Hacker News new | ask | show | jobs
by Zarel 3145 days ago
...I'm quite happy that Google stops my world view from expanding to include Chicago restaurants when I'm looking for food in Seattle.

I'm also happy it doesn't include information about swifts the animal when I search for Swift the programming language.

Even when it's about political stuff - I'm rather happy I don't get anti-vax information when I'm searching for health information. I'm also happy Google doesn't show me clickbait because it knows I'm not interested.

The idea of a filter bubble is definitely worth addressing, but I see no reason why getting more information music notes when I'm trying to search for information about C# is remotely beneficial. I can use Startpage or Incognito if I want to avoid filter bubbling, but for the vast majority of searches I do, relevance to me is useful.

But it's not best addressed through fear-mongering. Don't call it a child's imitation just because you don't understand how it could be useful.

5 comments

> Even when it's about political stuff - I'm rather happy I don't get anti-vax information when I'm searching for health information.

The tradeoff being that it won't give the anti-vaxxers information that refutes their claims. Is that _really_ what you want?

Ironically, this sort of pigeon-holing is mostly a problem with treating query text literally (or syntactically.) Language is full of shibboleths, and people indicate their worldview by the words they choose to use. The more naively a search engine treats the query, the more likely they are to find pages that only agree with them.
I honestly doubt the use of certain words are going to limit your search results more than the way google is 'personalizing' search results by looking at your history/interests and basing it's results around that, do you have any sources for your claim?
If you know the language to use, you can get into that world yourself. For instance [did dinosaurs exist]. Only creationists talk about dinosaurs that way, and so you only get creationist content returned. People with a conventional view of natural history never put "dinosaur" and "exist" together.

Contrast that with [did dinosaurs live] which has a mix of creationist and normal results, despite meaning the same thing.

Maybe. The naive approach definitely has a strong filtering effect, but there's more than one non-naive approach, though. Some of those are narrowing, and others are widening. Google seems to prefer the narrowing ones.
> Even when it's about political stuff - I'm rather happy I don't get anti-vax information when I'm searching for health information. I'm also happy Google doesn't show me clickbait because it knows I'm not interested.

But this is part of the problem, isn't it? You might not get the anti-vax information, but others will see that instead of REAL information.

And, be honest with yourself. Some of what you do want to see would be considered clickbait by others, it's just that Google already knows you well enough to show you things that fit your own world view.

The vast majority of Google results I see are very poor quality articles optimized for Google. Wikipedia is almost always the best option. Either the algorithm they are using seems to be optimizing for the wrong thing, or marketers have gotten really good at playing the game.
I spent a few years in the marketing world, doing copywriting optimized for search engines among other things. Good marketers are like professional gamers in that every time Google "changes the meta" they're uncannily quick to find new ways to game the system.
But personalization can also work against you. If for instance Google wants to favour one political candidate over another they could "personalize" your search results by boosting positive sentiment articles and filter out those with negative sentiment towards that candidate. And since Google search is still viewed as fairly objective it could have massive impact, and since it's personalized it's very hard to detect.

https://www.wired.com/2015/08/googles-search-algorithm-steal...

http://www.bbc.co.uk/news/election-2017-39938599

off-topic ... or is it?

> Even when it's about political stuff - I'm rather happy I don't get anti-vax information when I'm searching for health information.

Health information should not be considered political stuff. Unfortunately, it is ... at least in some countries.

It's interesting how this entire thread was spawned and derailed by the original essay being a bit too imprecise.

You all are taking this one argument out of context. He wasn't talking about restaurants. But everyone here is.

Obviously restaurants should be personalized. But no one was saying they shouldn't.

You have to take the argument in the essay and try to think of the most persuasive possible interpretation. Is that "If I search for something in Seattle, it would be stupid to return results in Chicago"? Probably not.

My webdev friend was excited that her personal site was being returned in her search results whenever someone searched a project she had worked on. I suggested that google was serving her personalized results. She searched in incognito. Her site was still being returned. Pretty good, right? She was getting exposure.

When we used a VPN, she was not in the results. Google knew that our searches came from our IP address, and that searches from our IP should include her site, since she mostly visited her site from our IP. Or something along those lines. Either way, it was a misleading worldview.

I'm going to be harsh for a second, but I mean this lovingly: Stop being naive. It's important for us to be skeptical of Google. They're the thousand-pound gorilla, and the moment they do more than wink and nod at their "Don't be evil" philosophy then we should start getting scared.

As far as I can tell, your essential complaint is that Google is sometimes providing results that are too relevant. You're OK with Google doing some amount of personalisation, but there's a line that they shouldn't cross.

How do you distinguish between "good" and "bad" personalisation at scale? How do their algorithms know what should and shouldn't be personalised? Do humans even agree on where to draw the line?

Google process literally trillions of searches per year, with each search taking a few milliseconds. You're asking them to make a complex tradeoff between providing completely irrelevant results for some queries and excessively personalised results for others. I don't disagree that they could probably do a better job of making that tradeoff, but I don't think that they're being malicious or negligent either. I think that they're making perfectly reasonable engineering decisions given the constraints of scale.

I don't think I am being naive. I don't blindly trust Google. I think that there are many important questions to be asked about how major internet companies collect, store and process our personal data. I think that America urgently needs to pass legislation equivalent to our General Data Protection Regulation. I think that there are significant concerns about the quality of information that people see online, but I think that publishers play a far greater role than Google in this respect.

https://en.wikipedia.org/wiki/General_Data_Protection_Regula...

There's definitely a scale I think — showing me local pizza places is great. What about guessing my budget & ruling out places it thinks I won't book? That's getting into a grey area & personally I'd find that uncomfortable/undesirable.

What about applying it to non-transactional stuff like news? Should it personalise results based on what it thinks I like?

Again with subjective/objective distinction — I know my friend will tailor movie recommendations to me based on mutual interests / past discussions.

I expect though that a library catalogue would point me to the same info on global warming no matter who I am.

The trouble at the moment is that Google is conflating these two types of interaction & the user doesn't know which of their queries are personalised (and to what extent / based on what criteria).

> How do their algorithms know what should and shouldn't be personalised? Do humans even agree on where to draw the line?

In my opinion, this is the crux of it — are we happy with algorithms filling this blank unfettered, based on their own learning. If not, it's something that we have to discuss and agree on, and then enforce / bring visibility to.

At present, Google aren't negligent (legally anyway, as we haven't set any bar) and may not be acting maliciously. But if we think change is necessary (at least for visibility of what's happening under the hood), we need to ask the questions around these services to drive that change

What bothers me is the attempt at omniscience. Give me an option! Let me have a checkbox for "tailor results to my location". Heck, turn it on by default. Just give me the damn choice.
Sometimes you just want the top results by PageRank, like in 1998 when Google was miraculous. I don't know how to get that anymore. I want information from out in the world - I generally don't want the contents of my own head reflected back at me.
Search, then click on the "tools" button and toggle to "Verbatim." It's a little more than just unpersonalized -- it also gets rid of things like synonym expansion, stemming, and allowing the search results to elide particular parts of your query. But it definitely gives you that 1998 feeling. :)

https://searchengineland.com/responding-to-complaints-google...

> As far as I can tell, your essential complaint is that Google is sometimes providing results that are too relevant.

My takeaway is that the whole thread of the first comment should be collapsed, which is usual for HN. Important topics get ignored or derailed.

pg once joked that Google's philosophy on search is the same as Scientology: What's true is true for you.

In our case, I just think it's interesting that my webdev friend was trying to ascertain truth about the world -- "is she associated with the project she worked on?" -- and the answer came up "yes!" for her even though it was "Nope" for the rest of the world.

Sure, it's an interesting question of which tradeoffs they should make. But as users, it's not really our responsibility to be concerned about that. All we know is that Google is acting a bit strangely.

To be clear, if Google stayed how it currently is, I'd have no problem at all. I'm just worried that google can go from strange to malicious at the flip of a switch. It's unsettling that they're the only realistic option. DDG has been picking up steam, but hopefully they'll do more than nip at Google's heels.

> To be clear, if Google stayed how it currently is, I'd have no problem at all. I'm just worried that google can go from strange to malicious at the flip of a switch.

What evidence do you even have that they didn't go malicious? And how could you gather such evidence if you don't have any?

>Obviously restaurants should be personalized. But no one was saying they shouldn't.

>>I'd much rather type "takeaway pizza chicago" than have google mind-read where I want the pizza delivered.

Some people are saying that restaurants shouldn't be personalized.

This was actually a problem for me recently. I just wanted to know the options for restaurants in an area of London. I didn't want results specific to me, I wanted all of the restaurants. I ended up not using Google since it just wouldn't give me the simple information I was looking for. Googles becoming less and less useful as time goes on, which is a real shame as they were excellent 10 years ago.