Google won't make search recommendations for "Islam is" | Hacker News Reader

Google won't make search recommendations for "Islam is"

scienceblogs.com

131 pointsgort98 comments

Everyone knows exactly why they don't make search recommendations, and is there really anything we can say about it that won't devolve into trolling, flaming, and deliberate misunderstanding?

Everyone knows exactly why

Sure, but did you know that they were doing it at all? I didn't until today.

I don't think it's removed for the reasons you think it is.

It seems to be driven by what people search for. First option under "dead b" is "dead baby jokes". Try "death and r" and "death and return of superman" is above "death and resurrection of jesus". And one of the options under "death squads" is "death squads health care" If people don't use "Islam is ..." all that often it's not going to be part of their suggestion tree.

PS: First option under "muslim is " is "muslim is wrong". Muslim is not actually a religion, but I think far more people in the US use it than islam.

Hmm - you could be right, but as the article takes pains to note, all of the obvious analogous searches do return suggestions, which I think is evidence against your idea.

Still, it's been pointed out that queries like "Muhammad is" still "work", which I think is evidence for your idea. :)

Right, I think the author of the article is not really interested in uncovering the (presumably) interesting heuristics for inclusion/exclusion in the autosuggest list, and is rather just throwing his prejudices out in the open.

In a related thought, try "Islam will"

"I am extremely" provides further amusement as a search suggestion prefix.

"why is my" is also quite entertaining, with many fun breakoff suggestions possible. "Why is my poop..." makes the list several times.

Yes, fix the algorithm by requiring a higher cutoff for what is shown.

Why tune the algorithm in a way that would make the service less useful just because some people are sensitive? I guarantee you I've been helped by google suggestions that occur less frequency on the web than, say, "islam is false" (which, judging from other religions, would be suggested if it were not blocked).

What about racist suggestions?

How does a frequency cutoff filter out racist suggestions? Or were you referring to another sort of automatic cutoff?

NB. Interestingly, "christians are" and "atheists are" don't return any results either.

I suspect it would filter out much of what is offensive. But you need to try it and see.

"white people are stupid" (without quotes) has 218 million results. It's one of the suggestions that google (presumably) blocks. That's many more results than a lot of useful suggestions ("michael jackson songs" has 88 million).

It turns out "(race or religion or nationality) are" is blocked for most races, religions, and nationalities.

Possibly the starting "Is" in Islam throws the googleboxen into a self-recursive Gödel gyroscope.

For example, if you try with related phrases such as "Muslim is" it works. So my guess would be algorithm breakdown because of "is" repeat.

Actually, I tried a few other examples using "minorities" as the start of the phrase and then using "are". They are all censored. Non-minorities aren't; e.g. "men are" works fine.

So my algorithm implosion theory is clearly wrong.

Not so...try "Isaac is "

Probably not. "Island is" works. It's not even that you don't find offensive things in the dropdown; it's that there's suddenly no dropdown.

Weirder still is that "Muslim is $ADJECTIVE" is ungrammatical. Typing "Christian is" turns up some nonsensical results, not the same negative ones that "Muslim is" or "Christianity is" do.

I don't know what your theory is ("everyone knows exactly why"), but I'll put forth my thesis.

1) Autosuggestions are a sample of the most frequent phrases google users type in

2) This autosuggest list is analyzed statistically and a sentiment score established ranging from 0 (friendly) through 1 (hateful).

3) If the autosuggest list statistics reveal a certain characteristic (e.g. all hateful, or very controversial, etc.), then no suggestions get returned.

If I'm wrong that at least in principle this is an automatic, data-driven behavior, then I really feel sorry for the sod who is maintaining a list of what phrases aren't kosher.

...so "buddhism is bulls__t" is where on that scale?

I think the counterexample to this theory is "Christianity is" which returns 7 negative suggestions, 1 ambiguous, 1 positive, and 1 t-shirt.

Another argument is that typing in "Islam i[letter]" where [letter] is some common consonant usually returns suggestions.

Maybe if it came from a muslim himself.

Google is just being wise here. Why would they risk offending people who are already offended?

Not many would be offended when the completion suggests "niggers are.." - it is a thing of past; "jews are.." - they are sane and educated enough not to go batshit; "indians are.." - pretty well integrated with the western society. Muslims are neither that educated nor integrated well with the western world and we have a stinking situation in hand like occupied countries and a threat of having more occupied countries.

The minute after Google suggests completions about Islam there will be forward mails like "GOOGLE IS RUN BY JEWS AND THEY HUMILIATE US ALL!!!1". It is an effective bit of propoganda and there's a possibility of causing a loss of market share. And mind you this is the least of what could happen. So, again, why would they even risk it at all?

why would they even risk it at all?

So, you're saying terrorism works. It's ok to offend groups who don't resort to terrorism to destroy freedom of speech.

That Google operates censorship in China and India is bad enough. Now you're saying the number 1 search engine in the west is right to censor.

I'm not saying terrorism works. I'm just saying Google is not a non profit organisation promoting free speech.

You are saying terrorism works. You are saying that one of the most important companies in the world should discriminate in favour of one religion, because the adherents of that religion will use violence to get their way.

Google has chosen to position itself as an ethical company. Google's motto is "do no evil". They want us to think they enshrine all that is good.

Discriminating in favour of irrational and violent religious fundamentalists is evil.

Oh boy is this what's going on Abu Ghraib? How you got that meaning from my post I'll never know.

> Muslims are neither that educated

That is an interesting comment.

I dont think your right BTW (generally in your comment). I expect it is simply Google is avoiding any fall out. It's one of those current affairs terms that eventually somone would think to Google then write a blog post titled "look what Google suggests for islam is..."

Lose-lose all round. Best to go for the option the least amount of people would give a crap about.

EDIT: google doesnt offer suggestions for any of your other examples either btw.

It's a bug. It will be fixed.

I think this may very well be true. It has always been the case that some search terms simply don't yield suggestions. The fact that 'islam' is among them, could very well be a mere coincidence.

Yes, everyone knows exactly why they're doing it but:

a) By censoring some terms they open themselves up to more criticism, they can no longer point out that everything is done algorithmically; and

b) It's still possible to get offensive search suggestions out of Google, try typing 'Allah is'

This isn't new. They've been filtering it for years, ever since some screenshots of what happens when they don't made the rounds.

From November 2005:

http://google.blognewschannel.com/archives/2005/11/30/does-g...

"Blacks are lazy"

"Blacks are stupid"

"Jews are evil"

"Jews are cheap"

Interesting. One also can't get "whites are", "Hindus are", etc. But one can get "Christianity is", "Buddhism is", etc.

Interestingly, the same is not true for Islam was ("founded", "founded by", "spread by the sword") or Muhammad is ("a false prophet", "satan", "the antichrist") or was ("born and raised in", "born", "illiterate"), or Allah is ("satan", "great", "not obliged") or was (which, intriguingly, has just a single suggestion, "a moon god").

I'm actually having a hard time finding anything else that is similarly blocked. Makes you wonder what the completion list had been popping up before they blocked it, eh?

It's not that hard to find other blocked terms. Just try something that's socially sensitive or offensive. Such as 'n----rs are ' or even 'chinese are ' (note: 'swedish are ' works).

I wouldn't be very surprised if they had some sort of algorithm to generate the blacklisted terms as opposed to manually cherrypicking them. We're talking about Google here.

Google could learn a lot by analysing which suggestions are presented and how the user reacts [chooses one, presses enter as is, leaves site, types something else]. You might be able to correlate this with how useful / how hateful the suggestions are.

Except that in many of those searches, the person is looking for something hateful.

My pet theory was that it was based on the length of the query, since "Islam" was the shortest word on the list of terms. That got shot down by "Allah is ". I guess it really is censoring.

Not censoring. Just not actively suggesting certain things. That stuff is all reachable if you want to type the whole thing yourself.

Perhaps there is too much noise or too even a distribution - if there are too many nearly-equally weighted results for "Islam is" to make a good suggestion, does it not even try?

I haven't seen it pointed out yet that it doesn't actually block searches for "Islam is". It only blocks recommendations. It's not like they're censoring actual results here.

What they're doing is simply disabling a minor feature for a small number of potentially inflammatory use cases at the expense of about 10 seconds of amusement for you. If this offends you, I would think it says more about you than Google.

I think it's more of a question of why is it just "Islam is"?

It doesn't exactly offend me; I just find it interesting and potentially troubling. It's part of a trend of going out of the way to avoid offending Muslims specifically.

I fear that things like Ireland's new blasphemy law, which came into force a few days ago, are the result of this trend. (Of course, it applies to mockery of all religions; in the West at least one can hardly get a law passed outlawing just mockery of Islam.)

I just read a comment elsewhere that such a trend "incentivises outrage", which I think is exactly right.

If we only respect religions that cut off our heads, eventually those will be the religions that rule us.

My reaction to this is that people who are easily offended must have some insecurity. Where this insecurity is not justified, it reduces my respect.

To be fair, religious affiliation seems unusually tightly bound to some people's sense of who they are. What's unclear to me is why certain religions have this worse than others, e.g. compare Muslim reaction to the Muhammad cartoons with Christian reaction to "Jerry Springer: The Opera".

I suspect if Jerry Springer: The Opera hd been written by a group or country that certain sects of Christianity want dead then the response would be the same.

It was groups of hardliners with their own agenda (aka not representing the religion) who, for the most part, provided the extreme reaction. The majority showed roughly the same response as to JS:TO.

I can't speak for anyone else here, obviously, but I've never had my head cut off, whether by Islam, Muslims, or any related noun. "The plural of anecdote is not data" and all that, though, so take my story with a grain of salt, eh?

Isn't that how most religions get their start?

What I wonder is how the autosuggestions are generally generated.

From what users type in?

If so, and certain phrases get censored from autosuggest (even though "Islam is" is specifically in the censor list, it is not the only phrase that is specifically on this list), then my conclusion would be that the majority of phrases that googlers type in that start with "Islam is" are hateful? Hence the exclusion. In fact, what phrases get excluded might even be data driven as you can deduce sentiment from the most common phrases that users type in.

> It's part of a trend of going out of the way to avoid offending Muslims specifically.

A lot of other "<something> is/are" terms are filtered too. I dont think this is specifically about offending Muslims.

If they are censoring the suggestion list, why do you assume they are not censoring the actual search results?

So much for letting the algorithms rule their results. They could at least be somewhat equal about it and do the same thing about christianity and judaism, but then again, they know that nobody is going to chop someone's head off when they see that the first result for "christianity is" is "christianity is bullshit".

How long is the civilized world going submit to this insanity? When is enough, enough?

They could at least be somewhat equal about it

In theory, I'm not sure that I find the addition of a whole bunch of exceptions to be better than the current situation.

In practice, these recommendations are all a bit useless.

Let's hope someone at Google is reading this and gets this fixed PRONTO. It is grounds for boycotting the service IMHO.

Huh? Since when common sense, avoiding trouble and trying not to offend socially sensitive groups count as 'grounds for boycotting'?

There's apparent favoritism here, or bias against, depending how you want to look at it. If there were no filtering going on either way, I'd have no complaint.

Google is probably the hardest company in the world to boycott successfully.

To be fair, they do block "jews are" and "blacks are". And, for that matter, "christians are". But you're right, they should be more complete.

It's tough to get every case of this and still have an autosuggest. They also got "muslims are" in the block list, but even then "muslims can" returns "muslims can not be trusted" within the top couple results.

Define "algorithm" because it looks like they still are letting algorithms rule results, they've just changed the algorithm. That seems like a pretty good idea to me, because if by 'algorithms rule results' you mean "page rank as released on Google 1.0", I'd be seeing some pretty crappy search results.

Why do you think "jews are" doesn't autocomplete? Or how about "jesus sucks"?

Another reason why I Google with the Bing

http://img.skitch.com/20100105-f2xhs1pm1bmi1wt24cng2hddyw.jp...

Cause everyone knows "Islam is the light baby doll"

I would assume they put things in the list on a trial basis otherwise they wouldn't be able to respond quickly to current events.

This is rather disturbing. I didn't live through a dictatorial government in my country, but 30 years or so one that was ruling for about 50 years was put down. The worst thing about a dictatorship is not the censorship itself that will cut the parts that they don't want to be read from a text or a poem, it is when the author himself limits its expressiveness and freewill to the walls built by the fear of the censorship, and will instead write the version of his feelings that he thinks it will not get cut.

Google is doing the same... actively self-censoring, the worst form of censorship.

Aren't they simply withholding phrase suggestions which they deem wrong? I.e. you can still search for anything ranging from the angelic to the vile, and they probably don't censor (but of course, there's no way for an outsider to ever know).

What I do wonder is whether Google, for example, actually filters out from its search results websites that are blacklisted by the Australian government, when people down under google.

If that is the case, why are they not being consistent? Why withhold suggestions only in the case of "Islam"?

They don't. They also withhold suggestions in the case of "jews are" and "white people are" and "hot teens".

How would the phrase suggestions in question be wrong? They have a severe case of political correctness!

If you search for "muhummad is " you get the results you are looking for.

"islam should" is decidedly accurate as well.

True.

But I would think that people at google would be thorough. Most perplexing.

There's been a popular trend of posting screenshots from Google search recommendations over the past few months; I wouldn't be surprised if one of those shots made it to Google Headquarters (though I'm certainly surprised they didn't notice it sooner!).

Also shame on PZ Meyers for yoinking that article from atheists.org.

There's actually an entire site devoted to it, in the style of FML/IMMD/etc.: http://autocompleteme.com/

(Sorry if I just wasted anyone's evening!)

He links to that article from atheists.org right before the blockquote. I'm not sure what he should be ashamed of.

I don't think a blacklist is plausible. There are similar phrases that yield negative results such as "is Islam"

Perhaps there simply are not enough queries for "Islam is" in your region.

I submit something that may be potentially useful, a guy who is going to open source out something that might be real interesting ->http://news.ycombinator.com/item?id=1032278 and I get no love and this ahem "sensationalist" blog with no proof whatsoever for the claims gets so many points - wow!

It doesn't make any claims (aside from claims about the behaviour of the Google search field, which are easily verified). It asks a question. But don't feel bad - whether your submission gets voted up is half luck: either a couple of people spot it and vote it up so it becomes visible to everyone, or not.

"Engineers are:"

http://dl.dropbox.com/u/147719/Picture%206.png

http://searchengineland.com/islam-is-blocked-by-google-sugge...

A [Google] spokesperson tells us: "This is a bug and we’re working to fix it as quickly as we can."

I wondered if its because one word is the prefix of the other. My first attempt, "ishmael is " also suggested nothing!

Then I tried a few others; "isaac is " and "israel is " which did come up with suggestions.

Oh well.

I found this pretty interesting, not sure if it still works:

http://emerging-patterns.com/tmp/CrimsonJihad.png

Maybe it's just a bug? Haven't you found phrases which you thought were very popular and yet couldn't find Google recommendations for them?

Well, try muslim is for that matter.

Which is grammatically questionable.

ok ok, even muslism is then.

Eh? My point was that you wouldn't necessarily expect them to have thought of it. They certainly won't have thought of this bizarre misspelling.

Fair enough, I am unfamiliar with the misspellings. Musli(s)m sounds perfectly normal to me as a non-native speaker of the English language, and for a bunch of other people on the internet for that matter hence the suggestions of google.

I just tried to say that out loud and find it quite a tongue twister.

Search engines should avoid being in the business of egging people on. That has to be one messy program/algorithm/filter. We put one together for a retail company emails and stopped 20% of the incoming mail. It was pretty funny what combos you can get.

Note: to replicate the stated behaviour, you might need to type in the space, e.g. "Judaism is " rather than "Judaism is".

try out "Obama is"

Interesting... It's fairly difficult to find male celebrity names that do not come up with "___ is an idiot".

chuck norris ...

clint eastwood