Are we all chickens?

Inductive reasoning is often the method we scientists apply in our everyday work. We appeal to evidence — observations obtained, in the best case, under various conditions — for supporting our theories. This approach, however, bears unpleasant surprises.

Most of our knowledge, whether related to our everyday life or to our scientific activities, appears to accumulate by induction. Does the method of induction lead to relevant knowledge? Bertrand Russell in Problems of Philosophy warned us about the validity of this method. He illustrated the problem with the story of a chicken.

Once upon a time there was a chicken in a farm. This was a good and honest, empiricist chicken. He absorbed facts from around himself and learned about the nature of reality that he was exposed to. The chicken noticed an interesting thing, every day in the morning the farmer showed up and fed him. He did not really understand why that was happening but hey he got fed every day. The farmer showed up, the chicken got fed. Every single day, under all sorts of conditions, hot days, cold days, rainy days, windy days, the farmer showed up and the chicken got fed. What kind of conclusion could the chicken draw about the relationship between the farmer showing up and him getting fed? Well, being a good empiricist, allowing himself to make inferences about the laws of nature, the chicken concluded that there must be a law of nature saying: every time the farmer shows up a chicken gets fed. A different chicken, a sceptical one, asked: How do you know? How do you know that there is a law of nature that connects us being fed with the farmer showing up? The first chicken replied testily that all evidence pointed to this law of nature, that every single day of their life when the farmer showed up they were fed, that there was absolutely no evidence to contradict this.

The good, empiricist chicken

One morning the farmer showed up and wrung the neck of the chicken. What? The chicken relied on good inductive reasoning, all evidence supported his theory, none of the evidence contradicted it and nonetheless he ended up being wrong!

It appears that the experience we make about the past produces in us strong expectations about what should happen in the future. Food that has a certain appearance generally has a certain taste, and it is a shock to our expectations when food with a familiar appearance turns out to have an unusual taste. These expectations are very strong and not limited to humans. Horses that have been driven always along a certain road will follow that path even when they are by themselves. Dogs expect to go out when they see the leash. Despite being so misleading — David Hume put it nicely by saying that the future is under no obligation to mimic the past — all these expectations exist.

In science, the attitude to judge new observations according to expectations produces at least two kinds of problems, which pose some sort of dilemma. The first kind of problem is that too often new, unexpected results are rejected simply because they are not expected. I am sure we all have some personal experience with this but one of the most striking case is probably the one of Lynn Margulis. In the ‘60s, Lynn Margulis wrote a paper on symbiogenesis, in which she proposed endosymbiosis as the evolutionary force that led to the emergence of eukaryotic cells. The paper was rejected by about fifteen scientific journals possibly because her theory did not fit with the expectations of the time about evolution. Eventually, the paper was published in Journal of Theoretical Biology in 1967, the theory was experimentally demonstrated for the first time in 1978, and it is currently supported by genome data and molecular phylogeny. Her work is considered today a landmark in modern endosymbiotic theory. The second kind of problem is specular to the first. Too many studies, generally based on some observations, which are often very limited in time (the typical duration of a scientific project is, in fact, only three years), are considered to constitute relevant knowledge by induction, simply because they fit with our expectations, and they are published. Now the dilemma is: should we accept the expected or the unexpected? Should we accept both? And if so, in which proportion? Of course, the most immediate answer here is: it depends. But there is another aspect that adds on top of all this: the publish-or-perish culture. Obviously, we all (scientists, scientific institutions, and funding agencies) want to make the right judgments and be able to accept and value unexpected as well as expected results, when they constitute true knowledge. But the publish-or-perish culture leads us to shortcut decisions with respect to the scientific questions we ask, the approaches we use, and the results we accept and value. The pressure to increase the number of publications exacerbates the problem of induction and creates knowledge based on unfounded expectations more often than not. This has led to wasteful research and unethical practices with the risk that good research gets buried by the increasing volume of poor work.

Not long ago, Nature asked Thomson Reuters to list the 100 most highly cited papers of all time. Thomson Reuter’s Web of Science includes databases covering the social sciences, arts and humanities, conference proceedings and some books. It lists papers published since more than a century ago. If you printed out the first page of every item in this index, the stack of papers would almost reach to the top of Mount Kilimanjaro, Nature discovered. What I find particularly worrying about this study is the staggering amount of published work that has gone completely unnoticed: more than half has been cited only once or not at all. Thomson Reuter’s Web of Science holds some 58 million items. What kind of conclusions should we draw from these numbers? That millions of publications and hence millions of scientists and their respective institutions are completely useless?

And I am not so sure if the new breed of open access journals can cure these problems. Typically, the claim is that the work published by open access journals should only be technically sound, novelty and hence scientific value is judged by the readership. But who is this reader that can read and judge the staggering body of literature being produced every year, especially when all potential readers seem to be busy only with writing?

Among the most influential views on the problem of induction are those of Karl Popper, presented in The Logic of Scientific Discovery. Popper sustained that induction should have no place in science. Science, in his view, is a deductive process in which scientists should formulate hypotheses and theories that are tested by deriving particular observable consequences. Theories cannot be confirmed or verified, he argued. They may be falsified and rejected or tentatively accepted if corroborated in the absence of falsification by the proper kinds of tests.

And yet, we are all chickens and scientific institutions (we again), being unable to see scientists differently than chickens in battery cages, are our henhouses.

Ago