Rex Kerr @ichoran@qoto.org

@emilymbender
How does one know whether the humanities are less gate-keepy? In some sense they have to be comparable, because the ratio of students to faculty is similar.

Wouldn't this mean that the gate-keeping is social rather than more obviously skill-based, and hidden, rather than open?

@JadeintheNorth @emilymbender

Oh goodness, no, Critical Theory is a minefield of unintentionally disguised psychological tricks that let you mislead yourself while feeling virtuous and wise.

You can't even get through Horkheimer, who arguably was the most science-friendly of the bunch, without stumbling over numerous instances of things like tying your self-worth to the validity of ideas, declaring that your project cannot be evaluated until it succeeds, and that since history is non-repeating you can't make comparisons.

Critical Theory is best used like nutmeg: a little extra flavor to liven things up. Don't make a meal out of it; you'll get poisoned.

@kristinmbranson @albertcardona That's along the lines of what I was thinking too, except that I think poster is probably very important since I would not want to see Persuasively Structured Nonsense, Episode 9185, but I might want to see Resources to Concisely Rebut Persuasively Structured Nonsense (Episodes 8510-9510) which would probably overlap substantially textually with the nonsense itself.

The idea is that some X I always want to see, some Y I never want to see, and some Z I only want to see if it's done really well. For instance: everything on whole brain C. elegans imaging, nothing on roller derby, and only really thoughtful state-of-Twitter-meltdown news or opinion.

Compared to solving a massively difficult text comprehension problem, solving an implicit reputation problem seems appealing. Of course, you then have additional risks of cliques and echo chambers as you trend towards only listening to the "right people", but there are various ways that could be addressed either algorithmically or by using synthetic training data (e.g. right person, wrong content). But maybe I'm wrong and the text itself has sufficiently accessible cues. I guess it makes a difference whether we assume access to something of the level of complexity of GPT-3. (I haven't had a chance to play with content learning myself, so my intuitions aren't very well-formed.)

@kristinmbranson Yes. Boosting comes to mind.

@kristinmbranson

I just had an idea--what if people could train their own algorithms, and there was an algorithm-federating algorithm? Then you could subscribe not to any individual's algorithm (including your own) but also to community wisdom for any community you care to define. (Or maybe a limited predefined set if the federating algorithm required its own training.)

I guess, as usual, a major part of algorithm-training would be defining the training data. Sharing either a training set or a classifier or both could be helpful.

With a little more intricacy, a system could be devised that would appeal to ML geeks and nobody else :)

@freemo I'm happy to lend some eyes and thoughts, though I am too busy to be on ongoing participant. But I can try for a focused brief effort, maybe aimed at a decent first draft.

Hopefully this could be of use. I have a good deal of experience with online spaces (being old enough to have been involved in many), and I have seen a variety of attempts at moderation and such in a variety of venues and seen problems of various sorts in action. Furthermore, I'm reasonably well-versed in the contemporary problem of echo chambers, amplification of misinformation, and the propensity to anti-social content, am familiar with the standard philosophical arguments in favor of free speech (including necessary preconditions and assumptions), and have at least a superficial understanding of how human psychology differs in online vs in-person interactions.

So although I'm not an expert in the area, for an amateur, I think I'm decently qualified to have some hope of making a worthwhile contribution despite a limited engagement.

@Romaq @ejg @freemo

Although you intended this as a commentary on this particular proposal, these are good points to consider for *any* editing effort. Working in the open is laudable, but so too is presenting work of sufficient thoughtfulness and quality to be worth serious attention.

For instance, I think this is a fantastic set of questions that every author of a scientific paper should review and ponder while preparing their manuscript. (Maybe only the first two sentences of #3.)

@kristinmbranson

Although I am *very* tempted by the idea of training my own algorithm (oddly, I had never considered that before!), I do not think that this can solve the fundamental echo-chamber problem that causes social media to amplify tribalism rather than being a site of public discourse where issues are debated and resolved.

So I think a societally valuable social media system will, to some extent, have to force people to eat their vegetables, kind of like Wikipedia does: You get citations with your claims, like it or not. It's good for you.

(But it does make sense for specialist social media to be a somewhat isolated chamber, because most people probably get tired of hearing yet another thread about spike detection or clustering of scRNAseq data.)

@kristinmbranson @teixi

Ugh, I'm really sorry you've had to go through this, Kristin. It's terrible, and the "solutions" are often a different kind of terrible when implemented naively online.

I don't think we (= humanity) know yet how to reliably create a community that is welcoming and provides psychological safety for atypical ideas yet doesn't have a substantial risk for harassment. The same freedom that protects one from being censored can be wielded as a weapon to harass. It's kind of like working on advanced technology but wanting it to have no military applications whatsoever; it seems almost paradoxical.

I don't think Mastodon's "no algorithm" model is promises a good solution, either. The reason is that online discourse, without the inputs we get in person, trends towards hostility (as compared to face-to-face interactions) even in the absence of an outrage-selection algorithm. So my hunch is it needs to be algorithm-heavy, with an algorithm that favors not the outrageous and false but the measured and evidence-backed. Wikipedia is an existence proof of this working (culturally, mostly, in their case). Setting moderation as an (abusable) floor below which one can't go doesn't seem nearly as good as naturally rewarding best behavior with increased exposure.

What one can, I think, do reasonably well with Mastodon is have a curated community (maybe only implicitly curated by inviting the right people) within which people are motivated to be pro-social, and outside of which...well...there are dragons and walled kingdoms and so on, and to a large extent that just has to be accepted for what it is. Drawbridge up or down, depending on one's proclivities, but the outside world is going to do its own thing.

@albertcardona I guess I could go through a representative sample of recent fMRI papers and check. Knowing whether a process is RAM-intensive or uses shaders does give you new understanding of the function of your computational device, even if it's not an algorithm; there I think the track record is not bad (especially if you ignore the discussion section). But this is just a general impression; to get at anything like "what percentage" we'd have to actually pick out papers and score them, I think.

@neurolili @albertcardona - Good points Lilli, but I do want to point out that the poor temporal resolution of old calcium indicators was one of the primary reasons why Janelia put so much effort into calcium indicators to get them to be fast, and also that even the slow ones at least have single-cell resolution. And speed is part of why GCaMP6 was so popular.

So I don't think that argument gets very far. (I didn't trust the slow calcium indicator signals all that much either.)

@albertcardona I pretty much agree with you, but I think it depends also on what you're looking for.

To use a programming/computer hardware analogy, if you want to figure out an algorithm, fMRI is useless. If you want to know whether your computation is hitting RAM hard or requires a lot of floating point operations or is engaging the graphics card a lot (and whether shaders are used), it's pretty handy.

Rather than fMRI being a case of looking for the wallet where the light is, it's more like it's showing you where to put the light to look for wallets (if it's done well).

However, the number of overinterpreted fMRI studies does rather disappoint me. The inference "we didn't see a fMRI signal in this area so it's not important in this task" seems to happen far too often.

Then again, this is one of the biggest mistakes of scientists in every area: assuming that absence of evidence is evidence of absence when the tools and/or statistics were not specifically being deployed to address that exact question (i.e. putting bounds on how absent something is).

As a rough heuristic, I find it depressingly accurate to assume that a sentence starting with "This isn't significant, so" is one that will end with an unjustified conclusion.