Hacker News new | ask | show | jobs
by sli 2504 days ago
On top of all that, Google's snippets aren't curated and therefore, aren't always correct. They can be (and almost certainly are) gamed. Users that don't click through open themselves up to carrying on being misinformed.
7 comments

I've found them to be incorrect so often on things when I would click through to the actual page or find a better link. I don't trust just the blurb for any answers any more.
I don't trust just the blurb for any answers any more.

I don't, either.

A site I used to own had a discussion forum on it. It contained a message along the lines of "Real Estate Agent X is a great guy. Real Estate Agent Y is a complete sleazebag."

The blurb that Google displayed for it was "Real Estate Agent X... is a sleazebag." And that was the first result for anyone who searched for that agent's name.

As you can imagine, I received many angry e-mails, phone calls, and legal threats. No, you can't explain to angry people that it's "just" an algorithm that told the world that they're a sleazebag.

I ended up editing the post so that Google would display a different version after its next scrape.

I think there's more to this... Google use lots of fancy Natural Language Processing stuff to extract that data, and unless the wording was very tortuous, I doubt it could make such a big mistake by chance.
They can get it painfully wrong last time. I came down with something like optic neuritis a few years ago. It's often one of the first signs of MS in many folk. When I googled something like "MS life expectancy", the blurb said something like "3-7 years" -- with subtext indicating it's 3-7 years LESS than average rather than "you're kicking it in 3 years".

Turns out I didn't have optic neuritis.

They suck. And something about the way they are presented seems to make people believe them.

I think it gives that one-shot answer to questions people have, even when the real answer is nuanced and multi-faceted.

I think they’re believable because google started by providing things that weren’t wrong. If you search for a time zone google shows it in your local time, if you search for currency conversion google does that. All those things that it’s done for ages, which were things that were also typically correct.

Then the snippets show up, and they are presented in a similarly trust worthy fashion. But the snippets are really just the really just the result of which ever site has the best SEO, and that’s often a really worthless metric these days. The time zone and currency stuff is easy, because it’s math, but opinions aren’t. The thing is though that even if google didn’t have the snippets, those sites that gets snippets would still be the top results that we clicked, and we’d still get the wrong information. That would probably be better, because it might be easier to spot obvious bad sources, but I still think there is just a fundamental flaw in how SEO professionals have learned to game the google bot to bring the world useless information.

I mean, part of it is certainly on google. No one in their right mind wants to comply with Google’s ranking terms, unless you make money from google searches. Which means a lot of useful personal blogs have dropped off the face of the internet, unless you’re really lucky to see them linked on a place like HN.

I wish libraries would band together and make a privacy focused and curated search engine, because librarians are actually kind of good at finding you the correct information.

It sucks. Sometimes the bold text is the exact opposite of the answer to the query I search for. It’s very misleading unless you click through and read the full context.
Yeah. I personally like the feature, in theory, as an end user, but the signal:noise ratio for it has not been great for me.
This is especially true where the answer is time-bound, which happens a lot in technical topics. Many times the snippet is for an earlier version of the language (but still with a high PageRank), or the Operating System (especially Android settings), and the most annoying at all: an ancient answer in an undated blog post.
Google is good at dating undated content. They keep track of the first time they've ever seen a bit of text, and assume it was composed then, even if it later gets copied to other sites.
For a recent search "report amex card stolen", google showed a phone number for a scam who asked for a social security number as soon as you called.
The websites that the results aren't curated either. Clicking through to the site could provide the same incorrect information.
The point is that Google frequently adds another level of incorrectness, that may not be identifiable without checking the source. This is pretty common on Wikipedia, and when people link to things in discussion forums, as well.

And anything Google does, is done at vast scale, which makes me, at least, think it might be substantially affecting society.

But that's the responsibility of that website. Of course it's bad if Google lists a site with wrong information as the first hit, but I think it's worse when Google blindly copies that false info and lists it as their own zero-click result. By doing that, Google itself takes responsibility for the information.

Although sometimes the site is actually correct and Google still gets it wrong by copying the info incorrectly or losing some context or qualifiers.

I loved zero-click results back when DucfDuckGo first introduced them, but I'm less enthusiastic about Google's implementation of them.

sometimes the blurb just has an answer to a different question. Websites are curated, unless its spam.
> Websites are curated, unless its spam.

Yes, but even when they are curated the curators are usually unreliable and sometimes malicious.

snippets are just a reflection of that. how is google faring better in that respect?
Those are sites google chooses are correct.
For example, this WaPo story, about YouTube videos for some medical queries that go to videos featuring quack remedies and anti-vaxxer misinformation.

https://www.washingtonpost.com/lifestyle/style/they-turn-to-...

> On top of all that, Google's snippets aren't curated and therefore, aren't always correct.

The “therefore” is misplaced; curated snippets aren't always correct, either.

People on the web take the risk of being misinformed, clicking or not.