I’m always interested in new takes on the demarcation between science and non-science, so after seeing the New Yorker write up on Michael Strevens’ new book, The Knowledge Machine: How Irrationality Created Modern Science, it seemed like something I needed to read.
Strevens begins by examining the two leading theories of science: Karl Popper’s falsifiability criterion, and Thomas Kuhn’s paradigm shifts. Falsifiability is the idea that science is primarily in the business of falsifying theories, and that for a theory to be scientific, it must be falsifiable, that is, it must make statements that could, at least in principle, be tested.
Paradigms refer to the prevailing assumptions during a particular period about how science works and what attributes a theory has to have to be considered science. These paradigms hold for long periods, until they begin to show cracks as scientist push their limits, eventually leading to their failure and a new scientific revolution and paradigm shift.
The problem with both views, Stevens argues, is that they don’t really account for what scientists actually do. His frequent example is Arthur Eddington’s 1919 experiment measuring the position of certain stars in the sky near the sun during a solar eclipse, an experiment which has historically been reported as a major victory for Einstein’s general relativity, demonstrating its predictions as more accurate than Newton’s.
But the reality is that Eddington’s results were a muddle that could have plausibly been interpreted as favoring either theory. For various reasons, Eddington had a strong preference for Einstein’s theory, so he spun the results, downplaying the results from one telescope that seemed to confirm Newtonian physics, and making allowances for another that seemed more friendly toward Einstein. History now recalls it as a validation for general relativity because the results held up in subsequent experiments. If they hadn’t, we’d likely remember it as a cautionary tale, if we remembered it at all.
Strevens cites many other instances where scientists played fast and loose with the data, including some pretty famous names. He also describes an observational study of lab work, and just how subjective and ad hoc many of the day to day evaluations of data turn out to be. Science is often said to be self correcting. Which raises the question, where does that self correction happen?
Strevens’ answer is that it happens in the “official” conversation, in venues like scientific journals. There, only one type of argument is allowed: empirical. He characterizes this as “the iron rule”, an uncompromising rule rigidly followed in science communication. The self correction comes if other subsequent papers fail to find the same results.
Note that this iron rule isn’t an assertion about how scientists reason personally, but about how they argue in these venues. For their own reasoning, pretty much any source of ideas are allowed: intuition, philosophy, theology, cultural memes, notions of theoretical beauty, personal biases, competition with other scientists, etc.
All of these influence how scientists do their work and which ideas they pursue. And many, particularly the one about competition, are crucial motivation for doing the often exceedingly tedious work of scientific observation and calculation. Scientists may cite some of them in informal or popular writing. But not in their actual scientific papers. There, Strevens argues, only empirical reasoning is allowed.
When reading this, I initially wondered how theoretical papers fit into this framework. But one of Strevens’ prime examples is Isaac Newton’s Principia Mathematica, pretty much the epitome of a theoretical work. So I think the thesis is that even theoretical work is going to have an empirical orientation.
Another aspect of the rule is that requiring only empirical justification allows for shallow explanations, explanations that pass the iron rule test, but fail many philosophical ones, such as Newton’s description of how gravity worked, while leaving unanswered what gravity itself actually was. (An omission that bothered many of Newton’s contemporaries.)
Strevens characterizes the iron rule as extremely counter intuitive. It doesn’t seem reasonable. Why not allow reasoning from all sources? This is why he thinks humanity went so long before discovering it. He sees it beginning with Newton, in the way Newton personally segregated his empirical and mathematical reasoning from his more philosophical and theological pursuits. Although Strevens recognizes its precursors in people like Tycho Brahe, Galileo, Francis Bacon, and Robert Boyle. But it only became the iron rule after Newton’s success and advocacy of what belonged in “experimental philosophy.”
Why did the iron rule develop when it did in the 17th century? Stevens points to the development of nations as distinct from religious identities, and the need many had to keep the modes of thought from these two domains separate. He argues that what was natural for Newton, may have spread and taken hold in that period exactly because it was a discipline that many people living at that time already had to develop for other reasons.
Toward the end of the book, Stevens ponders the future of science and argues that the iron rule shouldn’t be tampered with. He notes the argument from people like Richard Dawid for a post empirical science. He admits that he can’t be sure it would be disastrous, but sees the possibility as too dangerous to play with.
This is an interesting book, and I think it has some important insights, notably in the distinction between how science is conducted and how arguments are put forth in scientific writing. But the idea that science developed a central dogma: the iron rule, in the 17th century, and that we must adhere to it or risk disaster, makes me uncomfortable.
It is true that investigators became much more serious about observation during the scientific revolution. It was a strong contrast between early modern scientists and medieval philosophers, who tended to rate the authority of ancient figures over observation. But Strevens admits that Aristotle, millenia before the revolution, saw compatibility with observation as crucial. The issue is that Aristotle didn’t see it as the only criteria.
To be fair, Aristotle didn’t have the printing press, a mechanism for rapidly disseminating and responding to ideas. It was arguably this feedback mechanism in the 16th and 17th centuries, particularly with the rise of scientific journals, that made it far more obvious to early modern thinkers what worked and what didn’t.
Arguments grounded in empiricism led to lasting reliable results. Arguments grounded in other things didn’t. We can see in the various statements made by thinkers throughout those centuries leading up to Newton, a gathering consensus on what worked. (With some holdouts, like Descartes, arguing for rationalism.) It seems like if there is a central guiding idea in science, it’s focusing on what works.
And framing something as “the iron rule” makes it sound like science hasn’t changed since 1687. The fact is, what is acceptable in scientific papers is constantly evolving and varies tremendously by field. In many of those fields, what passed muster a century ago might have serious trouble today. In some cases, there have been changes just in the last few years.
Finally, despite the assumptions I made above about theoretical papers, I remain unsure exactly how they fit in with the iron rule. Many theoretical papers make no mention of empirical tests. They leave that to experimentalists. In some cases, it may be decades, or longer, before anyone figures out a way to test a theory. Although we can say those theories shouldn’t be taken as reliable until they pass empirical tests, so maybe they could be seen as pre-rule arguments, ones that shouldn’t be fully accepted until they do pass the test of the iron rule.
What do you think? Is Strevens on to something? Or are my concerns valid?