I made the Rat Purity Test: a rationalist-themed parody of the classic Rice Purity Test. It contains a list of "sins" (to use the terminology of the original test), and users can check off which ones they've done to receive a score (100-sins
).
After running it for a few hours, I started collected data on what people submitted. While I unfortunately missed the peak responses/time, I still managed to catch the stragglers (1.5MB of them). Ostensibly this is a violation of something, but I've established a line of legal defense:
The data is hosted here:
Here's a very quick and dirty first glance at the data.
It contains a total of 1719 submissions, with scores distributed as such:
Filtering out the responses with scores of {0, 1, 99, 100}, the remaining 1536 responses form the following cleaner curve:
First, let's take a look at the top responses. Everyone knows "counterfactual" – understandable, given that my audience is a bunch of nerds. These first dozen sins basically constitute the rationalist stereotype.
A secret belief in eugenics ranks in at 14th place, and based on a tweet I made, people seemed really surprised by this. Personally, it's not too surprising – most taboo ideas are far more popular than they appear, for obvious reasons. And the knee-jerk reaction to eugenics is far less common among rationalists, meaning they're far more willing to identify a belief they already hold as such.
Here's the tail end. I'm skeptical of that transition rate between 1st and 8th Jhana, and curious as to what the actual numbers look like.
Here's the tail end if we leave in the outlier responses:
Despite it getting 58 checks, I only know one person who's "Snorted a line of ketamine off a hardcover copy of "Probability Theory: The Logic of Science" by E. T. Jaynes". This gives us our Lizardman Constant: basically, ~3% of people are going to reply with bullshit, and any minority at that scale shouldn't really be considered.
I also appreciated the two joke responses – I could have just made the client submit the indicies of responses instead of the text itself, but that wouldn't have left room for fun stuff like this. And, with regards to "Intruded on the privacy": hey, at least I'm honest about it and you're getting a blog post. The original Rice Purity Test collects data, but you didn't know that. And they definitely aren't making those results public.
Next, we can look at the correlations between sins.
Quite a few of the most correlated responses are the more lizardman-esque questions, which makes sense. You usually either commit to the bit, or don't.
Lastly, before you explore the data for yourself, here's a few other strong correlations:
Attended a ratcamp (Atlas, ESPR, CFAR, or similar) - Been in a Hamming circle: 0.525
Been in a cuddle-puddle - Tried polyamory: 0.506
Met hit-rationalist henry - Been on a Jhourney retreat: 0.498
Experimented with meth as a nootropic - Met hit-rationalist henry: 0.386
Asked someone about your p(doom) - Made a decision under the assumption of short ASI timelines: 0.497
Asked someone about your p(doom) - Explained AI x-risks to family members: 0.495
Attended a ratcamp (Atlas, ESPR, CFAR, or similar) - Received money downstream of SBF: 0.451
Transitioned - Experimented with meth as a nootropic: 0.199
Transitioned - Tried acid "for the VOI": 0.210