The Structure of Scientific Revolutions
A book review on “The Structure of Scientific Revolutions” by Thomas Kuhn
The Structure of Scientific Revolutions: 50th Anniversary Edition
TL;DR
The Structure of Scientific Revolutions is an essay, written by Thomas Kuhn, which begins by arguing that the way science develops over time is actually not through accumulation but through a series of revolutions in which the foundations of a scientific field that were previously accepted had to be abandoned in favor of an entirely different set of foundations (where one paradigm replaces another). Throughout the essay, the author uses many historical scientific examples to ground his argument. The remainder of the book is on the structure of the revolutions that occur including: How revolutions happen in the first place, what a revolution looks like, how resolutions are resolved, and how progress is made through them.
Why I Read This Book
Over the past 7 or so years, I have spent most of my data science learnings focused on technical fundamentals, techniques, algorithms, programming, etc… The core technical details. But I felt like I have been missing the science component. I have never done proper scientific research.. frankly I’m not even sure what it means to do it. In this book I wanted to get a better understanding on the history of science and help me answer the question: What is science?
What Stuck
This book particularly resonated with me because of my experiences working in the data science field, and especially in my current role at Credit Karma. There are many times I’m asked to figure out and explain what is going on in our app. For example, I could be looking into why a certain product is getting a low amount of user engagement, or in a particular month why our funnel efficiency decreased on ios. In a typical scientific process, I usually come up with a few hypotheses and then start looking at the data to either back up a hypothesis or reject it. In this process it is common to go through a series of mini revolutions (this is an extreme simplification of the concept), where I’ve constructed a story (theory) of what is going on, I start putting materials together, and then another question pops up in which my theory doesn’t seem to answer. I have to go back to the data and either expand my initial theory or reject it and start from scratch. Eventually I land on a new theory that better explains and answers all the questions. Under this new theory, the same data I looked at under the old theory now tell an entirely different story.
Particular Concepts
The same scientific methods we use today are the same ones that have lead to prior incorrect theories. While Science is the best option we have to understand the natural world, it does not answer every question and can often lead to incorrect understanding. This is not a fault of science, but the fact that a paradigm can appear to answer many previously unanswered questions, and only through continued scientific practice is it discovered that certain items can’t seem to be solved under the present paradigm.
Most normal science is analagous to puzzle solving. The fundamentals, norms, techniques, and tools provided under the current scientific paradigm allow normal scientific practice to focus on solving an esoteric problem in which the paradigm says is solvable. Meaning working on something entirely novel and making new discoveries, while this is the image I typically have in my mind as what a scientist does, it does not represent reality.
Science is not some piecemeal process in which you can say, for example, that lavoisier discovered oxygen in 1778. There were multiple people who discovered “oxygen” around the same time and not everyone agreed on what it actually was. To some it was dephlogisticated air (according the phlogiston theory which is now rejected, oxygen was understood as air deprived of phlogiston). At what point is the discovery made? The first time someone notices a new type of gas? When it is observed and understood? When it is correctly fit into a paradigm? This is much more of a gray area than is recorded in many scientific texts.
After undergoing a revolution and performing science under a new paradigm, the exact same results and observations made previously are now seen in a new light with an entirely different understanding even though nothing physical changed about the observation or result.
Key Terms
- Paradigm, “The combination of law, theory, applications, and instrumentation which spring particular coherent traditions of scientific research.”
Book Summary
Science Progression by Revolutions, not by Accumulation
The Structure of Scientific Revolutions is an essay, written by Thomas Kuhn, which begins by arguing that the way science develops over time is actually not through accumulation but through a series of revolutions in which the paradigm that was previously accepted had to be abandoned in favor of an entirely different paradigm. I’ll use a castle analogy to better explain this point. If the entire body of scientific knowledge were to represent a castle, then science by accumulation is to mean that the castle was built such that the first scientists to come around laid the foundation, and the scientists who followed built on top of that, and the building continued until we reached the peak of the tower, but we can still see the work of the early scientists holding everything up in the stones at the bottom. Science as a series of revolutions is to say that the first scientists built a castle which held strong for a period of time, until the castle was destroyed in a battle and the new generation of scientists had to rebuild the castle from scratch. This destruction and rebuilding happens over and over again such that the tower that stands today has little to no resemblance of the tower that once stood.
Normal Science as Puzzle solving
A scientific paradigm, which is designated for a particular group of scientists and field of focus, defines the boundaries in which normal science can be practiced. Given a paradigm, there are a few primary functions that normal scientific practice serves:
“Determine a set of facts that the paradigm says is particularly revealing of things”. For example, calculation the atomic mass of elements in the periodic table, densities of a variety of materials, wavelengths, boiling points, etc.
The attempt to match paradigm prediction to reality. For example, say a paradigm would predict that the speed of light is greater in air than it is in water. This type of activity could involve building special tools or equipment to measure these two different speeds and validate what was predicted.
Further articulation of a paradigm. For example, say the paradigm states that gas pressure is related to volume. To better articulate the paradigm you would want a particular law that gas pressure follows with a change in volume (Ex. Boyle’s Law). It can be the scientists role to determine what that relationship actually is? is it a constant? does it change based on the type of gas?
Given normal science falls into these categories, science then becomes puzzle solving. The paradigm tells the scientist that there should be a solution, and provides the rules and restrictions to the problem, it is up to the scientist to then devise that solution. If the scientist fails to come to a solution, it is typically considered a failure of the scientist rather than a failure on the paradigm to allow a solution to be found.
Scientific Revolutions and Their Structure
A scietific revolution is a period of time in which one scientific paradigm transitions to another paradigm. There are problems the current paradigm doesn’t seem to solve, allowing alternate paradigms to compete. Over time an increasing number of scientists begin to work under the new paradigm, leaving some of the older scientists working on the prior paradigm. Not everyone makes the transition, some resist and stick to their old views even until the end of their life.
Structure
Anomaly and the Emergence of Scientific Discoveries
Normal science consists of solving puzzles that are known to be solvable under the paradigm. But what if over the years, and after attempts from multiple scientists, there is a continued failure to produce a solution. In addition, what happens if when performing an experiment, the scientist makes an observation that does not fit in with the paradigm? An example of an observation that does not fit in with the paradigm is the discovery of X-rays. Willhelm Roentgen was performing a typical experiment when a material in his lab started to glow caused from X-Rays, which as the time was initially unexplainable. These types of discoveries or anomalies can cause scientists to question the current paradigm, or begin to make modifications to it.Crisis and the Emergence of Scientific Theories
In the case where particular puzzles cannot be solved, or discoveries have occurred, the paradigm can enter a state of crisis. This is where the paradigm starts to blurr, there become many different versions of theories, where scientists have to make modifications to continue work. There is no standard agreed upon set of rules. And the continued work in the field raises more questions than answers.
“So long as the tools a paradigm supplies continue to prove capable of solving the problems it defines, science moves fastest and penetrates most deeply through confident employment of these tools. The reason is clear. As in manufacture, so in science – retooling is an extravagence to be reserved for the occasion that demands it. The significance of crises is the indication they provide that an occasion for retooling has arrived.”
In other words, you decide on a set of foundations and tools, and with acceptance, you work under assumptions that these are all correct which allows for an incredibly high velocity in work. Only when in a crisis state, whether increasingly solutions are failing to appear, or whether time to completion is balooning, do you need to retool, otherwise it is an extravagence. From the perspective of a data scientist, imagine every time you want to solve a regression problem you have to think, is minimizing the sum of squares really the best way to minimize error (a past Will would quickly go there)? That is exhausting and the difference from coming to a solution in a couple of hours vs a week. In some companies, data scientists have the luxury to explore all these types of questions, whereas in others, in order to come to an answer in a reasonable amount of time you have to stand on a set of foundations and assume certain techniques are ok to use for a particular application.
- Response to the crisis
All crises close in one of three ways:
- “Normal Science ultimately proves able to handle the crisis-provoking problem”
- “Scientists may conclude that no solution will be forthcoming in the present state of their field. The problem is labelled and set aside for a future generation with more developed tools”
- “A Crisis will end with the emergence of a new candidate for paradigm”
- Resolution and Progress of Revolutions
In the event of crisis, new theories and paradigms start to form. There are new camps of scientists, some that favor the existing paradigm and some that favor the new paradigm. There is always pushback on a new paradigm. Imagine working under a certain set of foundations for 10+ years and all of the sudden being told they could be incorrect. The scientist had likely solved many problems under the current Paradigm. There are years and years of work that fall under the current paradigm, which could now be considered throw-away. Not to mention the new paradigm doesn’t solve all the problems in the field. A paradigm rarely solves all of its problems otherwise there would be nothing further for the scientist to work on and the field would become a collection of technical tools used by the engineer.
And sometimes: “a new scientific truth does not triumph by convincing its opponents and making them see the light, but rather because its opponents eventually die, and a new generation grows up that is familiar with it” (Example: Karl Pearson as noted in The Lady Tasting Tea)
In the end, the paradigm that eventually wins out is the one that has the highest likelihood to solve more of the field’s problems.
- The Invisibility of Revolutions
To many people, the view of science progression appears to be linear. Why is science through revolutions not obvious? One of the reasons is that a primary method of teaching a paradigm to a new generation is through its textbooks. “Textbooks begin by truncating the scientists sense of the discipline’s history and then proceed to supply a substitute for what they have eliminated.” The way new scientists learn about a specific discipline is often through texts. If the text does not provide how the science got to where it is today, then the past revolutions will be hidden. I can relate here. In my early years of education, we rarely learned about the history of a field, but rather how the current paradigm came to be and the status of that paradigm today with its current applications.