The COVID-19 Forecasting Project at the University of Oxford is making advanced pandemic simulations of 150+ countries available to the public, and also offer pro-bono forecasting services to decision-makers.
Beeminder's an evidence-based willpower augmention tool that collects quantifiable data about your life, then helps you organize it into commitment mechanisms so you can keep resolutions. They've also got a blog about what they're doing here
Jane Street is a quantitative trading firm with a focus on technology and collaborative problem solving. We're always hiring talented programmers, traders, and researchers and have internships and fulltime positions in New York, London, and Hong Kong. No background in finance required.
Seattle Anxiety Specialists are a therapy practice helping people overcome anxiety and related mental health issues (eg GAD, OCD, PTSD) through evidence based interventions and self-exploration. Check out their free anti-anxiety guide here
.
AISafety.com hosts a Skype reading group Wednesdays at 19:45 UTC, reading new and old articles on different aspects of AI Safety. We start with a presentation of a summary of the article, and then discuss in a friendly atmosphere.
Giving What We Can is a charitable movement promoting giving some of your money to the developing world or other worthy causes. If you're interested in this, consider taking their Pledge as a formal and public declaration of intent.
80,000 Hours researches different problems and professions to help you figure out how to do as much good as possible. Their free career guide show you how to choose a career that's fulfilling and maximises your contribution to solving the world's most pressing problems.
Dr. Laura Baur is a psychiatrist with interests in literature review, reproductive psychiatry, and relational psychotherapy; see her website for more. Note that due to conflict of interest she doesn't treat people in the NYC rationalist social scene.
B4X is a free and open source developer tool that allows users to write apps for Android, iOS, and more.
Altruisto is a browser extension so that when you shop online, a portion of the money you pay goes to effective charities (no extra cost to you). Just install an extension and when you buy something, people in poverty will get medicines, bed nets, or financial aid.
Substack is a blogging site that helps writers earn money and readers discover articles they'll like.
MealSquares is a "nutritionally complete" food that contains a balanced diet worth of nutrients in a few tasty easily measurable units. Think Soylent, except zero preparation, made with natural ingredients, and looks/tastes a lot like an ordinary scone.
Metaculus is a platform for generating crowd-sourced predictions about the future, especially science and technology. If you're interested in testing yourself and contributing to their project, check out their questions page
Support Slate Star Codex on Patreon. I have a day job and SSC gets free hosting, so don't feel pressured to contribute. But extra cash helps pay for contest prizes, meetup expenses, and me spending extra time blogging instead of working.
My older brother died when I was a year old, so I grew up as the oldest, though not the first born. There was no way to make that clear.
I expect he’s only interested in the ratio between people’s answers to the killings questions, not their absolute values.
Several points of confusion:
“How many of your parents are professors?” I answered 0 because neither of them were technically full professors, but both had academic careers at a university with job responsibilities very similar to that of a professor, and I feel like my answer may not have been in line with the spirit of the question.
Whether your SAT score reported was the score achieved the first time you took the SAT — I took the SAT in 7th grade as part of a talented and gifted talent search program. It was unclear whether this kind of thing was supposed to “count” for the purposes of the question about whether the SAT score reported in the survey was from the first time you took the SAT. (I marked no since technically it wasn’t)
Did your parent’s job titles have “professor” somewhere in the title? Or were their job titles “Lecturer” or the equivalent?
This is actually a good point, as senior level university administrators would also grant the basic parental benefits of a professor, while an adjunct lecturer may not.
I am slightly puzzled by this, because I think near all academic jobs at university level would imply similar parental benefits (your parents are near certainly well-read, intellectual, their social circles will reflect that, and almost certainly better experts in one academic subject than any of your school teachers until college-level, probably in several).
Maybe it is because of differences in terminology, and you are implying differences in stability of income? Here, “lecturer” is a relatively steady for-life full-time job (barring radical government cuts or similar, once per decade) for a person with PhD and willingness to stay at university. The downside is that it is very much a teaching position, one does not have time for serious research.
Edit. And on contrary, I would believe that university administrators would be no different than any other administrator. No necessarily other benefits that having generic middle-class parents bring in general.
* In the “Overall artifacts” question, in the reference you have artifacts on the left, but one who reads the question closely will see that the answers have more artifacting placed to the right. I very nearly just answered based on proximity to the images, ignoring the labels; a less pedantic reader than I might have answered it either direction without noticing there was another. I worry that that makes this question’s results suspect.
* You promised me a place in the survey to complain about your definitions of siblings, but no such place appeared. Luckily I had no need to complain, but for people who do, it feels like you’re hoping they’ll forget when they don’t see a field. (Or you just forgot, yourself, that you’d promised.)
* You told me to skip to “Part 20”, but for some reason the next part is “PART TWENTY”. As a visual skimmer, I missed it the first time, looking for a “Part 20” to go with the instruction and the “Part 19b”.
* Similarly PART TWENTY told me to skip to Section 21, but the next section is PART TWENTY-ONE. After the previous I was on the watch for it, but here somehow _neither_ part of the name matches, even ignoring case. Yes, the names may be semantically similar but it’s impossible to scan _visually_ for.
* The survey reloaded itself and lost my entered data way too many times. I got through on attempt 4. (Android Firefox) Could it be made to save previous pages’ answers (and be cut into shorter pages), or even better, made to keep fields’ values even when the tab is unloaded from memory? It’s possible the fix would need advanced use of Google Sheets or else a different survey host. (Survey monkey or something?)
* But a solution that makes it robust to reloads might also make it robust to navigating out, which helps non-mobile users too. (And it’d make it less important to finish in one sitting.)
* When there’s information to be gathered, like SAT scores, income, etc., maybe list it all at the start of the page? That way people can get it to hand (or plan to skip it) all at once if they want, rather than pop in and out.
* Not counting data gathering, the survey still took me like an hour and a half to finish attempt 4, even knowing most of the first page from the previous attempts. So your “30 minutes” can get stuffed, or else be qualified that that’s a time for someone who skips most of it. I’m up way past my bedtime and I’m mentally throwing vegetables at you about it.
I realized after I hit “submit” that I should have switched to “don’t release anonymized data publicly”. Anything I can do about that?
Probably not. If you look at the instructions (https://slatestarcodex.com/2019/12/30/please-take-the-2020-ssc-survey/) Scott says that he is not willing to change answers to any questions, explicitly including the privacy question. Likely the best you can do is a polite request asking that in the future maybe the privacy question could be at the end, so that one knows which information is included.
If I might ask, though, what’s your specific fear about releasing the anonymized data?
The risk of being identified is significantly higher with the additional information from the second page.
BTW, am I the only one who, upon being told to imagine a “red star”, was thinking this instead of a starfish-like symbol?
I was thinking of something like that; I think I imagined a red-outlined bright-white-centered circular blob of light basically.
You are not the only one 🙂
Yeah I did the same thing hah
I picked up this survey in the middle of reading The Three Body Problem (which for those unfamiliar is a sci-fi novel largely concerned with astronomy) so no you are not the only one. I was very surprised to see the images on Twitter!
When I took the survey, I think I assumed that “have you completed your formal education?” means “are you studying right now?” and not, for example, “are you planning to ever get more formal education in the future?” but in hindsight I’m not sure what was the intent?
I took it was supposed be inclusive of both full-time students and also “dropouts” who entered workforce before finishing their formal studies.
I did not take your survey because on several issues I know an awful lot about your questions were obviously designed to elicit an answer that does not and cannot conform to the best and truest answer.
That being said, you are doing some brilliant work here.
Seriously, though, you need to find some people who you consider to be as intelligent as yourself and just talk to them for hours and hours and hours on the subjects you find fascinating.
I would love it if you would have conversations like that and report on them.
Please do not post this comment, I am just trying to be anonymously helpful.
Here’s my yearly exhortation to make the racial category question “choose all that apply” rather than “choose one”.
Scott, unfortunately all of your sleep questions are basically useless because none of them provides a clear answer about being woken up by children or partners.
For example, I wake up at roughly the same time every morning WITHOUT an alarm clock, because my son starts yelling at about seven.
There was no answer to come close to capturing that. Any conclusions you draw will be very very misleading because you aren’t account for likely a majority of the population.
Your questions basically treated waking up as something between yourself, your alarm clock, and your internal clock. That’s a poor reflection of a large population out there. And you left no way to let those folks distinguish themselves
I think a pet or child that more or less predictably wakes you up at the same time every day can be considered to fall under the same category as “alarm clock.”
But what if it doesn’t predictably wake you up at the same time? My point is that, Scott’s questions about sleep habits seemed very clearly to come from a frame of reference of a single person.
And look, I appreciate that this is a hobby endeavor for Scott, but if he wants to draw meaningful insights some of the survey questions really suffer a need for a few rounds of cognitive interviews to make sure he’s asking what he thinks he is.
Then your kid is a shitty alarm clock, get him repaired.
“Atheist but spiritual” is a category I have only ever seen on dating website questionnaires. What does it mean? “Atheist but I am not Richard Dawkins”?
Also surely it is quite easy to make the html open links in new tabs automatically?
That option is for people who don’t believe in gods but believe in some “higher forces” that have some meaningful impact on our lives (for example, make astrology work). That includes a wide range of beliefs from some folk mythology to New Age.
I don’t think that’s necessarily what it means. A lot of people who practice meditation would describe themselves as “spiritual” but don’t believe in any “higher forces”. My impression is that spirituality to these people has something to do with the nature of consciousness rather than external forces or supernatural phenomena. I suspect that people are going to interpret that option in a lot of different ways.
AFAICT most people who practice meditation do believe that it grants them borderline supernatural powers, so the “spiritual” label still fits.
Hm, I don’t think I’ve ever met a meditator who believes that, not that I’ve met all that many meditators in the first place. Different Worlds, I suppose.
You don’t believe in God, but you think there’s something more going on than just the base physics of unconscious fundamental particles. A non-materialist would fit in this category.
Most atheists are physicalists. I believe that position is completely falsified by the avalaible evidence (qualia), but current culture categorizes my views on this something akin to woo, so I self-report as spiritual to make communication work, even though in a better world my position would be the mainstream.
My problem is that “atheist but spiritual” sounds like “there is no god but when I practise mindfulness my friends the crystals speak to me of a Higher Power” whereas “not spiritual” sounds like “there is no god, and what do you mean you cry at the end of King Lear? It’s all just physics.” Neither represents my position.
The “when do you go to sleep” questions are a poor fit for people who work the graveyard shift (I’m a night auditor, and before that I was a night shift security guard, so I’m sensitive to this).
Loved the “Prisoner’s Dilemma against your clone” question; a practical chance to use superrational decision theories like TDT.
Yes, I thought that. I was initially inclined to cooperate on both versions, but eventually I picked defect against random and cooperate against clone, just to emphasise the difference.
Also, does night auditor count as blue collar or white collar? I picked white, because I wear dress clothes at work (dress shirt, dress pants, dress shoes, tie, and briefcase) but I get paid by the hour, not a salary.
Were you able to read/listen to podcasts/surf the internet/etc. when you worked as a security guard? Are you able to do those things now?
We were not supposed to use our phones, although a lot of people took them out for a few minutes at a time when they were out of view of the security cameras. I don’t have a smartphone, and a Kindle would have been rather conspicuous, but what I did have was a very small MP3 player. So I would put on one headphone in the ear facing away from the security camera and listen to podcasts and audiobooks in between rounds (I would make a round every 15 to 30 minutes depending on how sleepy I was feeling). I don’t know if that was allowed or not, because I wasn’t stupid enough to ask and destroy my plausible deniability; that way if they caught me and it wasn’t allowed, I could honestly plead ignorance, and if it was allowed, then there was no problem. Hence “it is better to ask for forgiveness than to ask for permission”.
Oh, yes. I would say I do at most 4 hours of actual work on a given night (that’s between changing shifts, preparing breakfast, processing reports, selling the last few rooms, checking-in late arrivals, answering the phone, and dealing with whatever problems in-house guests are having). The rest of the time can be spent browsing the internet, watching Netflix, reading books, doing homework, playing video games, or whatever else you want. It’s a great job if you can deal with the 11-7 schedule; a perfect fit for autistic NEET types. See what /r9k/ has to say about it.
Hey there, fellow night auditor here, in a high-end fancy place!
I also listen to podcasts every moment I’m working but not interacting with guests. I run a corded earbud from my left ear once around the arms of my glasses (to hold it in place), under my hair, down the back of my sweater, to the iPod Nano Gen 3 in the right-hand pocket of my dress.
I use such an ancient device because I can pause and play by touching the click wheel through the fabric of my dress. To anyone watching, it looks like I’m just brushing my finger against my dress. No need to ever look at the screen more often than the 1-3 hours a file might last.
I, too, only do four hours of work a night at the desk (often even less than that), and the rest is utterly unsupervised downtime. I hope on the hotel’s wifi and read, watch movies, do internet based chores, etc. It’s absolutely glorious.
If you are naturally nocturnal, and disciplined enough to follow every other rule and do everything absolutely right so managers never, ever ask questions, it’s one of the easiest, lowest-stress, (arguably over-paid) jobs out there.
Or people with irregular circadian rhythm, whatever the reason. (My answers are very untrustworthy guesstimate about median.) Improvement for the next survey: ask “when did you go to sleep (a) previous Monday, (b) previous Tuesday, … etc …, (g) previous Sunday” and “when did you wake up on …” and “were your bedtimes during the previous seven days typical or atypical for you”.
edit. Even more improvement, add tick boxes “which of these days you did/ did not were expected to go to work or study” (examples of “not expected to go to work: weekend, no assigned work shift, vacation, or similar)
> (option for non-Americans who want an option)
Haha, I believe I smell lessons learnt from previous surveys. For me it was a good thing it was there, because I accidentally clicked on something, and once you do that, there’s no way to un-click.
I’m thinking of using Scott’s Survey data to train a model, optimizing for different values in it (e.g. highest possible income, highest score on Openness, etc.) and seeing what survey responses would that hypothetical person have. Has it been done?
Wait, what? I’m not sure what the goal of this would be.
Fun mostly but also just giving you a profile for those highly X hypothetical people.
Would the hypothetical super-rich SSC reader be a republican or democrat? Would they cooperate or defect?
Would the hypothetical super-Neurotic SSC reader be taking certain medications? exercising? etc. etc.
Oh, I see. Fill in part of the survey and let the model fill in the rest. That’s a cute idea, actually.
I was a bit confused about what you were suggesting as input and output for the model, but it makes sense now.
More specifically the Input would be all questions except for the one I’m interested in and the output would be the question im interested in (e.g. income).
Then I’ll do gradient descent over the input (which can be just the median user’s answers) given a really high fixed value for the output, optimize over it until the model has changed the answers to whatever it thinks would give that high value for my output.
What do you count as “training a model”?. I’ve done linear regression before. AFAIK, no one has done “real” machine learning (e.g. a random forest or neural network), but I’m not really sure why that would be interesting since we usually want to be able to interpret the models afterwards, which “real” machine learning generally makes difficult.
In this framing the problem which the NN (or like GP + NN) will solve*is* interpreting in the first place – which features make X (e.g. income) have a higher value.
Random Forests are actually quite straightforward to interpret and generally do pretty well on tabular data (usually on par with deep NNs, even at shorter training times).
It’s a neat idea, but I doubt you’ll have enough data points to train anything remotely interesting.
There were > 8k survey responses last time. That’s a decent amount, though the problem has very high dimensionality.
“and don’t click on a link unless you’re sure it will open in a new tab and not throw out your responses so far.”
yeah this happened to me weeks ago when I was on the second page so I had to fill in the first page again as well, I hope the first page also got deleted otherwise it’ll be in the database twice.
I think the first page only gets submitted if you actually hit the “Submit” button at the bottom, so you’re probably alright.