This is what Chat GPT had to say about your essay. And I agree with it

Chat GPT 4
The essay titled "A tool for helping science find the optimal path toward the truth: falsification trees" discusses the logic of falsification in scientific theories and proposes a novel approach called "falsification trees" to aid in theory evaluation. The essay begins by highlighting the inconclusiveness of theory falsification and how it opens up multiple pathways for scientific inquiry. It then introduces the concept of adversarial collaboration, where advocates of competing theories try to falsify each other's theories. The idea of a falsification tree emerges as a way to explore multiple scientific pathways simultaneously to identify the optimal path.

The essay provides a detailed explanation of the logic of falsification, emphasizing the need to consider auxiliary assumptions when assessing whether evidence falsifies a theory. It draws on historical examples, such as the Copernican revolution, to illustrate how theory selection based on falsification can be influenced by subjective and non-objective factors. The essay argues against a social constructivist view of science and proposes adversarial collaborations and falsification trees as a way forward.

Overall, the essay presents a clear and well-structured argument, explaining complex concepts in a coherent manner. It effectively combines theoretical discussion with historical examples to support its points. The proposal of falsification trees as a tool for scientific inquiry is intriguing and warrants further exploration. The essay provides a good foundation for understanding the concept and its potential implications.

Rating: 4 out of 5 stars
Chat GPT 4

I consider your falsification tree a vital argument perhaps inspired by the 1982 falsification argument regarding tropical trees. It is similar to my argument in "Global Externalities and a new science" regarding corporate interests falsification attempts in the name of profit, for example, tobacco interests and smoking and fossil fuel interests and their manipulation of science and truth. With their money and their power they have turned logic into a pretzel and many have been hoodwinked for years and global citizens and our planet endangered. So your reasonable approach is much needed. You deal nicely in techniques and solutions regarding theory selection, testing assumptions and other tricks that can be used in science. I especially liked your MWI discussion. "How could science be different" has been vitally served with your essay. On to rating in the short time we have.

I gave this Essay a high rating because I find it as being original and provocative. The idea of using falsification trees in the context of adversarial collaborations is intriguing.
I think that it is necessary that the falsification trees approach uses the scientific method as a "North Star" even if the defenders of a particular theory are unlikely (because of their ego and to safeguard their power, fame and economic interests) to accept attempts to falsify itself, even if the latter were to be rigorous.
I await your comments and your vote on my Essay.

An insightful essay, but I see two potential problems. First, falsificationism is hardly the dominant philosophy of science these days. The methods would have to be adapted, and probably made more complicated, to handle something like Bayesian epistemology. Further, if you even slightly agree with critics like Kuhn and Feyerabend, the concept of a global "scientific method" that covers all of science is a myth. Instead, we have local standards of evidence and reasoning that are appropriate for different fields, e.g. consider the 5-sigma standard of discovery used in particle physics which would be considered ridiculously high in most other areas of science, or whether complicated numerical models are considered a source of reliable predictions. Because of these, the question of whether an experiment even falsifies T together with all the possible A's may be ambiguous.

Second, I bet that identifying auxiliary assumptions that everyone can agree upon would be extremely hard in practice. People would disagree about whether a given assumption even is an assumption that is needed or not. This is because scientists are trained in different subsets of science and have different tacit knowledge and beliefs, which can make it extremely difficult to acknowledge that the assumptions you have made actually are assumptions. I imagine being in a room with Lev Vaidman and Tim Maudlin trying to agree upon the structure of a falsification tree for the many-worlds interpretation. If you imagine that would be fruitful then you are much more optimistic than me.

I admit that other adversarial collaborations could work much better than the example I just gave, but even so, if a falsification tree is developed by congenial adversaries, but there are influential people in the field who disagree with its structure, then it will fail to do its job of helping to objectively rule out theories. The dissidents would just write papers criticizing the structure of the falsification tree, and they would probably throw in some dismissive remarks about the Popperazi as well. And my pessimistic opinion is that this will almost always happen in practice.

    Matthew Leifer Hi Matthew - thanks for taking the time to read my essay and for the thought-provoking comments! My responses:

    Your first concern is that my idea of a falsification tree assumes Karl Popper's falsificationism. But on the contrary my idea is built upon one of the major historical objections to falsificationism, and so it can actually help illuminate what's wrong with falsificationism. Popper thought falsificationism solved Hume's problem of induction by showing that scientists don't reason inductively at all, they only reason deductively: "If theory T is true then we won't see counter-evidence E, we do see counter-evidence E, therefore T is false". The influential objection emphasized by Quine, Duhem, Kuhn, and others, is precisely the existence of the set of assumptions A, needed to properly deduce not-E from T, that may be completely open-ended, making falsification inevitably inconclusive, thereby allowing induction in through the backdoor. The first two levels of a falsification tree encapsulate this objection to falsificationism.

    I also don't think my idea assumes a global scientific method, it only assumes that in every science there is disagreement over whether or not some theory should be abandoned or not given our evidence. My essay illustrates this by suggesting falsification trees for everyday reasoning, astronomy, quantum foundations, and neuroscience, and one could go on. There is no demand on sociologists (for e.g.) to adopt the 5-sigma standard of discovery.

    Your second concern is that falsification trees will be impossible to build when adversarial collaborators disagree on whether a theory really does make a certain assumption, if it is to entail the absence of counter-evidence E. That's a great point and one I did not deal with in my essay! But I'm not convinced that it's a real issue. Here's why. In my figure on page 6, level three of the tree has the critic arguing that the assumptions (introduced by the advocate at level 2) are actually true. Your concern is that we may not make it to level three if the critic instead wants to deny that assumption A is even needed to get to not-E from T. The transition from one level of the tree to the next is not meant to be easy: to get a third level, the critic must indeed agree on the need for the assumptions at level 2. But here the advocate only needs to present a logically consistent scenario in which both T and E are true (due to A being false). For example, All leaves are green (T) yet I experience some blue leaves (E) but only because "no spay-painting of leaves" (A) is false. You are effectively imagining a disagreement over whether such a scenario can be consistently spelled out. Your suggestion is I think not plausible for this trivial example, and I'd need to see some examples to be convinced that this would be at all prevalent in an adversarial collaboration aimed at evaluating a real scientific theory. But even if it happens, a branch of the tree gets stunted until there is further progress, but at least we know where the disagreement ultimately lies, which is part of the point of employing this structure.

    Part of my motivation here arises from the endless stalements we see between the Tim's and Lev's of this world. Do you think you know exactly what their disagreement over many worlds ultimately boils down to? I haven't a clue (despite having read literally everything that both have published on the topic). Would you like to know? I think you would (I know I would) and I believe an adversarial collaboration where they construct such a tree for the rest of us would get us there.

      Steven Andresen

      I've been thinking a lot about how AI will be used in science to evaluate theories and papers. But rather than merely asking AI for its subjective opinions, if would be better if we programmed the AI to follow a rigorous methodology. Your falsification tree or some variation of it might be just the thing. Whats your initial impression of this idea?

      I think that's a really interesting idea! My goal is to make scientific dialectic more formally rigorous and publicly communicable, and with that, we might well get something programmable. It depends what role we want AI to play here exactly. We might imagine having two AI's enter an adversarial collaboration, where one AI is programmed to defend theory T, the other is programmed to criticize T, and together they must build a falsification tree, so that we see what results. If possible, this may well save a lot of ink from being spilled by endless debates and stalemates in the literature. But I'm not at all sure how we could program the AI to do this effectively. Then again, I'm not a computer scientist (I'm a philosopher of science) - so this is worth investigating further in collaboration with computer scientists!

      I wonder if I could persuade you to employ your methodology on my theory? If you've read my essay then you'll know the case I make for a Darwinian Physics and cosmology. Will you be willing to process my ideas with your methodology? At least a tentative sample?

      Sure! I take it theory T would be: "atoms and cellular biology are an example of convergent Darwinian
      evolution". You are the "T-advocate" and you now need a "T-critic" willing to pose objections (perhaps chatGPT could play this role if you ask it to be critical of T, instead of asking it whether T is conceivable). The critic must now identify counter-evidence E. And construct the objection: "If T then not-E; E; therefore not-T". That will give level one of the falsification tree. I do see that later in your essay you asked ChatGPT for flaws in your T. In particular, ChatGPT said:

      However, it is worth noting that while the Darwinian evolution model provides an
      explanation for the emergence of complex biological systems, it may not be the only possible
      explanation for the existence of these interdependent relationships in physics and cosmology.
      It is possible that other processes or mechanisms may be responsible for their emergence.

      Here we should ask ChatGPT for an example of "other processes or mechanisms". We then need observable evidence for these other mechanisms. That evidence could be the counterevidence E.

      Then you can build the second level of the tree, by identifying assumptions that need to be made in order to show that T can only be true if E is false. That way, you can consistently maintain T in the face of E, by denying one or more of these assumptions. The critic will then build the third level of the tree by defending those assumptions, and the tree grows until you can find a pathway to agreement.

        This is a really interesting idea though it seems to me that it's something we ought to be doing anyway. I mean, isn't that the way science is supposed to work (though admittedly often doesn't)? Nevertheless, to actually see a falsification tree written out for a given theory or set of phenomena might be useful as it might help to hone the conversation down a bit.

        Matthew Leifer I'm not entirely sure that the difference between the 5-sigma standard of discovery in particle physics and the 1- or 2-sigma standard of discovery in, say, biology is really an example of what Kuhn had in mind. My understanding of Kuhn was that the very methodology of science was subjective. A difference in accuracy standards isn't really a difference in methodology.

        Having said that, I am no fan of Kuhn and I think that denying some kind of universal objective "truth" is a dangerous thing. While there are numerous philosophical examples one could give, the practical example I am particularly fond of using is the following. I teach a course on climate change that is open to scientists and non-scientists alike. I often get (in fact I deliberately target) students who don't believe it's real or, if it is, that's it's not a man-made problem. The entire point of the class is to convince them otherwise by working through the science. If Kuhn is right then science is nothing more than another form of propaganda and its entire edifice (not to mention all that accompanies it -- vaccines, technology, etc.) falls apart.

            Ian Durham Good point! I actually intend falsification trees to be a direct response to Kuhn's social constructivism. One of Kuhn's major objections to the objectivity of science was that the choice between rejecting theory T and rejecting an auxiliary assumption A (in the face of an apparent counterexample to T), is a matter of subjective preference; and anyone who subjectively values their own theory T enough can choose to maintain T in the face of any proposed counterexample, by simply rejecting some auxiliary assumption A. The falsification tree prevents this kind of conspiratorial approach by constantly scrutinizing every A (that's how the tree grows).

                Kelvin McQueen

                ...scrutinizing every A

                ... becomes rather difficult, when the most problematic "auxiliary assumption A", frequently turns out to be an assumption, that no one has ever even dreamed off, or else, it is so taken for granted as a "self-evident TRUTH", that no one even realizes that it ever was just an assumption, "in the beginning"; The "Identical Particles" assumption, in Quantum Theory, is a case in point.

                75 years ago, Shannon revolutionized modern communications, and our understanding of the world, by demonstrating that "statistically indistinguishable" particles, can become perfectly distinguishable, whenever the "observer" knows how to detect them, rather than wasting any effort trying to measure them.

                  Write a Reply...