Experiments have been a recognised part of scientific thinking since Antiquity, as shown for example by Aristotle. In the Middle Ages, the Franciscan brother Roger Bacon (1214–1294) rose to particular fame for his scientific experiments. However, it was not until the so-called scientific revolution that experiments achieved recognition as a key element of research methodology. Galileo Galilei was perhaps the most influential researcher in this respect. His inclined plane experiment provides a prime example.
What is an experiment?
What is special about experiments, and why and how can they give us insight into the natural world? Most likely, the key is that experiments simulate reality by establishing a situation where we can control known parameters (the independent variables) to find out how another parameter (the dependent variable) changes in relation to the former.
In terms of the philosophy of science, there are certain properties of which we should be aware: First, experiments are not equivalent to observed natural phenomena/reality in any direct sense. They simulate a real situation in an artificial setting, such as falling bodies in the inclined plane experiment. Galileo's real interest was in studying how bodies behave in free fall, but since this is difficult to measure for practical reasons, he had the idea of – somehow – 'delaying' gravity by using an inclined plane that permitted measurements to be made. Second, it is important to be aware that in an experimental setting we only have control of the factors that are known in advance. In principle, there will always be some possibility that unknown confounding factors will distort the results we hope to achieve. Third, it is also important to be aware that the instruments used to take measurements may occasionally add considerable uncertainty. For example, measurement rods may change with temperature.
Experiments and thought experiments
Special difficulties may arise with so-called thought experiments. In principle, it is hard to understand how they may have any role in empirical research whatsoever. Nevertheless, many of the most significant experiments have been based on thought experiments. The first to perform such an experiment in the modern era was again Galilei. He asked himself how the Earth could move through space without us having any sense of this motion. In a thought experiment, he transferred this situation to a rider on horseback, throwing a ball. When the horse moves, the ball will behave as though the horse's movement has imperceptibly been added to the movement of the ball. When the rider throws the ball up in the air, he can catch it again as though he was standing still.
Thought experiments are rarely seen as confirmation of a controversial theory, but they provide a good basis for understanding complex theoretical associations. In this way they are also similar to our modern computer simulations, where we feed in relatively complex initial conditions and then 'observe' how the system will subsequently behave. Their major advantage is that here too we have control over the parameters and can read an outcome that could not have been immediately predicted without this simulation. Climate models and their simulation using powerful computers are a good example.
Experiments and research ethics
The question is whether considerations of research ethics should urge us to post some 'warning signs' related to experiments.
The answer is obviously yes. First, we need to be aware that some experiments may be questionable in terms of research ethics, depending on the subjects on whom these experiments are conducted. The Nazi doctors who performed medical experiments on prisoners in concentration camps acted immorally in the extreme, even when the experiments were scientifically valid (which was not always the case). The same can be said of certain forms of animal experiments. Even the so-called 'Milgram experiment' in the social sciences (to test the lengths to which people would go in order to satisfy those in authority) involved a violation of the integrity of the research subjects and has gradually become a standard example of ethically problematic experiments in psychology and social science.
Second, it is a problem in terms of research ethics when the outcome of experiments is immediately taken as confirmation of wide-ranging assertions about reality. For example in brain research, there are some experiments about people's moral decisions (http://www.teachersdomain.org/resource/vtl07.la.ws.research.yourbrain/). Here, the subjects will typically be presented with a dilemma in which they must choose one of the alternatives. Simultaneously, the researchers investigate what parts of the brain are being activated when reward or punishment strategies are chosen, and it turns out that different parts of the brain are responding. This is interesting, but at the same time we need to be cautious in drawing too wide-ranging conclusions about moral judgements as such. Rarely do we have a choice between only two alternatives. As a rule, we may be able to change the parameters as well, and this will occasionally be the morally correct thing to do as well as what will make people feel most comfortable. In such cases, the findings should be restricted to situations that are sufficiently similar, rather than suggesting that the experimental situation reflects a larger complex of problems (such as moral decisions in general). Thus, a general requirement of research ethics is that the overall insight generated by an experiment must be soberly restricted to the conditions under which the experiment may shed light, to avoid seducing us into drawing conclusions that are not guaranteed in view of the setup and nature of the experiment.
A particular challenge in terms of research ethics lies in our general proclivity to engage in wishful thinking. Specifically, this means that a researcher should be especially cautious in drawing conclusions from experiments if closer reflection causes him to recognise that psychological factors may have influenced the interpretation. The best known example in science history is the alleged discovery of N-rays by the French professor of physics René Blondlot in 1903. He claimed that N-rays existed (that were different from the known X-rays). These rays could not be observed directly, but according to Blondlot only in particular experimental settings where the N-rays would affect the intensity of various light sources. The obvious point is that such a change in light intensity would require an extremely accurate observer. Immediately after the publication of Blondlot's findings, many others also claimed that they could observe the same phenomenon. In the years 1903–1906, approximately 300 articles on N-rays were published. Many French physicists shared Blondlot's ideas about the existence of N-rays, whereas German physicists, for example, were highly sceptical. (The question is, of course, whether they would have reacted in the same way if Blondlot had been German).
What we observe largely depends on what we expect to see at the outset. As long as the findings of an experiment are primarily based on the researcher's observations without any independent and reliable measuring device, experiments may be influenced by wishful thinking. Many experiments are prone to this phenomenon, for example when cells/blood corpuscles etc. are counted under a microscope. To avoid wishful thinking on the part of the researcher affecting the result, it is always advisable to double-blind such experiments. (Here, 'double-blind' means that the researcher and the subject are both unaware of how the experiment is set up for the specific test runs.)
This last point also underscores a key issue in the philosophy of science pertaining to experiments. Their significance as a methodological tool for science is derived from their ability to be replicated by others. It is not until an experimental result has been reproduced by others that the result has undergone quality assurance. Blondlot's alleged findings were disproved in exactly this manner: when the American researcher Robert Wood accompanied Blondlot in the laboratory to observe these phenomena himself, he could ascertain that Blondlot saw what he wanted to see, and not what Wood or others saw. It is a problem in modern science that experiments are rarely replicated, because they are often costly and there is little prestige attached to repeating what someone else has already done. Based on good research ethics, however, it should be mandatory to always repeat important experiments to be able to exclude bias in the experimental setup or in the observer's findings. The well-known social psychologist and Nobel laureate Daniel Kahnemann recently issued a strongly worded call to his peers to pay more attention to replications in their discipline, precisely because of worries about decreasing quality in their findings.
This article has been translated from Norwegian by Erik Hansen, Akasie språktjenester AS.