Slashdot Mirror


Google Unsure About Letting Users Vote On Search

narramissic writes "Google began running a live test last year that lets people rank and remove search engine results and comment on them. Testers were presented with different variations of the experiment, which the company first publicly detailed about two weeks ago in an official blog posting. For example, in one version of the test, people can only remove results, while in another they can append comments that only they can see, said Google software engineer Matt Cutts. But while implementing these features permanently would be a major step for Google in giving more participation to its users, the company remains undecided. 'It's a really fun experiment. I can't say for sure whether it will go live for everybody because we're always running a ton of experiments. Only some of those, the ones that are being very successful, are launched live for everybody,' said Cutts. In the meantime, Google is collecting data that offers some interesting search quality insights."

5 of 145 comments (clear)

  1. Great. by gamanimatron · · Score: 5, Interesting

    So with this, I could get even more spam alongside my search results. I've got the feeling that "Ext3nd your pinis at foobar.com" would be a pretty ubiquitous comment.

    --
    cogito ergo dubito
  2. Re:Please please pretty please? by topham · · Score: 4, Interesting

    You realize the link-farmers would figure out how to use this to their advantage, right?

  3. Filtering my own results by Aerynvala · · Score: 4, Interesting

    would be fine. But I really don't care to see everyone else's search choices. At most I would tolerate a Relevant/Non-Relevant sort of system. But even that would require oversight. I think it would just be too much overhead for Google.

    --
    http://transformativeworks.org/
    1. Re:Filtering my own results by Firehed · · Score: 4, Interesting

      Well god knows they have enough data on the entire world to implement ranking in a way that's weighted towards users similar to you (whether they've got people with the data mining skills to produce that kind of thing is another story - they're brilliant, but that's some damn tricky work). If was a simple vote-up/vote-down system, they could just use their existing organic results and maybe tint the background either slightly red or green for results that users tend to find less or more helpful (maybe using time-on-site data from their pool of analytics would be better, as pre-click votes are worthless and most people wouldn't go back to rank them after the fact).

      If they can't automate it, they won't do it. IIRC there was some post a while back about them tweaking The Algorithm something like 3000 times a year, but they never blacklist sites or rankings by hand. These days it's probably as much for DMCA protection as anything else, but introducing a human element (that exists only within Google) is a bad idea for bias alone, never mind the actual labor overhead.

      I'd say that it remains an interesting exercise, but should probably stay as such. I don't think all of the data mining in the world could successfully counter the collective stupidity of the human race.

      --
      How are sites slashdotted when nobody reads TFAs?
  4. Re:If you let people vote and comment on searches by moteyalpha · · Score: 3, Interesting

    It might be better to have an algorithm that could be incorporated into the search. I could write a simple xml file that was like a grep template that took the attributes of the information and its content and applied it before I even get my search results. If they are doing the search anyway I could upload a parametric that was a list of things I don't want or prefer in the result pages. The search algorithms themselves could be searched for one that suited a person's interests. If they like fluff and magic then filter it that way. I am sure that most people would not like the choices that /.ers would make as we would have search results that contained hexadecimal, binary or octal humor ( 101 ). Things like word length and word variability could be a possible discriminating factor. I like reading about "hadrons", but other people may prefer something that causes "hadrons" s/dr/rd/.