My philosophy is this: Correlational nutritional studies should be trusted because they are the only and therefore best source of verifiable information on maintaining healthiness. They are, however, very noisy and each study in isolation should therefore influence my choices very little. Meta-analyses should be given more weight because they tend to smooth the noisiness.
Scott Adams bemoans the fact that the wildly varying conclusions of diet science cause people's pattern recognition to conclude that this is not trustworthy information.
However, the problem is not pattern recognition. The problem is the weight that people give to signals. When each signal is given a weight of 1 or 0, the pattern recognition will never converge on the underlying trend. The media and the public must understand that some studies should be given more weight than others.
Thus, scientists have a duty to communicate which studies should be most influential. They CANT trust the media to this.
Scott Adams bemoans the fact that the wildly varying conclusions of diet science cause people's pattern recognition to conclude that this is not trustworthy information.
However, the problem is not pattern recognition. The problem is the weight that people give to signals. When each signal is given a weight of 1 or 0, the pattern recognition will never converge on the underlying trend. The media and the public must understand that some studies should be given more weight than others.
Thus, scientists have a duty to communicate which studies should be most influential. They CANT trust the media to this.