#ResearchQuality If you could have a mixture of experts to roast an academic paper, who would that set of experts be (eg judea pearl, philosophers of science, Cosma Shalizi, Tapio Schneider, Kaeberlein, analytical chemists, Dusan Materic, cybersec, etc)?

Not perfect (esp for the truly OOD kind of roast) but these are good starters! might be useful to see a database of reviewer comments (though reviewer comments are often biased and not perfect). Also some scientists like to get “high” off of being excessively “bad faith mean”, sometimes really unnecessarily.

it’s nice to always look at the skeptics - HOWEVER - skeptics sometimes persistently end up wrong too (gary marcus has that reputation, and it’s funny b/c his critiques don’t have terrible Brier score, it’s more that he misses the forest in some of the trees even though it’s true that other researchers miss another mycorrhizal network that HE is seeing)

there are some areas of research where reproducibility/data quality is especially poor (pharmacognosy/nutrition, early microplastics, educational psychology, social psychology, Alzheimer’s, [arguably] fMRI-based cogpsych). You might say the same for large areas of drug discovery, but they are more “sus” due to selection/filtering effects [studies done on drugs are generally way higher quality than ones on nutrition]

(a LOT of fields that have their OWN conferences dedicated to them - eg ISEV, EVs, pharmacognosy, microplastics have poor SN ratios and many poorly-controlled papers - yet most longevity researchers would still benefit way more from at least some more exposure to them!)

i know in the early days of longevity people had a lot of really strange “explain all” theories of aging that weren’t multi-modal enough, or they believed exercise slowed aging [it doesn’t, kaeberlein is known at going after them], or they didn’t sufficiently appreciate the power of CR. I know there is the Liao et al study on dif mouse strains and some naive people cite that as “against CR”, but this is more bc different mouse strains cannot be put on isocaloric diet).

also in early days there are also major inconsistencies in how people take care of animals or c elegans or w/e.

Cremieux on twitter has an interesting f-score/brier score - he’s really good at critiquing the “normies” (eg normies hating on SSRIs), but . He’s also alleged to be adjacent to HBD-people and mb insufficiently attentive to methodological issues in some of the more “ambitious” papers like those by kierkegaard, though people on twitter tend to forget this (and mb intelligence explains more than most people think it does).

um by cybersecurity i mean cybersecurity researchers. um think of andrew bunny huang?

Um also think of fields known to have researchers who know how to call out cognitive psychology papers for being “trash”/unbelievable (bc the reproducibility rate of it is way lower often due to it being complicated, but there are also so many issues with test subject internal variability and TEST CONDITIONS affecting performance [internal variabiity can explain a lot] and differences in *diagnosis*).

think of data quality people who know that a lot of EEG/neurofeedback research is very unreproducible.

and ofc think of people who legitly go after people who read too much into intelligence research (even though intelligence *still* explains more in psychology than any other singular variable - thus psych research doesn’t work unless you control for intelligence first - hence why a social psychology fMRI test I did had IQ test questions at the end for controls)… think of those who really questioned the flynn effect for once.

but those who find anomalies (from data ) [one kuhnian idea] are also very uncorrelated from the 1000s of cogpsych/psych/cancer/biology research that goes out that is of questionable quality

What is the average paper quality in these areas of chemistry: toxicology/analytical/organic/physical/atmospheric/bio/geo/astro/silicon/metal-organic/retrosynthesis/free radicals/quantum/pharmacokinetics?

what is the avg paper quality of spatial transcriptomics? there are already so many issues with some RNA transcripts not being captured, some decaying bc of thanatotranscriptomics/decay, AND NO ONE HAVING A PROPER LONGITUDINAL DATASET [bc the stress-response transcriptome is very dif from the avg transcriptome]?

[cf ron alfa]

what about quality of epigenetic methodology papers? (esp their f-ratio). esp those that use sulfite-based extraction, TIME-Seq, andrei tarkov, alex trapp papers? what about later vs earlier ones? collecting ALL the data is important because someone is going to create a better clock later