Gregory_Lewis

Researcher (on bio) at FHI

Comments

What is the increase in expected value of effective altruist Wayne Hsiung being mayor of Berkeley instead of its current incumbent?

I recall Hsiung being in favour of conducting disruptive protests against EAG 2015:

I honestly think this is an opportunity. "EAs get into fight with Elon Musk over eating animals" is a great story line that would travel well on both social and possibly mainstream media.
...

Organize a group. Come forward with an initially private demand (and threaten to escalate, maybe even with a press release). Then start a big fight if they don't comply.

Even if you lose, you still win because you'll generate massive dialogue!

It is unclear whether the motivation was more 'blackmail threats to stop them serving meat' or 'as Elon Musk will be there we can co-opt this to raise our profile'. Whether Hsiung calls himself an EA or not, he evidently missed the memo on 'eschew narrow minded obnoxious defection against others in the EA community'.

For similar reasons, it seems generally wiser for a community not to help people who previously wanted to throw it under the bus.

Use resilience, instead of imprecision, to communicate uncertainty

My reply is a mix of the considerations you anticipate. With apologies for brevity:

  • It's not clear to me whether avoiding anchoring favours (e.g.) round numbers or not. If my listener, in virtue of being human, is going to anchor on whatever number I provide them, I might as well anchor them on a number I believe to be more accurate.
  • I expect there are better forms of words for my examples which can better avoid the downsides you note (e.g. maybe saying 'roughly 12%' instead of '12%' still helps, even if you give a later articulation).
  • I'm less fussed about precision re. resilience (e.g. 'I'd typically expect drift of several percent from this with a few more hours to think about it' doesn't seem much worse than 'the standard error of this forecast is 6% versus me with 5 hours more thinking time' or similar). I'd still insist something at least pseudo-quantitative is important, as verbal riders may not put the listener in the right ballpark (e.g. does 'roughly' 10% pretty much rule out it being 30%?)
  • Similar to the 'trip to the shops' example in the OP, there's plenty of cases where precision isn't a good way to spend time and words (e.g. I could have counter-productively littered many of the sentences above with precise yet non-resilient forecasts). I'd guess there's also cases where it is better to sacrifice precision to better communicate with your listener (e.g. despite the rider on resilience you offer, they will still think '12%' is claimed to be accurate to the nearest percent, but if you say 'roughly 10%' they will better approximate what you have in mind). I still think when the stakes are sufficiently high, it is worth taking pains on this.
Use resilience, instead of imprecision, to communicate uncertainty

I had in mind the information-theoretic sense (per Nix). I agree the 'first half' is more valuable than the second half, but I think this is better parsed as diminishing marginal returns to information.

Very minor, re. child thread: You don't need to calculate numerically, as: , and . Admittedly the numbers (or maybe the remark in the OP generally) weren't chosen well, given 'number of decimal places' seems the more salient difference than the squaring (e.g. per-thousandths does not have double the information of per-cents, but 50% more)

Use resilience, instead of imprecision, to communicate uncertainty

It's fairly context dependent, but I generally remain a fan.

There's a mix of ancillary issues:

  • There could be a 'why should we care what you think?' if EA estimates diverge from consensus estimates, although I imagine folks tend to gravitate to neglected topics etc.
  • There might be less value in 'relative to self-ish' accounts of resilience: major estimates in a front facing report I'd expect to be fairly resilient, and so less "might shift significantly if we spent another hour on it".
  • Relative to some quasi-ideal seems valuable though: E.g. "Our view re. X is resilient, but we have a lot of knightian uncertainty, so we're only 60% sure we'd be within an order of magnitude of X estimated by a hypothetical expert panel/liquid prediction market/etc."
  • There might be better or worse ways to package this given people are often sceptical of any quantitative assessment of uncertainty (at least in some domains). Perhaps something like 'subjective confidence intervals' (cf.), although these aren't perfect.

But ultimately, if you want to tell someone an important number you aren't sure about, it seems worth taking pains to be precise, both on it and its uncertainty.

Evidence on good forecasting practices from the Good Judgment Project: an accompanying blog post

It is true that given the primary source (presumably this), the implication is that rounding supers to 0.1 hurt them, but 0.05 didn't:

To explore this relationship, we rounded forecasts to the nearest 0.05, 0.10, or 0.33 to see whether Brier scores became less accurate on the basis of rounded forecasts rather than unrounded forecasts. [...]
For superforecasters, rounding to the nearest 0.10 produced significantly worse Brier scores [by implication, rounding to the nearest 0.05 did not]. However, for the other two groups, rounding to the nearest 0.10 had no influence. It was not until rounding was done to the nearest 0.33 that accuracy declined.

Prolonged aside:

That said, despite the absent evidence I'm confident accuracy with superforecasters (and ~anyone else - more later, and elsewhere) does numerically drop with rounding to 0.05 (or anything else), even if has not been demonstrated to be statistically significant:

From first principles, if the estimate has signal, shaving bits of information from it by rounding should make it less accurate (and it obviously shouldn't make it more accurate, pretty reliably setting the upper bound of our uncertainty to 0).

Further, there seems very little motivation for the idea we have n discrete 'bins' of probability across the number line (often equidistant!) inside our heads, and as we become better forecasters n increases. That we have some standard error to our guesses (which ~smoothly falls with increasing skill) seems significantly more plausible. As such the 'rounding' tests should be taken as loose proxies to assess this error.

Yet if error process is this, rather than 'n real values + jitter no more than 0.025', undersampling and aliasing should introduce a further distortion. Even if you think there really are n bins someone can 'really' discriminate between, intermediate values are best seen as a form of anti-aliasing ("Think it is more likely 0.1 than 0.15, but not sure, maybe its 60/40 between them so I'll say 0.12") which rounding ablates. In other words 'accurate to the nearest 0.1' does not mean the second decimal place carries no information.

Also, if you are forecasting distributions rather than point estimates (cf. Metaculus), said forecast distributions typically imply many intermediate value forecasts.

Empirically, there's much to suggest a T2 error explanation of the lack of a 'significant' drop. As you'd expect, the size of the accuracy loss grows with both how coarsely things are rounded, and the performance of the forecaster. Even if relatively finer coarsening makes things slightly worse, we may expect to miss it. This looks better to me on priors than these trends 'hitting a wall' at a given level of granularity (so I'd guess untrained forecasters are numerically worse if rounded to 0.1, even if the worse performance means there is less signal to be lost, and in turn makes this hard to 'statistically significantly' detect).

I'd adduce other facts against too. One is simply that superforecasters are prone to not give forecasts on a 5% scale, using intermediate values instead: given their good callibration, you'd expect them to iron out this Brier-score-costly jitter (also, this would be one of the few things they are doing worse than regular forecasters). You'd also expect discretization in things like their calibration curve (e.g. events they say happen 12% of the time in fact happen 10% of time, whilst events that they say happen 13% of the time in fact happen 15% of the time), or other derived figures like ROC.

This is ironically foxy, so I wouldn't be shocked for this to be slain by the numerical data. But I'd bet at good odds (north of 3:1) things like "Typically, for 'superforecasts' of X%, these events happened more frequently than those forecast at (X-1)%, (X-2)%, etc."

EA Forum feature suggestion thread

On-site image hosting for posts/comments? This is mostly a minor QoL benefit, and maybe there would be challenges with storage. Another benefit would be that images would not vanish if their original source does.

EA Forum feature suggestion thread

Import from HTML/gdoc/word/whatever: One feature I miss from the old forum was the ability to submit HTML directly. This allowed one to write the post in google docs or similar (with tables, footnotes, sub/superscript, special characters, etc.), export it as HTML, paste into the old editor, and it was (with some tweaks) good to go.

This is how I posted my epistemic modesty piece (which has a table which survived the migration, although the footnote links no longer work). In contrast, when cross-posting it to LW2, I needed the kind help of a moderator - and even they needed to make some adjustments (e.g. 'writing out' the table).

Given such a feature was available before, hopefully it can be done again. It would be particularly valuable for the EA forum as:

  • A fair proportion of posts here are longer documents which benefit from the features available in things like word or gdocs. (But typically less mathematics than LW, so the nifty LATEX editor finds less value here than there).
  • The current editor has much less functionality than word/gdocs, and catching up 'most of the way' seems very labour intensive and could take a while.
  • Most users are more familiar with gdocs/word than editor/markdown/latex (i.e. although I can add and other special characters with the Latex editor and a some googling, I'm more familiar with doing this in gdocs - and I guess folks who have less experience with Latex or using a command line would find this difference greater).
  • Most users are probably drafting longer posts on google docs anyway.
  • Clunkily re-typesetting long documents in the forum editor manually (e.g. tables as image files) poses a barrier to entry, and so encourages linking rather than posting, with (I guess?) less engagement.

A direct 'import from gdoc/word/etc.' would be even better, but an HTML import function alone (given software which has both wordprocessing and HTML export 'sorted' are prevalent) would solve a lot of these problems at a stroke.

EA Forum feature suggestion thread

Footnote support in the 'standard' editor: For folks who aren't fluent in markdown (like me), the current process is switching the editor back and forth to 'markdown mode' to add these footnotes, which I find pretty cumbersome.[1]

[1] So much so I lazily default to doing it with plain text.

Examples of people who didn't get into EA in the past but made it after a few years

I applied for a research role at GWWC a few years ago (?2015 or so), and wasn't selected. I now do research at FHI.

In the interim I worked as a public health doctor. Although I think this helped me 'improve' in a variety of respects, 'levelling up for an EA research role' wasn't the purpose in mind: I was expecting to continue as a PH doctor rather than 'switching across' to EA research in the future; if I was offered the role at GWWC, I'm not sure whether I would have taken it.

There's a couple of points I'd want to emphasise.

1. Per Khorton, I think most of the most valuable roles (certainly in my 'field' but I suspect in many others, especially the more applied/concrete) will not be at 'avowedly EA organisations'. Thus, depending on what contributions you want to make, 'EA employment' may not be the best thing to aim for.

2. Pragmatically, 'avowedly EA organisation roles' (especially in research) tend oversubscribed and highly competitive. Thus (notwithstanding the above) this is ones primary target, it seems wise to have a career plan which does not rely on securing such a role (or at least have a backup).

3. Although there's a sense of ways one can build 'EA street cred' (or whatever), it's not clear these forms of 'EA career capital' are best even for employment at avowedly EA organisations. I'd guess my current role owes more to (e.g.) my medical and public health background than it does to my forum oeuvre (such as it is).

Why not give 90%?

Part of the story, on a consequentialising-virtue account, is typically desire for luxury is amenable to being changed in general, if not in Agape's case in particular. Thus her attitude of regret rather than shrugging her shoulders typically makes things go better, if not for her but for third parties who have a shot at improving this aspect of themselves.

I think most non-consequentialist views (including ones I'm personally sympathetic to) would fuzzily circumscribe character traits where moral blameworthiness can apply even if they are incorrigible. To pick two extremes: if Agape was born blind, and this substantially impeded her from doing as much good as she would like, the commonsense view could sympathise with her regret, but insist she really has 'nothing to be sorry about'; yet if Agape couldn't help being a vicious racist, and this substantially impeded her from helping others (say, because the beneficiaries are members of racial groups she despises), this is a character-staining fault Agape should at least feel bad about even if being otherwise is beyond her - plausibly, it would recommend her make strenuous efforts to change even if both she and others knew for sure all such attempts are futile.

Load More