5 ms·
> "unbiased" doesn't mean "unfiltered". For example, I don't want spam or malware in my search results. That seems like a cop-out to me. They're not filtering
by gruez 5y ago
> "unbiased" doesn't mean "unfiltered". For example, I don't want spam or malware in my search results.
That seems like a cop-out to me. They're not filtering "spam or malware", they're filtering "Russian disinformation". What's next? Maybe OAN is "right-wing disinformation" and we should "filter" that too? What about fox news? How far can you use that excuse to say you're not biased?
- verve_rat 5y ago> "Russian disinformation" Which is spam.
- remless 5y agoIf I search cure for cancer the pro bleach crowed will be annoyed that bleach doesn't comes up before chemotherapy. I'm assuming you haven't actually looked at the disinformation that Russian propagandists are putting out. They aren't just saying "Russia is justified in ensuring the independence of areas with people who consider themselves traditionally Soviets" this is "Ukraine is bombing itself to make Russia look bad". If you're arguing search engines should bias for what you want to see then it shouldn't be ranking russian propaganda highly for western audiences, if you are arguing search engines should bias for facts then by definition it shouldn't be ranking russian propaganda highly.
- kube-system 5y agoYou have rediscovered why it is difficult to make a search engine. This is a problem that applies to all search results, not just politically charged content: human judgement is always required to write a search algorithm.
- gruez 5y agoSo where does that leave us? Everything is biased? What does that mean when applied to duckduckgo's tweet from 2019? It's just a hollow slogan? I'm not sure that changes the conclusion much. The first impression I got from reading the comment was that the tweet from 2019 was a hollow PR message that they don't stand behind.
- kube-system 5y ago> So where does that leave us? Everything is biased? Yes, humans have biases. There's not a way around it. > What does that mean when applied to duckduckgo's tweet from 2019? Yes, that tweet is obviously a vague marketing message, but if you read the entire quote in context, it was in reference to filter-bubble biases. It wasn't a claim that DDG employees are somehow superhuman creatures immune to human biases. Obviously that's a ridiculous interpretation. Here's the source, in context: https://spreadprivacy.com/why-use-duckduckgo-instead-of-google/ https://spreadprivacy.com/why-use-duckduckgo-instead-of-goog...
- yifanl 5y ago> So where does that leave us? Everything is biased? Search results are created by humans and intended to be consumed by humans, so yes. (And just to head it off, a web crawler is ultimately just an abstraction for the humans who will later consume the search results, it just happens to have particularly fast and strict browsing habits) Now that said, accepting that "(un)biased" is an extremely broad term, I'd very easily believe that the intent of the tweet was to point at some specific type of bias that DDG (at the time) intended to avoid.
- paulcole 5y agoYou should check out Algorithms of Oppression, a very good book about the exact topic you’re asking questions about. Short answer yes, if humans are involved in making something it’s going to be subjected to the biases of those humans. https://nyupress.org/9781479837243/algorithms-of-oppression/ https://nyupress.org/9781479837243/algorithms-of-oppression/
- verisimi 5y agoNo - they are explicitly adding bias here. They are coding it into their algorithm. Frankly, I'd be surprised if there weren't implicit biases already there. But this is something new. Its overt, explicit.
- kube-system 5y agoThere is nothing about this that is close to new. Everyone who runs a real search engine on the public internet has a need to explicitly code in edge-cases from time to time.
- gibrown 5y ago> they are explicitly adding bias here. All search engines are explicitly biased. That is the point, they generate a ranking of results. Heck even how you tokenize text is an explicit bias of what you match against.
- paulmd 5y agoI mean, do you feel that Fox and OAN deserve the same "reputation score" as Nature? If so, why? For as much whining here about the importance of free and open discussion, it's unfortunately verboten to say it, but fortunately people will usually bring it up of their own volition. It is like the "why are all the highly educated people so liberal!?!?" thing, deep down you know the answer, and it's not actually "because of liberal indoctrination!", it's that truth itself has a bias, and so "unbiased" results that merely prefer truthiness can have a bias themselves. The fix for that isn't to present both as equally reputable (or worse, for the answer to be "unknowable" simply because "two parties disagree"), it's for the affected party to stop pushing disinformation and falsehoods. Not all sources are created equal. Not even all news sources are created equal. And if you agree to that, then we agree on the merits, and it's just the outcomes that you disagree with. (And lest you disagree - Fox News themselves went to court to argue that their programming is not "news media" and is not intended to be a source of truth, in response to slander/libel suits for saying and writing things that are factually untrue to service their political agenda. So even Fox News agrees they should have a lesser "reputation score" than even other news media.) https://law.justia.com/cases/federal/district-courts/new-york/nysdce/1:2019cv11161/527808/39/ https://law.justia.com/cases/federal/district-courts/new-yor...
- native_samples 5y agoI think you're under-estimating the quantity of articles Nature and associated journals publish that aren't scientific, or are outright wrong, just because they flatter the preconceptions of their readers. There's even a cute saying in science, "Just because it's published in Nature, doesn't mean it's wrong". Example: Nature published Flaxman2020, a modelling paper by the epidemiology team at ICL which claimed lockdowns saved 3.1 million lives. People who actually read the article realized that it was a massive exercise in circular logic - the model assumed that any reductions in case numbers could only be due to government intervention. It was then presented as proof to the media that lockdowns saved lives. It had many other problems - in fact the authors themselves in the paper itself that it was illustrative only and that in reality, the effectiveness of NPIs would be lower. The circular nature of the argument was pointed out immediately but Nature only published a response letter 8 months after publication, and never retracted it. As for your legal point, that doesn't mean much. Lawyers try any argument that might work. Facebook have argued in front of the court that their fact checking isn't actually fact checking at all, that it's in reality merely corporate political opinion, and that no reasonable person could possibly construe otherwise!