Converging Streams Form Raging Rivers

In this installment of the series I want to illustrate something very important in science and history -- the principle of converging evidence or “consilience” as it’s known in science. This is something Batchelor and Company use every single day, just as all humans do, and yet that group minimize, bob, weave and avoid acknowledging it with all their might when it regards the earth sciences.

While it might sound like a word from a Monty Python comedy skit, consilience says that evidence from independent, unrelated sources can converge to strengthen conclusions we make. This means that when multiple sources of evidence are in close enough agreement, the conclusion can be really strong even when the individual sources might not seem as strong individually.

Here's Webster's: CONSILIENCE: "the linking together of principles from different disciplines especially when forming a comprehensive theory."

Another way to describe this principle is that it's based on the ‘unity of knowledge’, meaning that if we measure the same thing several different ways, it should lead to answers that are very similar if not identical. Let’s take the Golden Gate Bridge for instance and measure its length using a laser rangefinder, or satellite imaging (GPS) or even a simple yardstick. If those three methods are any good at measuring things, they will all three return a measurement that’s close to the same.

Since each method relies on different natural phenomenon, if one method is in error it is extremely unlikely to be in error in the same way as the other two and a difference in result will be seen. For the same reason, if independent dating methods in geochronology are credible they should give us similar dates when tested independently. The principle is applied comparatively to entirely different fields of science where the results from chemistry or astronomy shouldn't contradict a result in geology, etc.

When several independent methods agree, this is strong evidence that none of the methods are in error and the conclusion is correct. For a group of converging measurements to be wrong, the errors would need to be similar for all measurements taken, which is extremely unlikely. This is how scientific theories reach a high degree of confidence – over time they build up a large body of evidence which converges to the same conclusions.

While thinking about how to illustrate this principle, I realized that we use consilience regularly while navigating from place to place in our daily lives. If we accept a single piece of evidence as golden, we might just end up knocking on the wrong door. If we look at the whole of the evidence and how it may or may not converge, we will far more likely be successful.

As an example, let’s say you and your running partner Phil are going for an outing. He has invited his sister Nico and you are to meet them both at Nico’s house – a house you have never seen. Really all you know about Nico is that she is recently married for the first time and according to Phil, she works as an architect and her husband is an IT professional.

In a voice mail, Phil gave you rough directions and the address as “123 Smitt St.” He told you that Nico’s house was an old but recently restored green Victorian classic with white trim and is located right next door to large white Presbyterian Church. Phil told you that he was going over early and would be there when you arrived. You entered the address into your navigation app and headed across town when it was time.

You followed the instructions on your smart phone and soon enough you were at the prescribed address – The street name was correct and right there under the porch light it said “123”.

Even as you earlier turned onto Smitt St. however, you noticed that you were entering a brand new residential tract and you began to feel a sense of unease regarding the accuracy of the directions. You had certainly imagined an older part of town with perhaps more trees would accompany such a landmark home. The house before you is certainly some shade of green and there is a vague sense of Victorian about the design, but the only church nearby is brick, Baptist and across the street. There is an older car in the drive, kids toys strewn amongst the weeds in the lawn and Phil's car is nowhere in sight.

Do you knock on the door? You might give Phil a call and ask why his car is not there. You might listen to the voice mail and confirm the address. What you likely won’t do it just happily move forward assuming you have found the correct house. Why not? Well, the lines of evidence have not converged to one safe conclusion. There is a feeling of disarray among the combined indicators. There is no consilience of evidence.

Of the different pieces of evidence available, only the numbers on the side of the house fit in with the entire expected picture when thinking of a recently married professional couple with no kids. Sure enough, when you double check the address, you realized that it was “Smith St.” and not “Smitt St.” that should have been entered into your smart phone. One letter makes all the difference and it turns out that Smith Street is just a few blocks over.

Let’s pause for a moment and consider this case where the surroundings or what we might call ‘secondary evidence’ didn't fit in with what most would call the ‘primary’ evidence – the address. Even though at the time you believed your location was totally in line with the primary evidence, you realized your error simply by looking at the convergence or consilience of the other lines of evidence. When you unconsciously took the principle of consilience into account, you were able to recognize that the chances of an error in your primary evidence were extremely high.

Let’s flip the situation. You immediately reprogrammed your smart phone for the proper address and followed its instructions. Once it told you that you were in the vicinity of your goal, you looked around and saw a large white church surrounded by decidedly early 20th century homes. As described, a cute, well maintained classic Victorian design home with white trim sat right next door to the stately church. Sure enough your hiking partner's car was parked next to the curb right by the entrance gate with steps leading up to the door. You parked your car behind Phil's, walked up the steps and confidently rang the bell – quick as a flash you are inside and as it turns out, you never even looked to see if the number on the mailbox was indeed “123” – you didn't need to.

So how do these two situations differ? In both cases you considered multiple lines of evidence and how they converged to come to an easy conclusion regarding the unity of knowledge. Each individual piece of evidence was considered as part of the whole. Not just the address, but the size of the trees, the style of house, the nearby church, kids toys, Phil's car etc. In the first case, the convergence of the secondary evidence was strong enough in opposition of the primary evidence to overpower the primary evidence and cause you to correctly question its validity. In the second case, the consilience of the secondary evidence was so strong in favor as to render the primary evidence literally unnecessary. Phil could have given you nothing more than the description and turn by turn directions and without knowing the street address you would have still knocked on the same door without hesitation.

Using the principle of consilience, we can much more quickly gain or lose levels of confidence in an idea, assertion, conclusion or hypothesis. Consilience is often critical in elevating a hypothesis to a theory. More independent lines of investigation reaching the same result lead to higher confidence in the ultimate conclusion. If one line of evidence produces a result that is at odds with the consilience of other multiple investigations, it’s likely that the error will be found in the single line of evidence rather than in the converged streams – and so that line of evidence becomes the focus of testing.

Consilience does not forbid deviations, but science wisely doesn't thoughtlessly allow one piece of evidence to overthrow a large body of converged work. It first focuses on learning and understanding why the single stream differs. Perhaps there’s an error. Perhaps there’s a misunderstanding. Perhaps discovered knowledge will lead to coherent integration of the odd stream. This was aptly demonstrated in our navigation example.

Science is never settled. Science is also not a democracy and we would do well to remember that scientific consensus is not the same as political consensus. Where the former is achieved through thoughtful investigation, the accumulation of independently converged evidence and intelligent discussion, the latter is often arrived at through a quagmire of opinion, compromise and presuppositional bias. Consilience and scientific consensus are a numbers game of evidence while political consensus is a numbers game of adherents. Consensus is great, but a scientific consensus built on anything other than consilience of evidence is hollow ground and not good science.

So how do we tell the difference between scientific consensus and political consensus? We must rely on that most productive backbone of science – the prediction. Good science leads to accurate and useful predictions. Consensus (scientific or otherwise) without testable predictions is mere dogma.