Systematic exploration of design space: votes, discussions/answers etc

TL;DR: A crowdsourced repository of long-tail knowledge can be built in a huge variety of ways. It may be beneficial to explore each dimension insofar it may help achieve the goal(s) of the site.

Because of my operations research background, I’ve been taught to avoid offering solutions without first considering the design space. Each solution (in our case, a site/community design) is a single point in a multidimensional space. Given an objective function and constraints, we can search for the combination of features most likely to maximize the objective function while meeting the constraints. Even if we cannot be sure of the shape of the objective function, we can at least grasp the intuition behind preferring one alternative over another.

Here are my thoughts on the dimensions and the scales involved:

  • Types of first posts: concrete questions / recommendation solicitations / stories / links / rants / homework tasks / do-my-job-for-me / spam / submarine ads
  • Types of subsequent posts allowed: long answers / speculative answers / link-only answers / comments / low-effort comments & banter / spam / submarine ads
  • Reputation/karma: scalar-valued / vector-valued (e.g. gold badge privileges on SO)
  • Reputation display: ostentatious / user profile only / hidden
  • Moderation transparency: full / partial / none
  • Use of AI for moderation: spam detection / flamewar detection / trolling detection / rudeness detection / post tagging / voting ring detection
  • Vote explanation: reason required (SE close votes, downvotes) / no reason required
  • Reputation economy: inflationary / fixed supply
  • Interaction with real life economy: votes bought or sold for real money / bounties bought or sold for real money (reddit gold) / ‘black market’ for posts (HN) / not allowed
  • Attention management: vote-dependent / votes + time / votes + time + secret sauce (HNQ!) / random bumping (Community user) / deliberate bumping (mods/users on an edit rampage)
  • Community size: <=Dunbar’s number / >Dunbar’s number
  • Editing privileges: original poster only / OP + mods / OP + mods + higher privilege users / everybody
  • Necroposting restrictions: late replies allowed at any time / late replies not allowed
  • Images in posts: allowed / limited / not allowed
  • Duplicates: actively hunted down / linked / no restrictions on rehashing ‘old’ topics
  • Information retention policy: post modification history can be altered / posts can be purged from the database / post modifications leave trails in the database / posts are stored on the blockchain

…could go on like that for quite some time (I’m sure I’ve left out many many more dimensions), but I hope you get the drift.

An important caveat: in the short run, it is the expert community that makes even a bad design viable, but in the long run, bad design ideas force subject matter experts out and ruin the community.

(Note there’s no mention of malicious management that can ruin ANY site).

What I expect from this topic: realistically, nothing. I’d be delighted if even one person could gain more clarity from it, or added the things I missed. However, I’m just an anonymous lurker with no desire to share my SE name (quite possibly, a company shill/concern troll, since y’all have no means to verify my identity). What’s more, I know this topic is not on the critical path to the MVP.


This will be helpful for us to answer, particularly in figuring out what we’re trying to build. Based on what I’ve seen in the various MVP threads around and in Discord, this is what we’re aiming for:

  • Types of first posts: Questions. This may be expanded later on, but we are designing for questions; if that can be adapted after community feedback, great, but that’s not what we’re designing for.
  • Types of subsequent posts allowed: Answers, of any length. Secondarily, “feedback” - the equivalent of comments, but yet to be fully hammered out.
  • Reputation/karma: Scalar and vector values. There is reasonable consensus that we should have some form of overarching karma value, as with reputation on SE. There is also reasonable consensus (same thread, and this other one) that privileges should be tied to related activities rather than to the central number.
  • Reputation display: Will be decided as part of specification.
  • Reputation economy: Inflationary.
  • Moderation transparency: High. We’re here because we’re not satisfied with how opaque SE is being.
  • Use of AI for moderation: Will likely be required as communities grow, but is not MVP.
  • Interaction with RL economy: Likely to be disallowed. Hasn’t been discussed, but I strongly suspect folks will feel it undermines the quality control aspect of voting. We’re still not a social network.
  • Attention management: Voting + time, at least. Voting represents quality control, which is good signal; time decay is almost inevitable as more posts are created. There may be other methods introduced later, but they are probably not MVP.
  • Community size: Depends where you look. Overall, greater than Dunbar’s number. The overall community will fractalize and create many smaller communities around a central thing.
  • Editing privileges: At least OP + mods + privileged users. We’re building a knowledge repository; wide editing is required. Extending this to everybody (a la suggested edits) has been considered, but is not MVP.
  • Necroposting restrictions: Allowed. See “knowledge repository” - knowledge doesn’t have a timeline.
  • Images in posts: Allowed. No restrictions on new users have yet been proposed, but may be necessary after MVP.
  • Duplicates: Linked, to start with. May transition towards hunting with introduction of review after MVP.
  • Information retention: Absolute, excepting extreme cases or legal requirements. Post history will not be modifiable by regular users; moderators may be granted the ability to redact after MVP, leaving a log of the redaction visible to moderators.

This is the start of a specification. There are, as you say, more dimensions and features to spec, but this is a good high-level overview.


Uh, no, quite the contrary, there seems to be a consensus that there will not be a single number.

While we can’t enforce it, it should be strongly discouraged. I’m not even a fan of SE’s bounties.

@gilles My reading of that topic gives me the impression that most people think there should be some sort of reputation. Not necessarily the only number that’s assigned/displayed/used, but that there should be at least one publicly visible “experience”-type number, as with reputation.


This is a pretty good list. I think the only thing I’d add is a certification-based reputation system, which is where users can “certify” that another user has a certain level of expertise. Design space there includes whether certifications are public or secret and whether the system used is provably sockpuppet-proof.

1 Like