Thanks, this is a piece of writing that warrants spending a bit more time considering! Clearly written. I definitely want to re-read it a couple of times before commenting.
A tool for helping science find the optimal path toward the truth: falsification
Had another read through. It looks like you're describing a scientific dialectic? One aspect that might be a challenging obstacle to overcome is that this would require the scientific community whole-heartedly embrace null results. Franklin et al point out that:
Modern science’s professional culture prizes positive results, and offers
relatively few rewards to those who fail to find statistically significant relationships in their data. It also esteems apparently groundbreaking results far more
than attempts to replicate earlier research. PhDs, grant funding, publications,
promotions, lateral moves to more prestigious universities, professional
esteem, public attention—they all depend upon positive results that seem to
reveal something new. A scientist who tries to build his career on checking old
findings or publishing negative results isn’t likely to get very far.
A last observation: this "tree" might work well in a field like physics, where replicability is not the foremost challenge. However, it may have significant limits in psychology, social sciences and perhaps to some extent even neuroscience. Because it will depend very heavily on null results, replicability is critical, but these same fields present manifold challenges in that regard.
I hope I have not misunderstood anything! Your paper is great food for thought and I think such a schema has the potential to play a crucial role in a reimagined scientific project that is focused on eliminating bias, unexamined assumptions, and non-empirical influences.
Alex
Thanks, these are really useful thoughts! You are partly right that falsification trees describe scientific dialectics. But they’re more than that: they make explicit the logic of such dialectics. Explicating a tree helps to show where the dialectic should go if the dialectic is to be adequately resolved. So, it’s more prescriptive than descriptive. The point of it is to help improve scientific dialectics, not merely describe them. The goal is to resolve them more efficiently.
On your second point, I don’t think the usefulness of falsification trees depends on replicability. In fact, I think they are useful even in fields that don’t deal with replicating experiments at all. Take philosophy, for example. An objection to a philosophical theory T might take the form “If T then C, not-C, therefore not-T”, where C is a perceived consequence of T (instead of an experimental prediction of T), and “not-C” states that C is an absurd consequence of T. Expanding “If T then C” into “If T&A then C” is a way of saying “you only get C from T by assuming A”. Often, philosophical debates work like this but end in stalemate because they do not explore the tree branches properly. Hence why I want to make the tree structure explicit. So, the usefulness of falsification trees transcends any particular scientific discipline, including those that depend heavily on null results. Having said that, I do say in the paper that I think falsification trees may be more useful in certain disciplines and may be most useful in neuroscience and foundations of quantum mechanics and I offer examples at the end.
I did not understand why you thought my ideas would require the scientific community to whole-heartedly embrace null results. And I did not understand the relevance of the quote, which I believe is from Randall and Welser, not Franklin et al. If you would like to explain what you meant here, I would be very interested. Thanks again for your great comments!
You're right about the attribution (I sourced it from Franklin). My second point is along the lines that, given the (bad science) tendency to pursue positive results, this growing lack of rigor might pose a challenge to the sustainable implementation of the falsification process. Or, more to the point, we'd need to create the underlying culture or motivation to adopt it. It's not a direct criticism so much as a statement about the current state of affairs (which I touch on in my own essay). The null results point was illustrative of the problematic underlying culture.
Alex I see! Thanks for clarifying. It is indeed a problem. I believe that adversarial collaborations have been proposed in part to combat this. These have seen some success e.g. agencies like FQxI and Templeton have been funding adversarial collaborations for precisely this reason. I see falsification trees as a way to help structure and guide what should happen in an adversarial collaboration. If that's successful adversarial collaborations may then receive further support, and so this may be one way to turn the tide on the problematic culture. I'm looking forward to reading your paper.
Indeed! And that has a parallel in the US/Commonwealth legal systems where an adversarial approach is followed to scrutinize evidence and allegations, and in that context it works really well. The adversarial approach to justice was one of the greatest innovations in jurisprudence, and one can see the merits of applying a similar approach in the sciences.
Hi, I enjoyed your essay, and I find it a useful tool to think about specific problems. So thanks for that! I do fear, however, that the recipe “Let's get more systematic in the construction and exploration of the tree” may hit the wall of exponential bifurcations. I wonder, however, whether some progress can be made by analysing the complexity of the trees derived from the different alternatives. In the end, I somehow feel entitled to reject a theory because it has many more branches than its negation. In a way, it is like saying: too many assumptions need to be ruled out for the theory to hold. If the alternative has much fewer branches, it looks less conspiratorial. Or doesn't it? Of course, for this to be valid, one should use a computational standard, so that all assumptions share the same complexity. Only with such a standard does it make sense to count how many assumptions are needed to support each alternative.
quote
One aspect that might be a challenging obstacle to overcome is that this would require the scientific community whole-heartedly embrace null results. Franklin et al point out that:
Modern science’s professional culture prizes positive results, and offers
relatively few rewards to those who fail to find statistically significant relationships in their data. It also esteems apparently groundbreaking results far more
than attempts to replicate earlier research. PhDs, grant funding, publications,
promotions, lateral moves to more prestigious universities, professional
esteem, public attention—they all depend upon positive results that seem to
reveal something new. A scientist who tries to build his career on checking old
findings or publishing negative results isn’t likely to get very far.
end of quote
I beg to differ. Here we go
Case in point, the Cosmological constant problem
By Quantum field theory, its 10120 times bigger than its observed
EXPERIMENTAL results
We DO see the Cosmological constant brought up all the time. I.e. see Sean Carrol, in his
Caltech lectures
It is a myth that there are NO NULL results. Certain categories of problems, are celebrated as now for the time being allegedly UNSOLVABLE. which is a mating call for people to try to do them. I as an example tried it too, and still do.
Secondly the Riemann hypothesis, i.e. this one
quote
Has someone solved the Riemann hypothesis?
The Riemann hypothesis will probably remain at the top of mathematicians' wish lists for years to come. Despite its importance, no attempts so far have made much progress. Nov 11, 2022
end of quote
The goofs galore as to both of these, are widely celebrated and HARD problems, get people to put solutions in
print which get debated.
Even the failures are interesting and instructive: Very instructive
We can look at the idea of "null results" as maybe tied into this one
quote
What does Gödel's incompleteness theorem say?
Can you solve it? Gödel's incompleteness theorem ...
In 1931, the Austrian logician Kurt Gödel published his incompleteness theorem, a result widely considered one of the greatest intellectual achievements of modern times. The theorem states that in any reasonable mathematical system there will always be true statements that cannot be proved.
end of quote
So there you go.
HINT null results occur all the time. And the INTERSTING failures may at times even lead to great papers.
I have seen that more than once
Andrew Beckwith I see what you are saying, but my emphasis is a little different (take a look at my subsequent discussion with the author). The comment you've quoted also has more to do with this problem you identified in respect of another essay on here:
"In a word, the bloat in this is in huge facilities like CERN with few rivals as to cross checking. As an example the Fermi data sets hinted directly at the Higgs, but when Fermi was partly decommissioned, it took YEARS before some of the signal analysis of the signals were dredged up as a partial comparison as to the Higgs experiment in CERN."
I think you are correct in your comment above.
I would really appreciate your thought on my paper ("Self-Appropriation and Respect in a Moral Scientific Community"), as I'm hoping for feedback on my reasoning there! Do you have an essay submitted?
Have a good day!
Andrew Beckwith Thanks for the great examples! You say "It is a myth that there are NO NULL results." I wonder if you also think it is a myth that "theory falsification is always inconclusive", which is how I put it at the beginning of my paper?
Note that I'm talking primarily about scientific theories, not mathematical theorems. The top level of the falsification tree on page 6 of my paper begins with an empirical objection (E) to the observable predictions of a scientific theory (T).
Alex Nice analogy! That makes me wonder whether falsification trees would be useful in legal disputes. They might also help make transparent to others why, in controversial cases, the evidence was deemed sufficient to incriminate someone.
- Edited
Ines Samengo Thank you! Great points. I briefly consider the problem of exponential bifurcations at the top of page 7 where I propose a solution involving team adversarial collaborations. The idea is that two teams of researchers compete in an adversarial collaboration, sub-teams bifurcate as the tree branches. Perhaps that will just lead to many 1-on-1 adversarial collaborations, where each collaboration becomes overwhelmed with further bifurcations. But if they publish where in the tree they got to, then the rest of the scientific community can help. And there may be efficient ways of shutting down branches as "dead ends", preventing further growth in those directions.
I think in many of the most interesting cases, it is going to be rather difficult to tell whether a theory has more branches than its negation, like my examples on page 7 of neuroscientific theories of consciousness and interpretations of quantum mechanics. But constructing falsifications trees may help determine this.
Dear CoralBear,
You have written a very important essay that can really change science towards the study of reality instead of the study of abstractions.
Therefore, I highly appreciate and understand your work.
My essay is devoted to the key facts that lead to new key laws that are not noticed by the generally accepted concept. But these laws may form a new science of studying reality without studying abstractions. I think you will also be interested in the elements of the deterministic functioning of the quantum solar system on the new laws that are given in the appendix, and which are similar to the quantum laws of the functioning of the Hydrogen atom.
I wish you success!
Dear CoralBear,
I really wanted to thank you for the comments about my essay, to which I have now replied, however, your very interesting essay did make me think of another path to truth. .
So, if a theory consists a a small number of very simple rules that have to be repeatedly applied many many times to achieve a result, and that result matches reality, then that theory is likely to be true. This assumes that the more times you apply the rules, the more likely any error in the theory is to show up, so if there are no errors after a large number of iterations, then your theory is probably on the right track.
It is like being shown a fractal and being asked to reproduce it with your own computer program. You might be able to do this with a complicated program, which you then change to accommodate the changes in the fractal as it grows (so that the predictions of the program match the fractal), and there might be lots of different complicated programs that could do this. However, if you found a really simple program that repeated in a loop over and over again and continuously and accurately reproduced the fractal, then you would rightly think you had most probably discovered the original fractal program.
Perhaps you could call this proof by construction.
I would like to think this is an original idea - but I'm sure someone else will have thought of it!
All the best.
I read your essay with great interest and I really appreciate your idea of falsification trees.
It could be useful to force a system of reciprocal falsification. In some disciplines such as theoretical physics, scientists seem to have reached a high degree of "autism" for which it is difficult to communicate outside one's own niche. This implies that, as you rightly note in your essay, a model can hardly be falsified except by someone outside that research niche.
It is important that the falsification takes place in compliance with the scientific method, on the scientific merits of the models. Those who are subject to falsification must be able to know the reasons for the criticisms raised and must have the right to defend themselves, a bit like in the peer-review. Otherwise, one falls into the highly anti-scientific behaviors and oscure politics that characterize arXiv, as I describe in my essay "The Name of the arXiv: when too much zeal is an obstacle to science".
Furthermore, it seems to me that your model of falsification assumes that scientific progress takes place through successive refinements of ideas. This is not always true, and some cases in which scientific progress takes place through revolution rather than evolution are precisely the ones you give in your essay.
Kelvin McQueen
<<My hope is that this will accelerate science down a route that leads us to nature's most fundamental truths.>>
I believe that it is necessary to completely abandon the idea of "falsification" and deal with the solution of the "millennium problem No. 1" - the ontological justification of mathematics (ontological basification), and hence knowledge in general. That is, the construction of an ontological basis of knowledge: an ontological framework, carcass, foundation.
Any theory that claims to be called "fundamental" must be ontologically justified.
Other working theories are "effective". Like quantum theory and general relativity, for example.
Here is the so-called. "The Big Bang Theory". In your scientific picture of the world there is a so-called. "Big Bang"?
Doctors of Physical and Mathematical Sciences Yuri Vladimirov notes in the article "PRINCIPLES OF METAPHYSICS AND QUANTUM MECHANICS" (No. 1, 2017, p 10):
<<At present, the main goal of theoretical physicists is to build a holistic (monistic) physical picture of the world based on a single generalized category. At this point in time, it is "seen" (interpreted) differently from the standpoint of the three named paradigms: a single vacuum in the field theory approach, a single geometry in the geometric worldview, or a single system of relations (structure) in the relational worldview. In our opinion, these are different names for the same desired physical (metaphysical) first-beginning.>>
[https://lib.rudn.ru/file/Metaphysics%20#1_23_2017%20print.pdf]
What does your method say?
Have you tried to build your "holistic (monistic) physical picture of the world on the basis of a single generalized category"?
What is in your scientific picture of the world "the desired physical (metaphysical) first-beginning". What is its ontological structure?
- Edited
Hi CoralBear
I see you have 6 ratings and so need another 4 to qualify for the next stage of the contest. As do I. Would you like to help each other get across the line by reading and rating each others essays over the weekend? Mine is titled "Age of Knowledge"
Cheers
Swan
Steven Andresen Sure, I'll go ahead and read and rate yours now. I was not aware of the 10 rating minimum to qualify - where did you see that?
Vladimir Rogozhin
<< I believe that it is necessary to completely abandon the idea of "falsification" and deal with the solution of the "millennium problem No. 1" - the ontological justification of mathematics >>
Thanks, I appreciate your comments. These are compatible, I believe. My essay recommends a strategy for using adversarial collaborations to more efficiently find solutions to scientific problems. So, I would envisage researchers who disagree about the ontological justification of mathematics collaborating and constructing falsification trees to map their collaboration. A proposed falsification need not be an experimental counterexample, it could also be, for example, a philosophical objection.
Donatello Dolce
Thanks, these are great comments! I'm curious to know why you think my model of falsification assumes that scientific progress takes place through successive refinements of ideas, since as you note, one of my primary illustrations was the Copernican revolution.