Red and blue buttons

On 2026-04-24, Tim Urban came up with a version of Lisatomic’s son’s old question (from 2023-08-13) about the red and blue pills. The question was, basically: there’s a red and a blue pill, everyone picks exactly one, and if more than 50% choose the red pill then whoever picks the blue pill dies (if anyone). In Tim Urban’s version, it’s buttons to press instead of pills to take, and it’s everyone on earth instead of only the X/Twitter poll-takers.

I did not see it immediately because I have Tim Urban muted, but it eventually went so viral that I had to take notice of it. The question quickly became known as “the button test”. I don’t like this, because I dislike that Lisatomic’s young son’s creativity got overshadowed by Tim Urban’s ripoff. A version involving account bans was run on Glosso, of which I’m a member. (I chose blue on Glosso, for reasons which I posted to the site, which are different from my reasons regarding the original problem.)

Recapitulating my reasoning

None of my moral reasoning about the question has changed since my 2024-11-26 post about the original variant with the pills. To summarize it: If you’re good and not evil, your only concern is to be as little culpable as possible in outcomes with deaths (outcomes with no deaths can be disregarded). Deaths only happen if we have both RW (Red Win) and FRU (Failure of Red Unanimity), where RW means Reds are majority (a moral problem for Reds) and FRU means Reds are not unanimous (a moral problem for Blues). Since both are required, neither side is uniquely to blame. But each Blue only contributes to FRU in that he adds himself to the death count (culpable for 0 or 1 deaths), while Reds are each partly responsible for the whole death count via the RW (and this must work like conspiracies to commit murder, where each conspirator gets convicted of as many counts of homicide as the conspiracy as a whole committed). Since the total death count, if nonzero, is necessarily ≥1 and likely >1, the correct answer is Blue regardless, since that option is culpable for fewer deaths. (See the original post for comparison of these assumptions about culpability with alternative ones, which I found implausible.)

It is worth highlighting that my reasoning was threshold-independent. I do not choose Blue because it is likely to work for saving everyone. I do not care whether anyone is saved. I want to minimize my culpability if deaths happen, and I do not care whether they do actually happen; it is wrong to choose Red even if the threshold for Blue “winning” is 80% or 99% rather than 50%.

To emphasize this, I have sometimes compared the problem to this: “Everyone in group A will kill everyone in group B tomorrow. Which group do you want to join?” Here it is clear that no one dies if everyone joins only A or if everyone joins only B; but it is extreme to hope for either kind of unanimity, especially with large populations. In this, it is like the pills/buttons problem, but with the threshold raised to 100% (which accurately conveys my disregard for thresholds) and the action of killing made explicit (since I have indeed reasoned that the original does involve homicide by the participants). Here my assumptions about culpability can be made perfectly clear:

1. it does not seem plausible to take away culpability from participants just because none of them set up the “A will kill B” situation, and
2. it is very clear that joining B is morally better, since “it is better to suffer an injustice than to commit one”.

Reds are evil

All this time since my 2024 post, I already thought Reds were morally worse due to the reasoning indicated. (I was already pro-Blue before I made the 2024 post, but with somewhat less explicit reasoning.) But this has not been very important in my mind. Most people are pro-murder in one way or another; it is a wretched world.

Observing the conversations since Tim Urban’s viral variant has made me more alert, however, to independent reasons why Reds are morally worse, due to different kinds of dishonesty that showed up often during the Red-Blue arguments, and only ever on the Red side.

1. Analogies that introduce uncertainty that wasn’t in the original. Already on 2023-08-14, we had seen Roko’s blender variant, but these dishonest analogies, designed to instill doubt that reaching the threshold will even truly work to prevent the deaths (which doubt is not allowed by the original statement), have only become more prevalent since the Tim Urban variant (see thread on this). I have seen analogies to drinking poison so that an antidote is administered, jumping in front of train tracks to ensure the train will be stopped, etc.
2. Self-centered framings of the question. Framings such as, “Red: you live, Blue: you maybe live”. This presupposes that your own survival is somehow privileged, which is an assumption which it is always extremely wrong to even slightly consider.
3. Dismissing the entire concept of morality, and/or of valuing it more than life. It is one thing to defend that Reds are somehow less culpable than Blues for whatever reason. But something only Reds ever did, and Blues never did, was to dismiss the entire concept of morality, and mock persons who valued being moral for holding moral values, and/or for holding these values above their own survival.

For these reasons, which are only indirectly connected to their choice, I am very willing to have a lower opinion of the character of Red-choosers as a generality. Empirically, and not a priori, I have determined that there is something terribly wrong with the kind of mind that chooses Red, which makes Red-choosers much less pleasant persons to be around; as usual with empirical so-called knowledge, I have no idea why.

This was part of my reason for choosing Blue on Glosso; I don’t really want to be on a social media website where I mostly have to talk to these vicious Red-choosing people. I truly do not appreciate their existence, and I doubt anyone else does.