When I first heard the expression “evidence-based medicine”, I thought: What, there’s another kind? I soon learned that a lot of medical practice is based on tradition rather than on clinical trials, and came to appreciate the movement that began in the 1990s toward improving medicine based on good evidence. Today, there are solid reasons to expect that clinical psychology, dentistry, education, and even philosophy should be evidence-based. But what is evidence?
Dictionary definitions of evidence are of no help, because they allow that any ground for belief could count as evidence. People make take their holy books, hallucinations, and gossip as grounds for their beliefs, but these are not good evidence. Similarly, some analytic philosophers think that they can accumulate evidence for hypotheses about knowledge, reality, and morality by consulting their own intuitions about stories they’ve made up. In thinking about the difference between thought experiments and real experiments, I’ve tried to compile a list of characteristics of good evidence based on practices that have become standard in science since the 1600s. Here are five important characteristics of evidence that can provide legitimate grounds for belief.
1. Reliability: A source of evidence is reliable if it tends to yield truths rather than falsehoods, as in systematic observations using instruments such as telescopes and microscopes and in controlled experiments such as those practiced by many scientists. In contrast, religious texts, hallucinations, gossip, and philosophical thought experiments often support falsehoods.
2. Intersubjectivity: Systematic observations and controlled experiments do not depend on what any one individual says, but are intersubjective in that different people can easily make the same observations and experiments.
3. Repeatability: A major source of the intersubjectivity of systematic observations and controlled experiments is that the same person or different persons can get similar results at different times, replicating the original experiments or observations.
4. Robustness: Experiments results should be obtainable in different ways such as using different kinds of instruments and methods. For example, different kinds of microscopes can be used to provide similar insights into cell structure.
5. Causal correlation with the world: Evidence based on systematic observation or controlled experiments is causally connected with the world about which it is supposed to tell us. For example, telescopes and microscopes provide evidence because reflected light enters the eyes of observers, stimulates their retinas, and generates perceptions in accord with neural processes that are causally regular.
Holy books, hallucinations, gossip, anecdotes, and philosophical thought experiments fall short on these requirements, whereas scientific experiments generally satisfy them unless they are incompetent or fraudulent.
Of course, there is no simple step from evidence to hypotheses, which need to be evaluated in ways that take into account alternative hypotheses that compete to explain the full range of available evidence. Medicine, psychology, and other fields are evidence-based when they evaluate competing hypotheses based on how well they explain evidence that is reliable, intersubjective, repeatable, robust, and causally correlated with the world.