6 ms·

How about addressing the fact that sexism related stories get flagged off his very own website with surprising speed (either manually, or by his own automated a

How about addressing the fact that sexism related stories get flagged off his very own website with surprising speed (either manually, or by his own automated algorithms)? Are we really going to blame algorithms now for promoting prejudice? The reason sexism-related stories fall down is not because of lack of interest or due to prejudice, but quite the contrary: too much interest, causing an imbalance in the upvote-to-comment ratio, and a subsequent sinking. This is HN. Come on, we should know better.

cjbprime 13y ago
> Are we really going to blame algorithms now for promoting prejudice? Your implicit claim is that a human can't write an algorithm that promotes prejudice. Are you interested in defending that claim? It seems very obviously false.
- vezzy-fnord 13y ago
  I guess you could. Although I'm not sure how exactly an algorithm can be inherently prejudiced, that sounds rather postmodern. You can make one that is inefficient or apply one that isn't suitable to a given data set, and end up with skewed results. But an algorithm promoting prejudice by itself? I don't know. I always thought mathematics was neutral. Either way, it absolutely does not apply here. HN's algorithms sinking sexism-related stories is a side effect. The fact of the matter is it will sink any story that has an imbalance in the upvote-to-comment ratio, whether it's about sexism, Haskell or the Church of the SubGenius. This is the same logic as saying that encryption is evil because it can be used by terrorists to hide information. Encryption always does the same thing, it's fundamentally neutral, regardless of what's being encrypted. The same way HN's algorithms are fundamentally neutral, regardless of what happens to trigger their procedures.
  - cjbprime 13y ago
    I think you're overcomplicating it by insisting on isolating the algorithm from its owner and environment. No-one else was doing that. If you don't want to apply a label to the algorithm, how about "Foo Bar's decision to run Algorithm X against Dataset Y resulted in consequences that predictably increased prejudice, so it would have been better if he hadn't done it"? If that's okay, note that this is pretty close to what parfe said -- he's clearly upset at the fact that pg chooses to run an algorithm on Hacker News, rather than at the algorithm itself.
    - vezzy-fnord 13y ago
      I won't comment on the efficiency of the algorithm (although it is questionable). Let's assume that it's a bad design decision. Your logic behind it being prejudiced is still faulty, as it ascribes a fallacious cause. The algorithm merely shuffles page rankings based on two criteria: upvotes and comments, both of which are user-supplied. The algorithm does not support prejudice, it simply behaves in the fashion it is programmed to. You're completely ignoring all of the mundane stories that go down due to the same factors, and cherry picking the ones about sexism.
      - DanBC 13y ago
        The algorithm also uses flags. Some people flag the sexism articles once the wingnits arrive because the threads are effectively useless by then.
  - moocowduckquack 13y ago
    Although I'm not sure how exactly an algorithm can be inherently prejudiced, that sounds rather postmodern. All known halting program/input pairs are inherently a little bit prejudiced.
  - kaitai 13y ago
    Along a slightly different track than you were originally running on, there are a number of ways in which an algorithm could promote prejudice. Consider a nice little machine-learning algorithm that analyzes characteristics of successful employees at a given company in order to select new hires that share those characteristics. If, for instance, the company had a couple rabid barbecue-lovers who made sure that all lunches primarily featured delicious pork ribs, and they made fun of the vegetarians, Jews, Muslims, and people with braces who couldn't chew the meat off the ribs, and so being a rabid lover of pork was strongly correlated with success in the company and being a non-pork-lover or hampered by braces was strongly correlated with simply leaving the company at an early stage to get a job with the rival across the street that had vegetarian and shrimp satay lunch options... well, it's just the algorithm at work, my friend! Math is truth! As a mathematician, I guarantee it. Time for dinner. heh.
    - vezzy-fnord 13y ago
      Once again, this isn't the algorithm being inherently prejudiced. It does what it's meant to do, but it's being applied to a prejudiced and unreliable data set. Don't blame the algorithm for the fallacies of the people using it. The only fault in your scenario is the PEBCAK.
daleharvey 13y ago
The algorithm was written with an intent and it can be changed. Its been very clear for a long time that almost all discussion about women in tech is as good as censored on hackers, which is a shame because its giving a lot of well intentions but naive people the impression that there is no problem.
- jiggy2011 13y ago
  Those threads always turn into flamewars that spill out into other threads.
  - daleharvey 13y ago
    Whereas the latest NSA disclosure, facebook privacy setting etc etc leads to a rational though provoking discussion that stays on topic and within its own thread?
    - jiggy2011 13y ago
      I think NSA discussions now are penalized too.
    - dictum 13y ago
      If they get more comments than upvotes in a window of time (IANpg but last I heard, it was the first 30-45 minutes after submission), they'll suffer the same fate as submissions about sexism in tech.
    - frandroid 13y ago
      It's fairly hard to produce hate speech towards the NSA. It's much more likely to happen when the topic of sexism is brought about.
- vezzy-fnord 13y ago
  Unfortunately it seems that no one can actually agree on the exact nature of what the problem is, so most discussions end up being just pointless shouting matches.
chadwickthebold 13y ago
Aren't upvotes a measure of interest? I would argue that the lack of upvotes on sexism-related posts indicates a lack of interest in discussing these issues on the part of the general population of HN. What you get is a small group of vocal people eager to discuss the topic while everyone else simply avoids the post. This is similar behavior to what is seen in a flamewar, but could we agree that there might possibly be a better way to rank posts that is able to differentiate between an important discussion that not many people want to have, versus a petty internet argument?
- cjbprime 13y ago
  > Aren't upvotes a measure of interest? I would argue that the lack of upvotes on sexism-related posts indicates a lack of interest in discussing these issues on the part of the general population of HN. The sexism-related posts have a large number of upvotes, yet they are penalized anyway by the controversial algorithm under discussion.
  - DanBC 13y ago
    Or maybe those threads are flagged by many people?
    - Karunamon 13y ago
      That was my understanding as well. Every single "sexism in tech" thread on HN devolves into some combination of the following: * The Tumblr SJW PC brigade flinging poo * The people who wish the first group would FOAD, flinging poo * The people who have something interesting to say lamenting the existence of the first two groups, who wish they'd both FOAD. * The people who make meta posts about the whole thing. I.e. boring, predictable, bile-filled, and more suited to certain Reddit and Tumblr communities who shall remain nameless. Every time this comes up, especially this high up on the page, my reaction is "Oh god, not this shit again..." I would have flagged it myself, but somehow I don't think flagging a PG article is conducive to my further ability to use the feature...
- dictum 13y ago

cjbprime 13y ago

> Are we really going to blame algorithms now for promoting prejudice? Your implicit claim is that a human can't write an algorithm that promotes prejudice. Are you interested in defending that claim? It seems very obviously false.

vezzy-fnord 13y ago
I guess you could. Although I'm not sure how exactly an algorithm can be inherently prejudiced, that sounds rather postmodern. You can make one that is inefficient or apply one that isn't suitable to a given data set, and end up with skewed results. But an algorithm promoting prejudice by itself? I don't know. I always thought mathematics was neutral. Either way, it absolutely does not apply here. HN's algorithms sinking sexism-related stories is a side effect. The fact of the matter is it will sink any story that has an imbalance in the upvote-to-comment ratio, whether it's about sexism, Haskell or the Church of the SubGenius. This is the same logic as saying that encryption is evil because it can be used by terrorists to hide information. Encryption always does the same thing, it's fundamentally neutral, regardless of what's being encrypted. The same way HN's algorithms are fundamentally neutral, regardless of what happens to trigger their procedures.
- cjbprime 13y ago
  I think you're overcomplicating it by insisting on isolating the algorithm from its owner and environment. No-one else was doing that. If you don't want to apply a label to the algorithm, how about "Foo Bar's decision to run Algorithm X against Dataset Y resulted in consequences that predictably increased prejudice, so it would have been better if he hadn't done it"? If that's okay, note that this is pretty close to what parfe said -- he's clearly upset at the fact that pg chooses to run an algorithm on Hacker News, rather than at the algorithm itself.
  - vezzy-fnord 13y ago
    I won't comment on the efficiency of the algorithm (although it is questionable). Let's assume that it's a bad design decision. Your logic behind it being prejudiced is still faulty, as it ascribes a fallacious cause. The algorithm merely shuffles page rankings based on two criteria: upvotes and comments, both of which are user-supplied. The algorithm does not support prejudice, it simply behaves in the fashion it is programmed to. You're completely ignoring all of the mundane stories that go down due to the same factors, and cherry picking the ones about sexism.
    - DanBC 13y ago
      The algorithm also uses flags. Some people flag the sexism articles once the wingnits arrive because the threads are effectively useless by then.
- moocowduckquack 13y ago
  Although I'm not sure how exactly an algorithm can be inherently prejudiced, that sounds rather postmodern. All known halting program/input pairs are inherently a little bit prejudiced.
- kaitai 13y ago
  Along a slightly different track than you were originally running on, there are a number of ways in which an algorithm could promote prejudice. Consider a nice little machine-learning algorithm that analyzes characteristics of successful employees at a given company in order to select new hires that share those characteristics. If, for instance, the company had a couple rabid barbecue-lovers who made sure that all lunches primarily featured delicious pork ribs, and they made fun of the vegetarians, Jews, Muslims, and people with braces who couldn't chew the meat off the ribs, and so being a rabid lover of pork was strongly correlated with success in the company and being a non-pork-lover or hampered by braces was strongly correlated with simply leaving the company at an early stage to get a job with the rival across the street that had vegetarian and shrimp satay lunch options... well, it's just the algorithm at work, my friend! Math is truth! As a mathematician, I guarantee it. Time for dinner. heh.
  - vezzy-fnord 13y ago
    Once again, this isn't the algorithm being inherently prejudiced. It does what it's meant to do, but it's being applied to a prejudiced and unreliable data set. Don't blame the algorithm for the fallacies of the people using it. The only fault in your scenario is the PEBCAK.

daleharvey 13y ago

The algorithm was written with an intent and it can be changed. Its been very clear for a long time that almost all discussion about women in tech is as good as censored on hackers, which is a shame because its giving a lot of well intentions but naive people the impression that there is no problem.

jiggy2011 13y ago
Those threads always turn into flamewars that spill out into other threads.
- daleharvey 13y ago
  Whereas the latest NSA disclosure, facebook privacy setting etc etc leads to a rational though provoking discussion that stays on topic and within its own thread?
  - jiggy2011 13y ago
    I think NSA discussions now are penalized too.
  - dictum 13y ago
    If they get more comments than upvotes in a window of time (IANpg but last I heard, it was the first 30-45 minutes after submission), they'll suffer the same fate as submissions about sexism in tech.
  - frandroid 13y ago
    It's fairly hard to produce hate speech towards the NSA. It's much more likely to happen when the topic of sexism is brought about.
vezzy-fnord 13y ago
Unfortunately it seems that no one can actually agree on the exact nature of what the problem is, so most discussions end up being just pointless shouting matches.

chadwickthebold 13y ago

Aren't upvotes a measure of interest? I would argue that the lack of upvotes on sexism-related posts indicates a lack of interest in discussing these issues on the part of the general population of HN. What you get is a small group of vocal people eager to discuss the topic while everyone else simply avoids the post. This is similar behavior to what is seen in a flamewar, but could we agree that there might possibly be a better way to rank posts that is able to differentiate between an important discussion that not many people want to have, versus a petty internet argument?

cjbprime 13y ago
> Aren't upvotes a measure of interest? I would argue that the lack of upvotes on sexism-related posts indicates a lack of interest in discussing these issues on the part of the general population of HN. The sexism-related posts have a large number of upvotes, yet they are penalized anyway by the controversial algorithm under discussion.
- DanBC 13y ago
  Or maybe those threads are flagged by many people?
  - Karunamon 13y ago
    That was my understanding as well. Every single "sexism in tech" thread on HN devolves into some combination of the following: * The Tumblr SJW PC brigade flinging poo * The people who wish the first group would FOAD, flinging poo * The people who have something interesting to say lamenting the existence of the first two groups, who wish they'd both FOAD. * The people who make meta posts about the whole thing. I.e. boring, predictable, bile-filled, and more suited to certain Reddit and Tumblr communities who shall remain nameless. Every time this comes up, especially this high up on the page, my reaction is "Oh god, not this shit again..." I would have flagged it myself, but somehow I don't think flagging a PG article is conducive to my further ability to use the feature...
dictum 13y ago