Google wants to rank websites based on facts not links | Hacker News Reader

Google wants to rank websites based on facts not links

newscientist.com

66 pointsvoyweb35 comments

Some facts are easy, like "capital of albania" or "population of shenzhen" or "age of clint eastwood".

But is consensus the right answer for more interesting facts? Does the consensus give the right answer for "minimum wage effect on economy" or "do firearms make us safer" or "crime rate of immigrants"?

Will Google localize their facts? For a question like "us supporting nazis in ukraine", will Russian residents get one answer while US residents get a different answer?

The old Microsoft Encarta encyclopedia had different facts for different countries. In the US edition, Thomas Edison invented the light bulb, while the British version said it was Sir Joseph Swan.

As Bill Gates famously wrote[1], "The facts depend on where you are coming from".

Google Maps used to show different national borders depending on where the user was located[2], maybe it still does. Chinese users saw Arunachal Pradesh as part of China, while Indian users saw it as part of India.

[1] https://web.archive.org/web/20120629032909/http://www.btimes...

[2] http://qz.com/224821/see-how-borders-change-on-google-maps-d...

To make matters worse, much of Google's Knowledge Graph -- their database of facts -- is taken from Wikipedia. There are many problems with this approach. Wikipedia facts are often unreliable. They are easy to vandalize. Many pages are unchanged for years because editor numbers are rapidly declining. Pages are also easy to game, both by external forces (spammers) and internal forces (bias seekers).

Wikipedia editors are now working as Google's unpaid fact checkers. It's true that people edit Wikipedia for a variety of reasons, but for many contributors the site was supposed to be a free resource: contributed to for free with the promise that the content would be free to the reader and free of advertising. Instead it is being used to further Google's profit.

    The old Microsoft Encarta encyclopedia had different
    facts for different countries. In the US edition, Thomas
    Edison invented the light bulb, while the British version
    said it was Sir Joseph Swan.

If you search for "lightbulb inventor" on Google [1] it currently shows three "lightbulb inventors": Thomas Edison, Hiram Maxim, and Joseph Swan.

[1] https://www.google.com/search?q=lightbulb%20inventor

Remember back in 2005 when the stories started circulating about how supposedly Fox News won a court case that legally allowed them to lie?

Regardless of whether or not that's what was decided^1, I still question people who assert that somehow there should be a law against the news lying. I suggest that any answer about a controversial subject such as the one you list, even (or perhaps especially) a consensus answer.

I don't especially accept that Google (or any company that profits from accurate information, which is basically all of them) is a better caretaker of facts than a government.

^1 http://www.snopes.com/politics/business/foxlies.asp

Ha. SEO is about to get a lot easier. SEO is search engine optimization. Easy is an adjective and is the opposite of hard. "Ha" is an expression of laughter.

"A source that has few false facts is considered to be trustworthy," -- this leads me to believe that the algorithm punishes for including false statements rather than reward for including true statements, so this method wouldn't work

There are already relevance scores for keywords and pages. I doubt that this technique will work, and, if it does, I doubt the algorithm will roll out with such an obvious hole.

Easy and hard are relative.

What is "fact" is actually highly subjective and puts Google in the place of picking the "winners" for arguments or perhaps pushing their own political and religious agenda.

Fact: Vaccines are Dangerous Fact: Vaccines are Safe and prevent spread of deadly diseases. Fact: I nearly died from a vaccine. Fact: Vaccines contained mercury[1] for decades and poisoned untold millions of people. Fact: Google now decides which facts you get to see.

I understand Google is trying to get rid of lots of useless add pages but determining "truth" is potentially just pushing an agenda and burying those facts they disagree with.

[1]http://www.fda.gov/BiologicsBloodVaccines/SafetyAvailability...

To put a finer point on it, Google's success is based on statistical machine learning, but interesting facts are rooted in expert rule systems and predicate logic.

Someone, somewhere has to make that call: the current situation is untenable.

It's a whole lot more tenable than appointing a referee, which is what Google is trying to become here. An organization with the power to "settle" contentious debates becomes a target for corruption, a single point of failure for the integrity of the entire system.

For example, maybe the Waltons decide they don't like the national conversation on minimum wage laws; could they lobby/bribe/extort/whatever their way into having their "facts" be treated favorably in search results?

Or could the Democratic party decide that Republican candidate so-and-so is a pedophile/spouse-beater/alien, and make that a settled "fact" by promising Google employees favorable tax breaks in exchange for aid in "narrative building?"

Under the current system, "facts" are democratized. Sure, sometimes bullshit makes its way into the popular consciousness, but it's more random than not. When I see some tripe about "OBAMA TO INJECT TRACKING CHIPS INTO AMERICANS USING THIS ONE WEIRD TRICK," I'm relieved to know that there is no amount of money or power in this world to make that an established "fact."

While i like to see some certain sites removed, i fear that this could give google too much power of what is considered "truth" and what not. Does anyone knows good literature about such things and their implications?

It might make google's results less relevant. People don't believe in the truth. They follow the general consensus. They might switch to a different search engine because google 'gives bad results'. Mob mentality is real.

From the article it sounds like it considers general consensus to be the truth. Which sounds worse in my mind.

According to the article, Google's measure of truth is actually a measure on consensus. Which I find rather troubling.

In a sense, it already is. The link-based ranking system is a measure of popularity.

This concerns me as well. Frequently, I already know the "facts" on a particular subject, so what I'm looking for is some interpretations of those facts. Some information on how those facts actually apply to real life.

For example, here's a fact: 1.999... == 2. That's fairly well known to those who care, but it's the fact itself, but the implications of this fact that underpins an entire category of mathematics.

>>i fear that this could give google too much power of what is considered "truth" and what not

This was my initial reaction...then I realized Google already has this power just by choosing what comes to the top of their search results. Many truths are also highly contextual, which makes them hard to pin down.

It'll be interesting to see where this goes. My guess is it'll be short lived (like many of Google's projects), as regardless of how well their 'truth' algorithms work, people will go to where they can find the information they want (like the other commenter here mentioned), not the truth.

don't you think with the data they already have, they would have already devised some algos to know what is truth? its just accuracies of the truths that may vary in those algos

This sounds like an inverse information measure -- ranking sites highly based on their redundancy (repeated content).

And the people who are in the business of repeating their content in ever more inventive and adaptive ways are advertisers. So over time this seems likely to devolve into promoting the adverts and infomercials ahead of the content.

For instance rather than farming links, the spammers would switch to farming quotes or paraphrases. Expect "one weird trick..." to turn up a lot outside of just the advertising box, and the battle to be how then to stop the algorithms from classing it as truthy.

So if they count the absolute number of errors in a page, that's yet another incentive to cut content in even more pages, that drive even more ads, and even more profit for Google.

On the other hand, if they count the relative truth in a page, pages will bloat with marginal content, will get harder to read, people will scan them more shallowly, and as a result we will end up having different versions of the same content page but with marginal content tilted on one direction or another. So we will need to use more web search. So another win for Google.

Sounds fair :)

Google makes money (via ads) by showing the results the user wants, not by showing the results the user should want. If I want to go to an anti-vaccination site to reinforce my pre-pre-existing beliefs, it's hard to imagine Google will care, let alone make it more difficult to find it.

The thought of websites doing "fact stuffing" at the bottom of a page instead of keyword stuffing makes me smile.

Better title "Google is researching ways to rank websites based on facts along with links"

Not as sexy, but a bit more accurate.

This is an interesting case, and will probably be useful in some verticals, but not all. It's widely thought and sometimes verified by Google that they treat different verticals differently in search. "Payday loans" searches are treated differently from "New Macbook Air" for example, in which case we don't have to see it as an all or nothing approach. For example, today you only get the right-hand knowledge box on some searches, but not all.

I also wonder how New Scientist blanketing their page with ads like it's 2003 is affecting their rankings?

>> Facts the web unanimously agrees on are considered a reasonable proxy for truth. Web pages that contain contradictory information are bumped down the rankings.

But if there are contradictory sites then the web is not in unanimous agreement, is it? Something about this statement bothers me. Is there no chance that the web could be in "unanimous agreement" about something which is wrong? Seems to me that human populations make this mistake quite often, and since it is humans (for the most part) who write the web the same fallibility should show up there.

The article makes it sounds like they are replacing PageRank. According to the paper it's orthogonal to PageRank. It's just another signal to consider. The other assumption seems to be that this will cause site authors to fact-stuff. However it sounds like they more interested in using wrong facts to penalize a site than using correct facts to promote a site.

Is it just me or are google results getting worse with time? Five or six years ago I used to find everything on the first page of the first search, but these days I have to keep "rephrasing" my searches to find what I want.

Anddd this is why I no longer use Google but rely on DuckDuckGo.

So you use Yahoo.com which itself uses Bing.com.

DuckDuchGo (DDG) is a meta search engine that relies on the Yahoo & Yandex search API (http://en.wikipedia.org/wiki/Yahoo!_Search_BOSS) and adds some related Wikipedia/Yelp/Yummly/etc. snips.

So if Bing.com implements a similar rank algorithms your DDG/Yahoo results will too. http://en.wikipedia.org/wiki/DuckDuckGo

It's meta-searchengine-ness is exactly why I personally use it. A `!` away from any search engine I need.

True.

An open search engine was on "Show HN" on Saturday: https://news.ycombinator.com/item?id=9122397 -> https://deusu.org

Another open search engine from last year: https://news.ycombinator.com/item?id=6152839 -> http://www.gigablast.com

The European Union sponsored the development of an European Google competitor: http://www.heise.de/newsticker/meldung/Quaero-Erster-Vorlaeu... -> https://www.exalead.com/search/

I wonder how this will affect media companies' pagerank.

I wonder how the Goog will gauge the facts that Google sucks and that Google does evil.

Reply on news.ycombinator.com