I couldn't find any consensus on this on meta for a while, and this seems to be particularly helpful when answering decision-problem challenges. Should swapping truthy and falsy be allowed by default, when answering such challenges, unless overridden?
For example, when testing primality, one could output a falsy value for primes and a truthy value for non-primes. Is this allowed by default?
Defaults apply by default (duh), but they do not override the challenge specification. In contrast, the challenge specification can override our defaults in any way it deems fit. Thus, all answers should do what the challenge asks for.
If the challenge asks for a truthy value for primes and a falsy value for non-primes, output a a truthy value for primes and a falsy value for non-primes.
However, if the challenge only asks for two consistent and distinct outputs for primes and non-primes and specifies no further restrictions, truthiness no longer matters.
Consensus exists to help us deem what we think is appropriate for all challenges, not to override what challengers say. The challengers are probably much better equipped to make specific decisions about what is best for their challenge than for us to make sweeping defaults about all challenges.
Note that for decision problems where two consistent values are requested, (which seem to be becoming more popular because of issues with truthy v. falsy), swapping values is perfectly acceptable just by the definition of the challenge.
Having to negate the truthiness of a certain value sometimes only adds unnecessary bytes to a submission. Unless explicitly overridden in the challenge body, swapping truthy / falsy values should be allowed by default.
Note to challenge writers – You may want to read this meta discussion about replacing truthy / falsy values (which is a concept which some languages don't have) with other methods, suitable for decision-problems.
