The Problem with Digg

A recent up-and-comer web destination has been

Promoted as a democratic competitor to Slashdot, Digg has differentiateditself by the fact that the community (in true “Web 2.0″ form)determines the visibility of stories – as opposed tothe paid editors of Slashdot – voting stories up by”Digging” them. Those stories that are voted upon enough getboosted to the “front page” (orthe front page of one of the sections, like software), vastly increasingtheir exposure, while the stories that don’t catch on quicklydegrade to irrelevance in the Digg world.

The problem I have with Digg is the same problem I see withmany community-driven sites: It isn’t a large body ofdomain-knowledgeable, unbiased, critical evaluatorsspending needed time to evaluate the worthiness ofsubmissions, but instead is largely a bunch of fly-by visitorsthat are often going with whatever has the feel of anappropriate story, going with whatever the herd thinks (insituations like that, where voting is free, people often expresstheir feelings superficially, feeling obligated tocontribute to the democracy but preferring to do so with minimaleffort). Groupthink in action.This same principal applies at Flickr as well, where the mostinterestingpictures are the pictures from people heavily involved in thecommunity, and that have already appeared as interesting (e.g. mostof the viewers aren’t talent scouts out watching the raw talent -instead they’re watching the big leagues and commenting on who’sgood). It’s not that those pictures aren’t interesting – thereare often fantastic pictures in that set – but rather it’sjust that they are selected from a very small set of the availableFlickr photos.

This same problem appears in Slashdot moderation, where it’squite easy to game the system. You can ensure that your commentwill get moderated up to Score:5 by following acouple of simple posting rules.

  • Post early – right after the story appears
  • Post near the top – if you didn’t get in early, then post as areply to one of the early stories. Most moderators are lazy andjust want to find something to blow the points on quickly
  • Longer posts usually get moderated up (even if it’s redundantlysaying the same thing)
  • Say something cluefully anti-Microsoft (not just an empty M$rant – the crowd is too cynical about that), pro-Firefox,pro-Linux, or contrarian pro-Microsoft (anti-Linux will never getyou points, but if you’re pro-Microsoft in a contrarian way -the “I’ll probably get moderated down for this…” – you’ll get thevote of the contrarian crowd).
  • Make a post soliciting an insightful comment. e.g. “Couldsomeone explain to me, an unpolished rube, why…?”
  • It’s critical to get the first mod up, by following the aboverules, because many mods look at the already culled posts, scanningthe Score:2 and Score:3 posts for something to moderate up.

This is possible because many of the people with mod pointssimply want to exhaust them as quickly as possible – thisis actually encouraged by Slashdot – so they moderate up whateverhas the feel of a prototypical Score:5 comment. Even if itisn’t based on the referenced article, makes a nonsensical point,or is a brutally obvious karmawhore, it will be Score:5 in no time.

Of course Digg might not be ideal, and even though the highestranked stories are examples of Groupthink herding in action, it’snot like the editors at Slashdot do a better job. Most of themdon’t even read their own site (evident by the incrediblenumber of duplicate stories. I visit Slashdot once or twice a day,yet even I manage to immediately spot the duplicates that peoplewho get paid to do this fulltime amazingly can’t), and many of thestories they pick are dated and of marginal interest to most of thecommunity.

My ideal situation would be a composite site – I’d love to seethe best of both worlds, where it isn’t a select group of apatheticemployees doing the selection, nor is it a random group of herdingindividuals engaged in groupthink, but rather it’s myown anointed group of selectors. This could be accomplished ina Digg type setting by allowing me to agree or disagree with theselection of a story. As it learns my opinion of stories, complexassociative data mining could be used to build a set of storiesagreed to by a set of individuals with similar selectioncriteria. I don’t want to have to manually select “friends” orbuild any web of trusts, but rather it should be easy to implementbased upon my tastes correlated with the tastes of others.

In a simplified form, what I’m talking about is implemented by,albeit in web toolbar form.