10. Randomization avoids bias
11. Seek replication, not pseudoreplication
12. Scientists are human
13. Significance is significant
14. Separate no effect from non-significance
15. Effect size matters
16. Study relevance limits generalizations
17. Feelings influence risk perception
18. Dependencies change the risks
19. Data can be dredged or cherry picked
20. Extreme measurements may mislead
We have here a comprehensive agenda for our next summer schools, haven’t we?