- You may want to read Freud's early work, from Studies in Hysteria to his metapsychology papers like "The Unconscious" where he shows in detail his thought processes, coming up with hypotheses and refuting them. The classic critique of Freud that you see in Popper is a gross misrepresentation, even a critic of psychoanalysis like Grunbaum thinks it's bad.
You might be interested in the French school of psychoanalysis, which honestly makes much more sense in the context of French philosophy of science, which was rationalist compared to the Anglophone-analytic tradition. There's a very powerful and interesting epistemological tradition from Bachelard, Canguilhem, Cavaillés, Koyré—which spawned a psychoanalytically informed epistemological tradition in the Cahiers pour l'Analyse, and Lacan wrote a few key texts on science, like the brilliant "Science and Truth". Jean-Claude Milner has a brilliant book on Lacan and science, "A Search for Clarity" that I recommend, Lacan is much better imo than the characterization you see in Sokal and others. Lacan's analysis anyway is that psychoanalysis is not a science, but has the emergence of science as a condition—I recommend Milner's account.
Also, Freud wrote up one of the first diagrams of neural networks in his unpublished Project for a Scientific Psychology. There's a book, Biology of Freedom, by a Lacanian and a neuroscientist, on this topic. Also Liu's awesome book Freudian Robot on the intersection between psychoanalysis and the history of AI
Freud's Introduction to Psychoanalysis - it';s a transcript of his lectures for the course of the same title in Vienna - is also quite useful and shows his scientific bent.
>All Bayesian inference is subjective! Subjectivity is the foundation of the entire Bayesian statistics program.
Not sure exactly what you mean by this, but its a bit triggering :-). If by subjective you mean something like 'dealing in degrees of belief' and/or 'depending on external prior information' than I would much rather be the explicitly "subjective" Bayesian than be implicitly credulous and even outright wrong the way many routine applications of frequentist statistics are.
What I was saying here (and what Meehl is saying in the Lecture) is just because some inference is subjective does not mean it is invalid. And it does not mean that the method isn't empirical.
I 100% agree with you. One of the best selling points of Bayesian methods is they encourage explicit statement of assumptions.
I know of that paper but only have read the abstract. I should give it a closer read.
For whatever it's worth, I think we'd be better off embracing how subjectivity and normativity pervades scientific thinking and quantitative methods. I think this, more than p-hacking or QRPs, better explains why some programs run into stalemates while others flourish.
That's interesting. I had absorbed the idea that psychoanalysis didn't work, but i realise now that this was probably from reading Eysenck before his frauds were exposed
Adolf Grünbaum’s philosophical critique is still worth reading, which Meehl acknowledges in one of his papers. However, going by the criteria Grünbaum relies on, astronomy would have to be classified as unscientific as well.
A major difference is that you can RCT psychoanalysis because you can create an abstract reference class for "treatment" and "outcome." But this puts us into the dumb box where "science-based" = "RCTs"
There’s also the effect the analyst has on the patient — by presenting his conjectures regarding the latent content, the analyst can exert influence on the patient’s future behavior (even if it’s just introspection).
There are definitely studies that show "benefits," though if we watched the rest of Meehl's course, it's hard to conclude one way or the other that psychoanalysis works. You can convince yourself of which ever side you'd like to believe.
By the way, the distinction Willems makes between “manifest variables” and “latent variables” is taken straight from psychoanalysis:
“The distinction between manifest (signal) variables and latent (auxiliary) variables as used here is quite appropriate in analogy with the use of these terms in psychology. Latent variables are reminiscent of Rosenbrock's partial state and of the internal variables used off and on in electrical circuit theory.”
The bait and switch of all therapy being presented as “scientific” or “evidence based” is what we are actually referring to by this. It is evidence based that psychodynamic therapy improves clinical outcomes for those who engage in it; this is different than the theory, interpretations, principles are scientific or evidence based. These are not testable (or readily testable, as you mention), falsifiable ideas. It then falls to what your philosophy of science, but at least in regards to the rest of medicine, it is unscientific. Of course. this does not also not mean it is valueless or not an approach to knowledge, but it is a different mode of inquiry
Some notes
- You may want to read Freud's early work, from Studies in Hysteria to his metapsychology papers like "The Unconscious" where he shows in detail his thought processes, coming up with hypotheses and refuting them. The classic critique of Freud that you see in Popper is a gross misrepresentation, even a critic of psychoanalysis like Grunbaum thinks it's bad.
You might be interested in the French school of psychoanalysis, which honestly makes much more sense in the context of French philosophy of science, which was rationalist compared to the Anglophone-analytic tradition. There's a very powerful and interesting epistemological tradition from Bachelard, Canguilhem, Cavaillés, Koyré—which spawned a psychoanalytically informed epistemological tradition in the Cahiers pour l'Analyse, and Lacan wrote a few key texts on science, like the brilliant "Science and Truth". Jean-Claude Milner has a brilliant book on Lacan and science, "A Search for Clarity" that I recommend, Lacan is much better imo than the characterization you see in Sokal and others. Lacan's analysis anyway is that psychoanalysis is not a science, but has the emergence of science as a condition—I recommend Milner's account.
Also, Freud wrote up one of the first diagrams of neural networks in his unpublished Project for a Scientific Psychology. There's a book, Biology of Freedom, by a Lacanian and a neuroscientist, on this topic. Also Liu's awesome book Freudian Robot on the intersection between psychoanalysis and the history of AI
Thanks for the references!
Freud's Introduction to Psychoanalysis - it';s a transcript of his lectures for the course of the same title in Vienna - is also quite useful and shows his scientific bent.
>All Bayesian inference is subjective! Subjectivity is the foundation of the entire Bayesian statistics program.
Not sure exactly what you mean by this, but its a bit triggering :-). If by subjective you mean something like 'dealing in degrees of belief' and/or 'depending on external prior information' than I would much rather be the explicitly "subjective" Bayesian than be implicitly credulous and even outright wrong the way many routine applications of frequentist statistics are.
Oh sorry, I did not mean it to be a dig!
What I was saying here (and what Meehl is saying in the Lecture) is just because some inference is subjective does not mean it is invalid. And it does not mean that the method isn't empirical.
I 100% agree with you. One of the best selling points of Bayesian methods is they encourage explicit statement of assumptions.
I assume you've seen this? I think of it every time someone applies subjective/objective to a statistical method: http://www.stat.columbia.edu/~gelman/research/published/gelman_hennig_full_discussion.pdf
I know of that paper but only have read the abstract. I should give it a closer read.
For whatever it's worth, I think we'd be better off embracing how subjectivity and normativity pervades scientific thinking and quantitative methods. I think this, more than p-hacking or QRPs, better explains why some programs run into stalemates while others flourish.
That's interesting. I had absorbed the idea that psychoanalysis didn't work, but i realise now that this was probably from reading Eysenck before his frauds were exposed
Adolf Grünbaum’s philosophical critique is still worth reading, which Meehl acknowledges in one of his papers. However, going by the criteria Grünbaum relies on, astronomy would have to be classified as unscientific as well.
I mean, you can make a case, no?
Both are observational, so yeah, you can.
A major difference is that you can RCT psychoanalysis because you can create an abstract reference class for "treatment" and "outcome." But this puts us into the dumb box where "science-based" = "RCTs"
There’s also the effect the analyst has on the patient — by presenting his conjectures regarding the latent content, the analyst can exert influence on the patient’s future behavior (even if it’s just introspection).
There are definitely studies that show "benefits," though if we watched the rest of Meehl's course, it's hard to conclude one way or the other that psychoanalysis works. You can convince yourself of which ever side you'd like to believe.
I have to register my objection to using "latent" as a noun, especially in plural form.
Noted. But I'm going to play my poetic license card.
By the way, the distinction Willems makes between “manifest variables” and “latent variables” is taken straight from psychoanalysis:
“The distinction between manifest (signal) variables and latent (auxiliary) variables as used here is quite appropriate in analogy with the use of these terms in psychology. Latent variables are reminiscent of Rosenbrock's partial state and of the internal variables used off and on in electrical circuit theory.”
The bait and switch of all therapy being presented as “scientific” or “evidence based” is what we are actually referring to by this. It is evidence based that psychodynamic therapy improves clinical outcomes for those who engage in it; this is different than the theory, interpretations, principles are scientific or evidence based. These are not testable (or readily testable, as you mention), falsifiable ideas. It then falls to what your philosophy of science, but at least in regards to the rest of medicine, it is unscientific. Of course. this does not also not mean it is valueless or not an approach to knowledge, but it is a different mode of inquiry