2 ms·

Based on observed behavior, I never believed that was how it actually works there. It seems to actually just be an upvote/downvote system.

dredmorbius 3y ago
Classically, Slashdot moderation: 1. Applied only to comments. Stories are selected by (paid staff) editors. 2. Was enacted by a subset of members, who would rotate through the active moderation pool. 3. Each moderator could only moderate a small number of comments, after which they dropped out of the rotation. 4. Each comment could be moderated only within a specific range, I believe -1 to +5. Each moderator's vote moved the comment by one unit up or down. Negative-voted comments were hidden or collapsed. The second and third aspects were highly effective against moderation brigades or collaboration. It was unlikely that any one group would receive and be able to coordinate moderation activities at any one time, whether to promote or bury comments. The limits may have been too extreme, but the effect was useful. The fourth aspect meant that a given comment wouldn't converge on some ultimate value (say, as with a five-star rating system, with all the issues involved there), but rather each moderation action shifted score by a full point. The result was ... a fairly uneven ultimate rating of content. Yes, really bad stuff tended to get buried, and really good stuff ... might ... make it to the top of the page. But an awful lot of mediocre-to-good content just sort of languished in the middle. Reading large comment pages felt ... kind of bleh. (Mind: HN's comments threads can often feel the same way, though they seem somewhat less prone to this.) I've long felt that averaging weighted ratings of content, on a 3- or 5-point scale, would be better. A comment could approach but not attain a perfect score (3 or 5), and the overall moderation could be "fuzzed" by the uncertainty given the number of votes and points assigned. (One problem / annoyance of voting systems is the false certainty of perfection in rating items. Reddit adopted fuzzing to address this, though somewhat poorly in my view. Using a voted range + random fuzz factor might do better at placing items at various levels within a displayed ranking, achieving a more fair final result.) Though I've also come to feel that crowdsourced moderation has its own tremendous limitations. The assessments of a thousand laypeople really shouldn't outweigh that of a domain expert (acting in a well-justified and good-faith manner, of course). And many moderation systems really don't seem to have given deep thought to what they are trying to measure. Something I've written about at length: <https://web.archive.org/web/20200629055317/https://www.reddit.com/r/dredmorbius/comments/28jfk4/content_rating_moderation_and_ranking_systems/ https://web.archive.org/web/20200629055317/https://www.reddi...>

dredmorbius 3y ago

Classically, Slashdot moderation: 1. Applied only to comments. Stories are selected by (paid staff) editors. 2. Was enacted by a subset of members, who would rotate through the active moderation pool. 3. Each moderator could only moderate a small number of comments, after which they dropped out of the rotation. 4. Each comment could be moderated only within a specific range, I believe -1 to +5. Each moderator's vote moved the comment by one unit up or down. Negative-voted comments were hidden or collapsed. The second and third aspects were highly effective against moderation brigades or collaboration. It was unlikely that any one group would receive and be able to coordinate moderation activities at any one time, whether to promote or bury comments. The limits may have been too extreme, but the effect was useful. The fourth aspect meant that a given comment wouldn't converge on some ultimate value (say, as with a five-star rating system, with all the issues involved there), but rather each moderation action shifted score by a full point. The result was ... a fairly uneven ultimate rating of content. Yes, really bad stuff tended to get buried, and really good stuff ... might ... make it to the top of the page. But an awful lot of mediocre-to-good content just sort of languished in the middle. Reading large comment pages felt ... kind of bleh. (Mind: HN's comments threads can often feel the same way, though they seem somewhat less prone to this.) I've long felt that averaging weighted ratings of content, on a 3- or 5-point scale, would be better. A comment could approach but not attain a perfect score (3 or 5), and the overall moderation could be "fuzzed" by the uncertainty given the number of votes and points assigned. (One problem / annoyance of voting systems is the false certainty of perfection in rating items. Reddit adopted fuzzing to address this, though somewhat poorly in my view. Using a voted range + random fuzz factor might do better at placing items at various levels within a displayed ranking, achieving a more fair final result.) Though I've also come to feel that crowdsourced moderation has its own tremendous limitations. The assessments of a thousand laypeople really shouldn't outweigh that of a domain expert (acting in a well-justified and good-faith manner, of course). And many moderation systems really don't seem to have given deep thought to what they are trying to measure. Something I've written about at length: <https://web.archive.org/web/20200629055317/https://www.reddit.com/r/dredmorbius/comments/28jfk4/content_rating_moderation_and_ranking_systems/ https://web.archive.org/web/20200629055317/https://www.reddi...>