Monday, November 27, 2006

The Demarcation Problem

The following is an essay I submitted for one of my classes last year. It deals with the demarcation problem of science, so I figured you might find it interesting.

The field of science has been responsible for many significant steps forward in civilization, and true scientists are granted great amounts of respect in modern society. As such, many pretenders find it beneficial to claim to be scientists themselves. This makes the Demarcation Problem—distinguishing science from non-science, pseudoscience, and religion—an important issue and possibly a useful tool in exposing these pretenders. In this paper I intend to examine the history of demarcation, analyze the boundaries of science with non-science, pseudoscience, and religion, and propose standards for distinguishing science from these fields.

- - - - -

The first attempt to demarcate science came from the Vienna Circle, a philosophical society formed at the Vienna University in 1922. They developed the theory of Logical Positivism, which stated that the only meaningful statements were ones that were derived from empirical observations (also known as verificationism). This most clearly demarcates science from religion (which was their particular goal), in that it implies that religious and metaphysical statements are meaningless. There were problems with this theory, however. For one, it failed to distinguish fields such as art, which is also based on empirical knowledge (though generally artificial and subjective), from science. The theory also inadvertently condemned mathematics as being meaningless, as it wasn’t based on empirical evidence.

These problems were noticed by the philosopher Karl Popper. He pointed out that a theory could be meaningful without being scientific and that a standard of meaningfulness would not necessarily coincide with a demarcation of science. Popper’s proposed alternative was falsificationism: If a theory is falsifiable, it is scientific; if a theory is not falsifiable, it is not scientific. This theory has shown to be much better at demarcating science than verificationism was. Using the example above, art shows no falsifiability as the field is highly subjective, and each person is entitled to their own artistic preferences. Also, since mathematical theorems are proven, they are not falsifiable and thus not scientific—though this doesn’t mean that they aren’t meaningful or useful for science.

Falsificationism has some problems of its own, however. Almost every theory or paradigm has certain anomalies which could be said to falsify it. However, due to the statistical nature of many experiments, if a large number of experiments are done, it becomes inevitable that some will display seemingly anomalous results. Falsificationism also has the curious property that it makes many mundane statements scientific. For instance, the statement, “My eyes are blue,” could be falsified if the one looked at the speaker’s eyes and found that they were not blue, so under falsificationism, it must be a scientific statement.

The most recent views on the demarcation of science come from Thomas Kuhn. He stressed that science operates in two different manners: what he calls “normal science” and “extraordinary science.” Normal science is mostly a problem-solving phase, where scientists solve problems within the current paradigm. In this phase, the standard of falsificationism still has merit. Extraordinary science is what happens once a significant number of anomalies has built up, making the current paradigm no longer viable. A new paradigm is developed at this time, and the scientific community undergoes a “paradigm shift” and gradually switches. Since all paradigms have anomalies, particularly at their early stages before they’ve been fully refined, a strict interpretation of falsificationism would rule out all of these new paradigms almost immediately. This isn’t what the scientific community judges, however. The new paradigm is judged on its ability to solve problems. If it solves more problems or more important problems, it is generally accepted over the previous paradigm. Thus, in this stage, the standard of whether a paradigm is scientific is its ability to solve problems in normal science.

- - - - -

Distinguishing science from non-science or religion is generally a simple matter, as the two generally make no claims to be scientific. When they do make claims of being scientific, they become pseudoscience. Even so, it is useful to recognize the distinctions, as science doesn’t always specifically declare itself as such. For the boundary between science and religion, the verificationism view works quite well in the majority of cases. Science deals primarily with empirical matters, while religion deals primarily with spiritual matters, which cannot be empirically and objectively observed.

Occasionally, however, these will overlap. Religion might make claims about empirical matters, or science might make claims about unobservable phenomena. In these cases, the methodology of the two is what differentiates them. Science relies on objective experiments to gain knowledge, and there is a high degree of community in it that helps to confirm or disconfirm theories and settle debates between paradigms. Religion is more subjective, and every individual has a different interpretation. Due to its objective nature, two individuals practicing science within the same paradigm will generally come to the same result. In religious debates, disputes are attempted to be settled by appeal to authority—generally, the word of the clergy, holy texts, or a prophet—but this is never perfect due to the subjective element, and there will always be different interpretations of a religion.

When distinguishing science from non-science, it’s easiest to start by defining the core of what science is, and then define non-science as the fields which fall outside this definition. At it essence, science is a system of acquiring knowledge about the physical world through objective experimentation and observation. Therefore, we can immediately classify fields such as art and business, which do not involve the acquisition of knowledge, as being non-science. Mathematics, which doesn’t acquire knowledge about the physical world, is also non-science. Engineering is focused on the practical application of scientific knowledge, so while engineers may practice science at times, the field as a whole is non-science.

There is also the subdivision of science known as the social sciences, comprising fields such as history, economics, and sociology, to consider. They use methods similar to the natural sciences, but study human behavior instead. One large limitation they have is that experimentation is rarely feasible, and so all studies must be observational. This makes it a lot more difficult to test theories, and the unpredictability of human nature only adds to this problem. But just because a field comes with difficulties doesn’t mean it can’t be scientific—it just isn’t as reliable. The social sciences do fit many of the criteria of science, but it is necessary to keep in mind that due to the complexities of human nature, the results aren’t as reliable as those obtained in the natural sciences.

Now we come to the division between science and pseudoscience. Pseudoscience can be defined simply as something which claims to be scientific but isn’t. Classic examples of pseudoscience include astrology and the belief in ESP. The belief in the existence of ESP provides a good example of how falsificationism can work to classify it as pseudoscience. Since it’s impossible to analyze every event on earth taking place at any time in history, it would be impossible to prove that there isn’t at least one legitimate case of ESP. As such, it’s impossible to falsify it, so it’s unscientific.

Astrology, on the other hand, is a field for which falsificationism fails us. Since it relies on predicting the outcomes of earthly events from analyzing the stars and other planets of our solar system, it could theoretically be falsified if these predictions are false (or show no statistical improvement over random, similar predictions). In fact, this is exactly what has happened. Experiments have been done repeatedly with astrology showing that its predictions show no statistical merit. But the fact that some experiments have been done which falsify it isn’t enough to say it’s unscientific—all paradigms and theories have had a small number of experiments performed which give contrary results. It is simply the result of random chance that with a large number of experiments performed, some will have extraordinary results. Even if many or most of the experiments give evidence against the theory, it’s still possible that the theory is correct and this was just an extraordinary occurrence.

If we were to rely solely on falsificationism, we would be forced to accept that those who practice astrology are practicing a science. Therefore, it’s apparent that we will need a further standard to judge whether falsifiable fields are scientific. For this, we can use some of Kuhn’s views on science: primarily that it is, in the end, a communal practice. Therefore, we can rely on the opinion of the scientific community on whether or not a theory actually has been falsified. In the case of astrology, this is most definitely the case; the chance of all of the experiments which disprove it being the result of random chance is so slim that astrology can safely be said to be falsified, and thus is a pseudoscience.

There is an obvious problem in applying this standard too strictly: All new paradigms and theories go through a period in which they are put on trial by the scientific community, and it is the norm that they are presumed “guilty until proven innocent.” Because of this, we must make an important distinction. Newer paradigms and theories that have yet to be fully fleshed out and whose “trials” have yet to be resolved are classified as “protosciences.” These protosciences shouldn’t be classified as pseudosciences as they haven’t been dismissed by the scientific community. But conversely, they also shouldn’t be given the weight that established sciences have, as they have not yet proven their merit. An example of a modern protoscience is String Theory, which roughly states that subatomic particles are shaped like coiled strings, and the different motions of these strings correspond to different properties and interactions of the particles. At this time, the proponents of String Theory have some evidence which could be said to support this theory, but they don’t have a way to test it—it isn’t falsifiable. It will possibly become falsifiable at some point as the theory evolves, and it is not immune to scientific review, so calling it a pseudoscience would be a definite misnomer.

- - - - -

One of the reasons that demarcation is an important problem is in the admissibility of scientific testimony in court. “Junk science”—the popular term for pseudoscience—has been showing up in an alarming number of lawsuits in attempts to win money when there is no actual merit to the case. Up until 1993, the preferred method for accepting scientific testimony was for the judge to determine simply whether the evidence presented had “gained general acceptance in the particular field in which it belongs.” While this standard was reasonable, it was often impossible to apply it. How would the opposing attorney prove that the evidence was not generally accepted? A disagreeing expert could be brought in, but how would that expert prove to be more reliable? The difficulty here was what allowed so much junk science to sneak through.

In essence, the legal system needed to come up with its own demarcation between science and pseudoscience that was simple enough for the judge to understand and apply. This was done in the Supreme Court’s opinion of the case of Daubert v. Merrell Dow Pharmaceuticals in 1993. The “Daubert Standard” included two measures by which an expert’s testimony would be admissible: relevancy and reliability. The relevancy prong is simply whether the testimony is relevant to the case. For instance, a chemist could testify that acid could be used to burn off one’s fingerprints, but if fingerprints were actually found at the crime scene this piece of information could be deemed irrelevant. The reliability prong was the part used to determine whether the testimony was actually scientific. The Supreme Court gave “general observations” of what made scientific testimony reliable, though they stressed that it wasn’t supposed to be an exacting checklist:

  • Empirical testing: the theory or technique must be falsifiable, refutable, and testable.
  • Subjected to peer review and publication.
  • Known or potential error rate.
  • Whether there are standards controlling the technique's operations.
  • Whether the theory and technique is generally accepted by a relevant scientific community.
The first point here is simply the falsifiability criterion. The fifth point shows a critical divergence from asking for communal acceptance of the theory; it asks only for communal acceptance of the methods used to obtain this theory (this is also the intention of the fourth point). This is important as it allows for novel ideas and evidence to be presented, as long as they were arrived at in a scientific manner. The second and third points are most important within the legal context for which this standard was intended. The second point means that the testimony has been exposed to the possibility of criticism from the scientific community, so if the evidence presented has been argued heavily against, it would be easy for the opposing council to show this. The third point also opens up a possible argument if the error rate is high; the opposing council could argue that this high error rate decreases the reliability of the testimony as evidence.

- - - - -

This gives us the following three-step process in order to determine whether a theory is scientific: First, determine whether the theory is potentially falsifiable (a good measure of whether it’s empirically-based). If it’s falsifiable, determine whether the methods used to obtain it are generally accepted by the relevant scientific community. Finally, examine the criticism of the theory by the scientific community, and whether or not this warrants claiming that it’s falsified. If it’s falsifiable, obtained in a scientific manner, and hasn’t been falsified by the scientific community, then it’s scientific. This method isn’t perfect, and one should keep in mind the aforementioned classifications of social sciences and protosciences which may or may not end up classified as science under this standard. Of course, science is ever-evolving, so many sciences of today will likely become obsolete, just as Phlogistics and the Geocentric Theory have. In their day, they were scientific, but if one were to practice them today, they would be pseudosciences.


Anonymous said...

Nicely done. What kind of mark did you get on that?

Also, you mention "the legal system". Do you mean strictly the legal system of the USA, or have other countries adopted the 1993 case as precedent, as well?

Infophile said...

We rarely get our final exams back at this school, so I can only guess at the grade on that exam given my final score and how I did on the other papers. From that, the score was probably close to 85% (keep in mind that in the Canadian system, 80% corresponds to the American 90%, so that would be an A).

And yes, I was referring to the American legal system there. The professor was American, so I just left that assumed for parsimony.

It's exceedingly rare, as far as I know, for any country to accept a judgment of a foreign court, so this has never been referenced specifically in any foreign cases (that I could find). Additionally, pseudoscience in the courts is primarily a US phenomenon, as it's at the intersection of a high psuedoscience quotient and a sue-for-anything mentality. This leads for little need for a standard in other countries, so cases are generally decided on an ad hoc basis. Depending on the cultural standards, this can be good or bad, but generally, I believe having a standard such as this would be better in almost all cases.