4 ms·

This will be abused by trolls more than used by the "good guys". It's just a matter of time before some trolls decide to use this feature to create more lies ab

This will be abused by trolls more than used by the "good guys". It's just a matter of time before some trolls decide to use this feature to create more lies about elections, hitler, holocaust, immigrants, and every conspiracy theory out there. You're going to search "who did 9/11" and have false websites trying to "fact check" the top pages (which at least today seem to be sane).

mitchty 9y ago
Yep, 4chan will have a field day with this. I think the goal of this is to harness the wisdom of experts, but can't help but think the end result of this will just be more noise in signal to noise search ratio. Without human input as to what is correct and real of these inputs, its just another input that can be gamed.
- username223 9y ago
  Pretty much. Google was sad when it realized that 10^x layers of "Deep Learning" couldn't detect propaganda, and it will be sad when trying to get people to do it for free doesn't work, either.
  - jkeat 9y ago
    I could see this working if sites with a good track record were trusted more when making claims. A site's trustability could be punished when one of its fact checks is flagged for review and discovered to be wrong. Will be interesting with Wikipedia in the mix. It could be the web's fact checking backbone.
    - JetSpiegel 9y ago
      > I could see this working if sites with a good track record were trusted more when making claims. Google mentions this in the docs, but if it's page ranking, that can be gamed, and if it's "pay us to be a fact checker", well...
      - jkeat 9y ago
        It could be "if your fact check gets flagged a lot we'll review it and punish your reputation if it's inaccurate." Not ideal to leave Google in control of Truth (and they might not want the burden), so it'd be interesting to offload that responsibility to Wikipedia, which seems to have a pretty good system for coming to a consensus.
notatoad 9y ago
Assuming Google trusts anything marked as a fact. Which would be a ridiculous assumption.
- tyingq 9y ago
  They seem pretty lax about "rich snippets" already. Perhaps they don't trust them, but they present them as things to be trusted.
beejiu 9y ago
It seems like you need to be manually approved by Google? (Hence the "If your organization is interested in implementing, or seeing issues with, ClaimReview, submit your contact information here".)
userbinator 9y ago
I suppose "trolls" also includes companies with an interest in pushing their own agenda, which historically have been far more influential than the more diverse and random Internet trolls.
- mtgx 9y ago
  Or say ... Russian propaganda machines. And now we've come full circle.
Laremere 9y ago
The linked page explains that fact checks aren't guaranteed to be displayed, but use an algorithm similar to page rank to choose which if any checks should be shown. Fact checking doesn't automatically mean people will believe the fact checker over the top link anyways. Exposing people to disagreement to the issue from the start may be important because once people form a belief, they're much more likely to hold that believe despite contradictory evidence.
moomin 9y ago
Indeed. Google already has a huge astroturfing problem and there is no evidence this is going to do anything to address this at all. Incidentally, those who think ML would help need to take a look at the training corpus.

mitchty 9y ago

Yep, 4chan will have a field day with this. I think the goal of this is to harness the wisdom of experts, but can't help but think the end result of this will just be more noise in signal to noise search ratio. Without human input as to what is correct and real of these inputs, its just another input that can be gamed.

username223 9y ago
Pretty much. Google was sad when it realized that 10^x layers of "Deep Learning" couldn't detect propaganda, and it will be sad when trying to get people to do it for free doesn't work, either.
- jkeat 9y ago
  I could see this working if sites with a good track record were trusted more when making claims. A site's trustability could be punished when one of its fact checks is flagged for review and discovered to be wrong. Will be interesting with Wikipedia in the mix. It could be the web's fact checking backbone.
  - JetSpiegel 9y ago
    > I could see this working if sites with a good track record were trusted more when making claims. Google mentions this in the docs, but if it's page ranking, that can be gamed, and if it's "pay us to be a fact checker", well...
    - jkeat 9y ago
      It could be "if your fact check gets flagged a lot we'll review it and punish your reputation if it's inaccurate." Not ideal to leave Google in control of Truth (and they might not want the burden), so it'd be interesting to offload that responsibility to Wikipedia, which seems to have a pretty good system for coming to a consensus.

notatoad 9y ago

Assuming Google trusts anything marked as a fact. Which would be a ridiculous assumption.

tyingq 9y ago
They seem pretty lax about "rich snippets" already. Perhaps they don't trust them, but they present them as things to be trusted.

beejiu 9y ago

It seems like you need to be manually approved by Google? (Hence the "If your organization is interested in implementing, or seeing issues with, ClaimReview, submit your contact information here".)

userbinator 9y ago

I suppose "trolls" also includes companies with an interest in pushing their own agenda, which historically have been far more influential than the more diverse and random Internet trolls.

mtgx 9y ago
Or say ... Russian propaganda machines. And now we've come full circle.

Laremere 9y ago

The linked page explains that fact checks aren't guaranteed to be displayed, but use an algorithm similar to page rank to choose which if any checks should be shown. Fact checking doesn't automatically mean people will believe the fact checker over the top link anyways. Exposing people to disagreement to the issue from the start may be important because once people form a belief, they're much more likely to hold that believe despite contradictory evidence.

moomin 9y ago

Indeed. Google already has a huge astroturfing problem and there is no evidence this is going to do anything to address this at all. Incidentally, those who think ML would help need to take a look at the training corpus.