When asked my opinion on various science-related topics that are in the news, my usual reply is, “I don’t know.” It’s not that I’m incapable of knowing. It’s that I haven’t studied the topics in enough detail to have a well-grounded opinion. My scientific expertise lays elsewhere, in a less popular news cycle.
HOWEVER
If I were asked to develop a well-grounded opinion and had the time to do so, I would follow an approach that has withstood the test of time: the scientific method. My take is that while many have heard of this approach, only a few truly understand it, and fewer still employ it to its full capability. So my objectives here are to 1) share what this method entails, drawing largely from John Platt’s excellent article titled “Strong Inference” (1964), 2) provide examples from the evolution of thermodynamics to highlight key points, and 3) encourage you to embrace this approach in your own work.
Briefly speaking, the first step in the scientific method is INDUCTION. One gathers data, experiences, and observations and then induces a hypothesis to explain it all. In the second step, called DEDUCTION, one assumes the hypothesis to be true and then follows a rigorous cause-effect progression of thought to arrive at an array of consequences that have not yet been observed. The consequences inferred in this way cannot be false if the starting hypothesis is true (and no mistakes are made).
Thermodynamics generally evolved as laid out above. Rudolf Clausius reviewed years of data and analyses, especially including Sadi Carnot’s theoretical analysis of the steam engine and James Joule’s extensive work-heat experiments, and induced: dU = TdS – PdV. J. Willard Gibbs took this equation, assumed it to be true, and then deduced 300 pages of consequences, all the while excluding assumptions to ensure no weak links in his strong cause-effect chain of logic. To the best of my knowledge, he made no mistakes. Multiple experiments challenged his hypotheses; none succeeded. Gibbs’ success led scientists to view Clausius’ induced hypothesis as being true.
In parallel to the above efforts, which established classical thermodynamics, was the work of Clausius, James Clerk Maxwell, and Ludwig Boltzmann, among others, to establish statistical mechanics. One of my favorite examples of the scientific method in practice came from this work. Based on the induced assumption of the existence of gaseous atoms, Maxwell, an expert mathematician, deduced a kinetic model of gas behavior that predicted the viscosity of gas to be independent of pressure, a consequence that he simply couldn’t believe. But being a firm adherent of the scientific method, he fully understood the need to test the consequence. So he rolled up his sleeves and, together with his wife Katherine, assembled a large apparatus in their home to conduct a series of experiments that showed… the viscosity of gas to be independent of pressure! This discovery was a tremendous contribution to experimental physics and a wonderful example validating the worth of the scientific method.
HOWEVER
There’s a critical weakness in the above illustration. Can you spot it? It’s the thought that a single hypothesis is all you should strive toward when seeking to solve a problem.
Be honest with yourself. What happens when you come up with your own reason for why something happens the way it does? You latch onto it. You protect it. It’s your baby. It’s human nature because, when all is said and done, you want to be right. Ah, the ego at work! And it’s exactly this situation that can do great damage to science. People become wedded to their singular “I have the answer!” moments and then go forward, ‘cherry picking’ evidence that supports their theory while selectively casting aside evidence that doesn’t. And it is exactly this situation that inspired John Platt to take the scientific method to a higher level: strong inference.
Platt proposes that the induction process, illustrated below, should lead to not one but instead to multiple hypotheses, as many as one can generate that could explain the data. The act of proposing “multiple” ensures that scientists don’t become wedded to “one.” The subsequent deduction process assumes that each hypothesis is true, whereupon the resulting consequences are tested. Each hypothesis must be testable in this process, with the objective of the test being to effectively kill the hypothesis with a definitive experiment. Recall that a hypothesis can’t be proven correct but can be proven false. All it takes is a single data point. If by logical reasoning and accompanying experimentation the proposed hypothesis doesn’t lead to the specific consequence, then the hypothesis is assumed to be false and must be removed from consideration. As Richard Feynman famously stated, “It doesn’t matter how beautiful your theory is, it doesn’t matter how smart you are. If it doesn’t agree with experiment, it’s wrong.” Only the hypothesis that cannot be proven false, the last one standing, is taken to be the correct hypothesis. Even then, this does not constitute a proof. A hypothesis is only taken to be correct, for the time being, if it offers a means to be tested and if those tests can’t prove it incorrect.
While the illustration below suggests a linear process, in reality, the process is more likely to be iterative. The initiating step typically occurs once a problem or an unexplainable observation is detected. At this point, it is critical that a statement of the problem be written out to ensure clarity and bring focus. As more is learned about the problem, as hypotheses are proposed and tested, as some hypotheses are eliminated and others are expanded to multiple sub-hypotheses, the entire process, with evolving and more detailed problem statements, may repeat itself, over and over, until a single detailed testable hypothesis remains.
Returning to this post’s opening, while I don’t have the time to invest in researching the various sciences being debated today, I do have the time to read those who are doing the research. My criteria for trusting their conclusions? Whether or not they followed Platt’s strong inference model. I want to see the collected data, ensuring that no cherry picking or selective elimination has occurred. I want to see that dissent was encouraged and not ignored. I want to see multiple hypotheses laid out on the table. I want to see an intelligent experimental attack on each and every hypothesis. I want to see the reasoning that leaves hypotheses standing or falling. If I see all of this, then I trust.
I encourage all scientists, no matter the field, to embrace strong inference. Yes, it takes time. But thinking that this process could be short-circuited because you believe you know the answer will eventually lead to problems. As a PhD engineer and friend of mine once said, “some of my biggest errors were when I didn’t follow the methodology.”
A fitting conclusion to this post is the below wonderful quote from Louis Pasteur that captures the essence of Platt’s strong inference model.
“What I am here asking of you, and what you in turn will ask of those whom you will train, is the most difficult thing the inventor has to learn. To believe that one has found an important scientific fact and to be consumed by desire to announce it, and yet to be constrained to combat this impulse for days, weeks, sometimes years, to endeavor to ruin one’s own experiments, and to announce one’s discovery only after one has laid to rest all the contrary hypotheses, yes, that is indeed an arduous task. But when after all these efforts one finally achieves certainty, one feels one of the deepest joys it is given to the human soul to experience.” – Louis Pasteur, Nov. 14, 1888, in a speech given at the inauguration of the Pasteur Institute in Paris.
Thank you for reading my post. The above illustrations are from my book, Block by Block – The Historical and Theoretical Foundations of Thermodynamics. My thanks to Carly Sanker for bringing her great skill to creating them from my ideas. She is an excellent artist.
Special thanks to Jim Faler and Brian Stutts for introducing me to John Platt’s work and also for their contributions to this post
Identifying the Micro behind the Macro 