0 reactions
FoogHow to sort through crankery?
**Disclaimer:** I am not trying to put down any groups of people; there have been many threads on here and StackExchange which discuss the reasons for bad mathematical work and its consequences. I would venture to say that nearly student, especially in a discipline like math which focuses on adding layers of abstraction https://en.wikipedia.org/wiki/Fundamental_theorem_of_software_engineering#:~:text=The%20fundamental%20theorem%20of%20software,an%20extra%20level%20of%20indirection.%22, oes through a brief crank phase (consider the Dunning-Kruger effect), but the problem does not stop at writing easy-to-debunk research papers.
Now for the post:
Whenever I feel a bit suffocated by work in arithmetic/algebraic geometry, I try to "take a breather" by exposing myself to different (e.g. model theory, metric number theory), sometimes niche (e.g. hypercomplex analysis, differential algebra), concepts in math. Problem is, whenever you stray too far from the more modern, serious research areas (say, arithmetic geometry, PDEs, homotopy theory, etc.) and into more "enticing" stuff (say, fuzzy logic, anything "quantum", etc.), you are met with a lot of cranks. While it's easy to spot someone trying to do angle trisection or prove GRH with "10th grade math", it's not so easy to judge a book written by an associate professor at some obscure university working in a niche research area.
I've personally interacted with some serious cranks through a complex systems/AI lab I used to be a part of. While I'm perfectly capable of spotting if something is "wrong", I had wasted a ton of time listening to some hand-wavy, superficial babble. As such, I am hesitant to read many books/papers as 1. it is difficult to judge something you are currently learning (something, something, Dunning-Kruger...) and 2. it is risky to spend days reading through something and trying to make a sound judgement after the fact.
Let's say I'm looking at a book. Here are some (soft) filters I like to go through:
* *What are the authors' educational backgrounds?* Even if they have a PhD, but don't work in a related area, I'm cautious.
* *What is the authors' publication history?* It's an easy instant disqualification for even a single viXra paper, but it's tough to sort through conference and journal rankings, especially in hyped topics like AI. I also get cautious when it's tough to find any of their publications except for two papers on academia.edu r the like, or they don't have a website.
* *What does the book read like when I skim through it?* In the preface, do they claim to solve an unsolved problem over the course of the next god-knows-how-many pages? Do they (rigorously) prove things?
* *What about the publisher?* Cambridge University Press has solid work, but Elsevier does little quality control.
* *Using Scott Aaronson's list https://scottaaronson.blog/?p=304 r John Baez's (humorous) index https://math.ucr.edu/home/baez/crackpot.html.* airly self-explanatory.
For example, yesterday I was looking into Giardina's "Many-Sorted Algebras for Deep Learning and Quantum Technology". He's got a PhD, seems to be a former researcher in a "Bell Labs" type of environment, but is talking about obscure algebraic logic in the context of two hyped areas. The book also does not have proofs, just a bunch of "examples". Ultimately, I'm not going to chance it (unless one of you has some informed insight). **Another disclaimer:** I am not trying to put down Giardina; I'm just trying to illustrate how I have a certain amount of chips to gamble (say, one chip = 1 hour) and I want to maximize my return using a limited amount of information.
So, over the course of all of this, several questions I have are:
1. What are some other filters you use? I won't usually have domain expertise to make a judgement call on material, so as above, I usually resort to asking questions about the authors (who I've never heard of) themselves.
2. I've listed *red* flags, but what are some *green* flags you might suggest?
3. (If applicable) How can experts help "moderate" the field and how could one request feedback on a particular work (aside from posting in this sub, I guess)? There's been recent talk on the alphaXiv project from Stanford and how it could be disastrous if not done properly. However, maybe something a little softer is needed. arXiv sometimes uses a referral system, but they still get some "disproving FLT using only undergraduate math" garbage, so there's definitely room for moderation. Say, instead of immediately nuking certain papers from the site, leave a big warning banner linking to a separate section of the site which lists specific complaints from *carefully certified* mathematicians.
4. I mentioned journal and conference rankings, but these are generally a crapshoot, so are there any particular metrics you recommend for my 2nd "soft filter" bullet point?
5. What are some niche (maybe even initially crank-seeming) works that you have gone through that are rigorous and educational? Doesn't necessarily need to be in an aforementioned area.
Feel free to add anything I might have neglected (including any criticism of what I've said above)!
#science
source
**Disclaimer:** I am not trying to put down any groups of people; there have been many threads on here and StackExchange which discuss the reasons for bad mathematical work and its consequences. I would venture to say that nearly student, especially in a discipline like math which focuses on adding layers of abstraction https://en.wikipedia.org/wiki/Fundamental_theorem_of_software_engineering#:~:text=The%20fundamental%20theorem%20of%20software,an%20extra%20level%20of%20indirection.%22, oes through a brief crank phase (consider the Dunning-Kruger effect), but the problem does not stop at writing easy-to-debunk research papers.
Now for the post:
Whenever I feel a bit suffocated by work in arithmetic/algebraic geometry, I try to "take a breather" by exposing myself to different (e.g. model theory, metric number theory), sometimes niche (e.g. hypercomplex analysis, differential algebra), concepts in math. Problem is, whenever you stray too far from the more modern, serious research areas (say, arithmetic geometry, PDEs, homotopy theory, etc.) and into more "enticing" stuff (say, fuzzy logic, anything "quantum", etc.), you are met with a lot of cranks. While it's easy to spot someone trying to do angle trisection or prove GRH with "10th grade math", it's not so easy to judge a book written by an associate professor at some obscure university working in a niche research area.
I've personally interacted with some serious cranks through a complex systems/AI lab I used to be a part of. While I'm perfectly capable of spotting if something is "wrong", I had wasted a ton of time listening to some hand-wavy, superficial babble. As such, I am hesitant to read many books/papers as 1. it is difficult to judge something you are currently learning (something, something, Dunning-Kruger...) and 2. it is risky to spend days reading through something and trying to make a sound judgement after the fact.
Let's say I'm looking at a book. Here are some (soft) filters I like to go through:
* *What are the authors' educational backgrounds?* Even if they have a PhD, but don't work in a related area, I'm cautious.
* *What is the authors' publication history?* It's an easy instant disqualification for even a single viXra paper, but it's tough to sort through conference and journal rankings, especially in hyped topics like AI. I also get cautious when it's tough to find any of their publications except for two papers on academia.edu r the like, or they don't have a website.
* *What does the book read like when I skim through it?* In the preface, do they claim to solve an unsolved problem over the course of the next god-knows-how-many pages? Do they (rigorously) prove things?
* *What about the publisher?* Cambridge University Press has solid work, but Elsevier does little quality control.
* *Using Scott Aaronson's list https://scottaaronson.blog/?p=304 r John Baez's (humorous) index https://math.ucr.edu/home/baez/crackpot.html.* airly self-explanatory.
For example, yesterday I was looking into Giardina's "Many-Sorted Algebras for Deep Learning and Quantum Technology". He's got a PhD, seems to be a former researcher in a "Bell Labs" type of environment, but is talking about obscure algebraic logic in the context of two hyped areas. The book also does not have proofs, just a bunch of "examples". Ultimately, I'm not going to chance it (unless one of you has some informed insight). **Another disclaimer:** I am not trying to put down Giardina; I'm just trying to illustrate how I have a certain amount of chips to gamble (say, one chip = 1 hour) and I want to maximize my return using a limited amount of information.
So, over the course of all of this, several questions I have are:
1. What are some other filters you use? I won't usually have domain expertise to make a judgement call on material, so as above, I usually resort to asking questions about the authors (who I've never heard of) themselves.
2. I've listed *red* flags, but what are some *green* flags you might suggest?
3. (If applicable) How can experts help "moderate" the field and how could one request feedback on a particular work (aside from posting in this sub, I guess)? There's been recent talk on the alphaXiv project from Stanford and how it could be disastrous if not done properly. However, maybe something a little softer is needed. arXiv sometimes uses a referral system, but they still get some "disproving FLT using only undergraduate math" garbage, so there's definitely room for moderation. Say, instead of immediately nuking certain papers from the site, leave a big warning banner linking to a separate section of the site which lists specific complaints from *carefully certified* mathematicians.
4. I mentioned journal and conference rankings, but these are generally a crapshoot, so are there any particular metrics you recommend for my 2nd "soft filter" bullet point?
5. What are some niche (maybe even initially crank-seeming) works that you have gone through that are rigorous and educational? Doesn't necessarily need to be in an aforementioned area.
Feel free to add anything I might have neglected (including any criticism of what I've said above)!
#science
source
| 25,000 mlx | |
| 27 | |
| 0 |