A question long argued in the philosophy of science is the demarcation problem. How to we distinguish science from non-science? Karl Popper famously proposed falsifiability as a criteria. To be science, a theory must make predictions that could turn out to be wrong. It must be falsifiable. Theories that are amorphous or flexible enough to never encounter this test, aren’t scientific. This standard was famously sharp enough to cull Marxism and Freudean psychoanalysis from science.
Falsifiability has a lot going for it, but it also has a lot of issues. For one, when we say “falsifiable”, do we mean falsifiable in practice today? If so, then a lot of exploratory work done by scientists is non-science. This would include Copernicus’ work on heliocentrism, Albert Einstein’s work on relativity, or Peter Higgs and colleagues’ work on the Higgs mechanism. None of these theories were testable while these scientists were working on them. In the case of Copernicus and Higgs, it was several decades before they became testable, that is, falsifiable.
Reportedly, Popper was actually more careful than this. His proposed standard was falsifiable in principle. So to be scientific, a theory must be testable in some foreseeable manner.
But even this can be problematic. August Comte infamously predicted in 1835 that we would never know the composition of the stars. Speculation seemed pointless. But within a few decades, stellar spectroscopy was developed and we did actually start to learn about stellar composition. Likewise, when Einstein, Podolsky, and Rosen published their paper in 1935 on the EPR paradox, it was criticized by many as metaphysical navel-gazing, until John Stewart Bell figured out a way to test it 29 years later.
On top of this, as Sabine Hossenfelder recently pointed out, new theories can be about existing data. If the new theory explains the existing data better than an established theory, that is with fewer assumptions and perhaps simpler constructs, then it may replace that older theory, without ever producing its own unique falsifiable predictions.
Even more problematic, many successful scientific theories, while having reliably testable predictions, also have predictions that can’t currently be tested. For several decades, general relativity predicted gravitational waves, but they were only actually detected a few years ago. From what I understand, aspects of the role of pressure in general relativity remain untested.
And every scientific theory is essentially a metaphysical statement, a conclusion, or series of conclusions reached inductively. Anyone who has studied epistemology is familiar with the problem of induction, most famously analogized by black swans.
All of which means that the dividing line isn’t sharp, but long and blurry and requires a lot of judgment. I tend to think, rather than a sharp demarcation, it’s better to think in terms of a spectrum.
- Reliable models
- Rigorous exploration
- Loose speculation
- Falsified notions
1-Reliable models, are the ones most clearly science, and represents the most successful theories, such as general relativity, quantum mechanics, natural selection, etc. Often the predictions of these theories are reliable enough for technology to be built using them.
I considered calling 1 “settled science”, but that implies that successful theories are never overturned. Most famously, Newton’s laws of gravity reigned for centuries, before Einstein overturned them with general relativity. However, Newton’s laws remain reliable enough that NASA mission planners use them for most of their calculations. Newton’s laws are no longer the most reliable model, but they remain very reliable for many purposes. Which is to say, very successful theories, at least the mathematical components, are unlikely to ever be completely dismissed.
2-Rigorous exploration, is disciplined theoretical speculation. As noted above, scientists have to have space to work in this realm, since too many theories in 1 began here. But what characterizes rigorous exploration from the next category is that these theories are either extrapolations from theories in 1, or tight speculation involving one or a few assumptions, assumptions narrowly motivated to fit the data.
3-Loose speculation is where I think there start to be legitimate concerns about whether what’s happening is scientific. In this category, there may be numerous assumptions, with each assumption an opportunity to be wrong. Or the assumptions may be motivated by a desire for a certain outcome, not to explain the data, but perhaps to meet personal biases and intuitions.
I gave examples of 2 above. For 3, based on what I’ve read, string theory arguably belongs in this category. I think some other speculative notions, such as many exotic theories of consciousness, belong here too.
Many people would relegate all multiverse theories here, but I think they have to be looked at on a case by case basis, since some are either extrapolations of successful theories, or have minimal assumptions, and a strong case can be made for them being in 2. (None are currently in 1.)
But I would include Tegmark’s mathematical universe hypothesis in 3, along with a lot of other philosophical metaphysical speculation. This is often stuff that, strictly speaking isn’t impossible, but has non-data motivated assumptions and is the hardest to imagine ever being testable.
4-Falsified notions, the last category, is, simply put, fantasy. Generally for this stuff to be reality would require that one or more theories in 1 be wrong. Astrology, paranormal claims, creationism, intelligent design, and a lot of similar notions go here. If it’s presented as science then it’s fake science, pseudoscience.
Only 1 represents a reliable view of reality. But as noted above, this is science, and nothing is immune from possibly being overturned on new data.
2 represents what I often refer to as candidates for reality. Many will be right, others wrong, but we can’t currently know which is which.
3 might, in principle, turn out to be reality, but the probability is very low, low enough that the skeptic in me tends to just assume they’re wrong.
And 4 is the province of honest entertainers or dishonest charlatans.
It’s worth noting that even putting theories into these categories takes judgment, and many might sit on the boundaries.
But I think the main takeaway is that just because something isn’t in 1, doesn’t mean the only other option is 4. It’s not just reliable science or fantasy. There’s a space for exploratory science at least. I’m actually pretty sure science as an overall enterprise wouldn’t work without that exploratory space.
Unless of course I’m missing something? Am I being too permissive with these categories? Not permissive enough? Or just missing the ball entirely?