Joel Spolsky compares the evolution and propagation of content rules on Stack Overflow and Wikipedia in a talk he titled How to Have Nice Things.
YOUTUBE bvEAuSHJOBU Streamed live on Nov 6, 2015.
It is easy to have high-quality small communities. Bigger ones fall apart. Why? See Own Worst Enemy
A community must discover what is on-topic vs. off-topic.
Wikipedia is a tertiary source, not a repository of world knowledge or true facts, it is merely a summary of what "authorities" consider to be true.
Stack overflow is more about the artifact, the answer, than the conversation. No shopping lists, no opinions. No overly specific. No debugging. No jokes. Not about community.
Poor terminology drove away good people who might have been able to contribute one day. For example, when they changed "closed" to "on hold", meaning your question can't be answered yet, question revisions jumped from 5% to 10%.
closed ⇒ on hold too localized ⇒ off topic not constructive ⇒ primarily opinion-based not a real question ⇒ unclear what you're asking
Great community-generated sites eventually discover non-obvious rules, crucial rules, but rules that surprise most people. Thus, constant explanation is necessary.
Both stack overflow and english wikipedia have about 30,000 five-edit-per-month contributors, but that number is growing for stack overflow, possibly due to making rules clear.
Stack overflow moderation scheme: community team 15 people, moderators 800 people elected for 150 sites. Deal with fraud and abuse, every psychological disorder, proportional to incidence in the population. White hat vs. black hat trolls.
Joel remembers his usenet experience with comp-lang-c. September ⇒ dumb questions ⇒ no-archive ⇒ faq ⇒ obscure content only ⇒ diminish community. In the end they have only those who know and like the rules.
Contributing doesn't have to be easy, but obstacles must be aligned with necessary rules. It is a mistake to think visual editor will find better editors, just more editors.
Stack overflow will suspend users for one day to one year, which works surprisingly well.
Meta community are those who are so enamored with stack overflow that they want to do more than answer questions, participate at a higher level, ultimately progress to little known chat room.
A diverse community needs diverse individuals representing as themselves, say, female, makeing other females feel like they are in the right place.
Stack overflow will delete micro-aggressions, no matter how well intentioned. Its narrow focus makes this easy. Possible too in Wikipedia. Redit and 4chan not so.
.
I came to this video by way of conversation in the wikipedia research list where the topic is often editor retention. archive 
I wrote to that list:
In the followup Q&A around minute 50 Spolsky described his subject as long-tail in contrast to notable content on Wikipedia. Is this a distinction often made in user-generated content research? Can we, for example, characterize the rules for acceptable material as favoring a long-tail or a fat-front? I ask because my current work with federated wiki is largely about allowing diverse rules for acceptability.
See Elinor Ostrom who with colleagues developed a set of design practices for commons governing rules.
See Analytics of Empathy where data analysis limits abuse in online games. flp
See Plural Identity for possible solutions to the psychological realities of protracted conflict.