Unpinning makes sense for some questions and not others
Stack-Exchange has three types of questions on the network:
- Ones with an objective answer that is correct, and a different answer may become more correct over time
- Ones with a combination of objective and subjective components to answers
- Ones with almost entirely (but not completely) subjective components to answers
All three of these types are answerable on stack because all three types allow for expert opinion to avoid wrong answers. It is possible for there not to be a correct answer to a problem while simultaneously having very definite wrong answers. For example, "How would I solve world hunger in the world I am creating for my novel? (parameters are x)" would be a fine question for Worldbuilding.SE, it doesn't necessarily have one true "right" answer, it does, however have at least one very wrong answer "have [insert disaster] wipe out everyone on your world". Sure it solves the problem of world hunger, but leaves the querent without a world to build in. Expert opinion allows answers to avoid the wrong, and provide potential good solutions to the problem.
Whether or not a site should use the unpinning feature or not depends on what mix of these three types of questions there are on that site. Sites that have more of types 2 & 3 should retain pinning, sites that have more of type 1 should remove it.
What does pinning actually achieve?
Pinning an answer gives the querent an active opinion on which answer solved their problem. Unpinning changes that opinion to being more passive (readers have to scroll past the first answer to get to an unpinned accepted answer, which is the use case this change is attempting to solve).
This decision seems to be based on the assumption that the querent is the least expert person in the discussion. Making this assumption in all cases is a dangerous road to go down.
The experiment this change was based on, was run on the site which has a plethora of objectively answered questions (Stack Overflow). In that situation the querent will be the least expert person in the room.
Someone, on the other hand, asking "How should I handle situation x which is causing issues for reasons x & y" is not a situation where the querent is the least expert person in the discussion, it's a situation where they may be the most expert person on what constitutes the best answer.
The moment you bring a subjective topic into a question is the same moment the fundamental assumption that the SO experiment was investigating shifts dramatically. For example, the querent cannot possibly put down all of the history and nuance of the interpersonal relationships (because space), however they are uniquely qualified to spot what is the best answer for their situation once presented (in that situation).
Example Site Classifications
For sites where there is an objective answer to a question (Physics, Mathematics, Stack-Overflow, etc.), the querent's opinion doesn't matter as much on what is the best answer, as the best answer for the site is "what is the highest quality and most correct answer", which is a criteria that can change over time.
For sites where instead there is frequently going to be a subjective answer (RPG, Parenting, Politics, etc.). Answers to these types of questions will (commonly) have two components, objective facts (e.g. game rules) and subjective opinion (how those game rules should be interpreted and applied). The querent's opinion on what solved their problem should hold more weight for these questions as it is a prominent and visible signal of what solved the subjective part of their problem.
There is a third set of questions which are (almost) totally subjective, those dealing with interpersonal relations. Given the focus of some sites (like RPG.SE), we allow a subset of those (e.g. managing unruly players in my game, or x player's behaviour is unfun). For these types of questions the most correct answer is the one which solved the querents question the best, and this can only be indicated by pinning their accepted answer.
So which of the two approaches should we take?
Given the different mix of questions that sites can have, it should be an opt-in situation for a site to de-pin accepted answers, as opposed to an opt-out of the de-pinning. This would produce the least harm.
As a counter point, I think a better approach would be to try and classify most sites in the network based on the balance of these three buckets of questions on their site. Then provide an initial rollout of de-pinning to those sites with primarily questions of type 1, with an opt-out for these sites. It should then be opt-in for all other sites.
When a site is in beta, it should be a configurable criteria by the mods of that site, so they can experiment and figure out which approach works best for their fledgling site. The final decision should then be made once the site graduates out of beta.