Will Duolingo implement (profanity and/or other) filters?
Bonus question from the series Ask us about Duolingo, Round 2!
Will Duolingo be implementing a profanity filter - or any other filter - in the forums? Are course developers restricted in any way in the vocabulary they choose to teach in their course? In general, how does Duolingo view the contradiction between restricting language used and taught here - particularly restrictions based on culturally-specific norms not shared by all users - and a free and open language learning environment?
This is an interesting question and worth an open discussion.
As you know from our guidelines, we generally err on the side of openness and tend to only take action when discussions get spammy or abusive. The same general principles apply to course creators. We celebrate and encourage curiosity, questioning, and cultural understanding. But hate speech doesn't translate well in any language.
While profanity filters have not been implemented across the board, we have tested them at various points on different platforms and are considering rolling out more protective filters in the future.
We understand this challenge is ever-present and that even the clearest guidelines are subject to debate. That's why our staff and moderators often discuss edge cases and consider each situation carefully before taking action. (And why you should contact firstname.lastname@example.org if you feel you’ve been unfairly moderated.)
(On that note, moderating is harder than it looks! Everyone should try this quick NYT comment moderation quiz!)
So we pose the question back to you: how do you all view the contradiction between restricting potentially offensive language and keeping a free and open language learning environment? We're listening!
My opinion on swearing is about the same for all platforms and situations and is really quite simple: if it helps prove a point or emphasize a point, it is fine and is just using language in speech and writing, something that Duo ought to support.
But if it is there just to be rude, licentious, or just plain obnoxious, it should be restricted.
(Yet another situation where I believe there should be a 'report' option on Duolingo...)
In comparison to other issues on Duolingo, I would say that profanity is a small one. I would rather have time and effort be spent on the larger, more prominent issues than on a profanity filter.
I agree, it would be cool to have that option though. For some languages like spanish, it's but then there're some like Russian where it's more tricky. I'm up for it.
I don't think any word should be filtered when discussing the word or language itself. It's when people start hurling those words as weapons against others that they should be removed.
We learn a lot about cultures through understanding which words are taboo or considered innapropriate.
Do not narrow language learning to cater to a small group of sheltered home schoolers. They either need to grow up eventually or have their parents shelter them some more by turning on the filters.
I agree with you on that. "Bad words" usually don't start out as bad. Someone comes along and associates them with something or someone bad.
If you are familiar with swear words you obviously aren't that sheltered ;)
I think profanity is an important part of language to learn, not least to understand if it is used towards oneself. Or to avoid accidentally using it. (A well meaning French teacher once gave me a phrase she said meant "I don't mind" but really meant "I don't give a..." which ended up causing me problems.)
However, I do not think it should be part of the tree as some people are very sensitive to it. Perhaps either a bonus skill or allowed in clearly marked forum threads.
Then there's the question of where to draw the line about profanity. It's virtually impossible to become truly fluent in French for example without at least a few words that a direct translation into English would be regarded as profane but which are accepted as every day parts of the language in French.
Using language that is by its nature abusive to attack others is unacceptable, accusing people of things they haven't done, including because they used language in its proper sense is inexcusable, and censoring language used correctly because someone somewhere in the US has come up with an ephemeral slang usage is totally unacceptable - not least because it reinforces the slang usage that would otherwise die out in a few months or years and helps to drive the real word out of existence.
If there was a profanity filter, how would you tackle the issue of swear words that mean something completely normal in another language, such as ❤❤❤❤❤❤❤ Hell being a beer brand?
Not to talk of all the words within a language that can be used offensively but are otherwise innocent.
It might be possible to have an algorithmic component to solving this problem, but an algorithm isn't going to be able to solve the whole problem:
The Scunthorpe problem is only the start of the difficulty of finding naughty words in a passage of text without any false positives.
Offence and naughty words don't necessarily go together. We teach young children to avoid naughty words because they don't yet have the experience or maturity to cope with more complex rules. This is a site for grown ups, so we need to do better. If a commenter named EricCartman were to post a comment saying "why should we listen to you, you're ginger, and all gingers are the spawn of satan", then that's offensive, and needs dealing with, but has no naughty words. On the other hand, there are sometimes discussions where it's necessary to use a naughty word - this discussion is an example; I've linked to a discussion of the Scunthorpe problem, and I think I was right to do so, even though that article contains a very naughty word.
These things are complicated, and a simple profanity filter is going to solve the wrong problem.
Actually, I don't think is specifically "a site for grown ups." There are plenty of children and young teens who participate. There are no rules against this and, in fact, Duolingo for Schools tends to draw in kids who would never have found the site for themselves or been directed to it by their parents.
The site does indeed welcome children and teens, and hooray for that. Perhaps it would be better to say "a site for people willing to behave like grown-ups".
One of the great things about communicating by text is that I've no idea who I'm talking to; it may be that one of the contributors to this discussion is very young. If that was the case, that contributor will be taken as seriously as anyone else, and will face no prejudice because of their age.
We shouldn't accept offence or needless profanity on these forums, but not because young people shouldn't have to put up with it; rather because no-one should have to put up with it.
We also shouldn't accept limitations on our ability to discuss languages and languages learning, regardless of the age of people using the site. I stand by my decision to discuss the Scunthorpe problem in my earlier post, and I welcome anyone to this discussion, regardless of their age, as long as they're willing to join in a grown-up discussion.
It most definitely is not "a site for grown ups" and I can't imagine why someone would think it is. It used to have a lower age limit of 13 in the terms (linked to at the bottom of the page), but then that age limit was removed.
You can be rude with someone even without using any "bad word", so there is no point for that.
Also there are words that are perfectly fine in one language and not in some other one, so it would be really weird to put some sort of filter as you could probably mistake a good question with something else.
I'd prefer you guys to take your time do develop more courses and improve the existing ones.
I swear a lot in real life but I've actually come to appreciate the no profanity rule on Duo and I find that it makes discussions nicer, so I think you should definitely keep that. But implementing it with a filter won't work. Whichever way you design that filter, people will get around it. What you need is more active moderation and better organization of the forums. Only real people can create the discussion climate that you want.
(I'm a course contributor for Swedish and I moderate sentence discussions in our course.)
Well.... I think a filter with a warning message is quite the way to go.
Whoever wants to work around the filters is intentionally violating the guidelines.
If I had been filtered and warned, I'd certainly pick other ways to say what I want, but not other ways to use the bad word.
I guess that wouldn't hurt, but the way I see it it would only help with a very small portion of the problem.
The concept of a profanity filter greatly disturbs me, because in practice it is impossible to remove all terms that can be used offensively, without also removing a great number of posts that are using the word in an innocent context; either because the word has both a legitimate use and a regional, offensive one, or because it exists in more than one language with completely separate meanings.
I agree with MacIomhair's suggestion about a separate, locked area for teaching about which words are taboo in various cultures. Different cultures also have different standards of acceptability, and that too is something that needs to be learnt and accepted.
It causes me no problems to avoid using profanity on these forums, because I do not normally use it in everyday life (I make no promises about what I'll say if you drop something heavy on my foot!)
Nevertheless, I always try to learn the most common expletives in a new language as soon as possible. I have no intention of using them, but I do wish to avoid them.
Firstly, as a beginner, I may sometimes mispronounce (or misspell) something - this is most likely to cause problems if this results in unintended profanity.
Secondly, these words provide a flag not to proceed further with a conversation. When you first start using a foreign language, there will be many times that you get a reply that you do not understand, and may have to ask for explanation. But if I ask (for example): "Excuse me, is this the bus stop?" and get the response "F&&& you, b&&&", I do not need to understand what precisely the words behind those ampersands meant, but I do need to understand that the speaker is hostile, that asking for an explanation would be unhelpful, and that I probably need to get out of this situation. (Yes, I may occasionally break off conversation with someone not aggressive, but who is simply extremely casual in their use of expletives, but I would prefer to make that mistake than the alternative ones.)
Briefly: I support whole-heartedly Duolingo's zero tolerance of language used offensively (whether or not it involves actual profanity), and I can understand its decision to also forbid casual use of profanity for emphasis etc.
But linguistic and cultural mutual understanding also requires learning what terminology on sensitive issues, although innocent in one's native language, become offensive if translated literally into another. We need to be able to point this out to one another, without the explanation being automatically deleted by the filter.
Another shortcoming of profanity filters is that they can't distinguish intonation. There is a very great difference between 'Hi!' while waving at a friend and 'Hi! You!' to a stranger, which signals aggressive or, at the very least, unfriendly intent.
Equally, is perfectly possible to insult or belittle someone using correct, standard vocabulary. It is a matter of context and intent.
I don't like automated profanity/sexuality filters, because computers are basically stupid and remove posts that may be perfectly appropriate because they contain terms that in other contexts may be considered inappropriate. A filter that holds posts for review might be ok, if the moderators have enough time to handle the traffic. But - I remember one thread here where a woman wanted all reference to any alcoholic beverages removed from the site, and another where a young person felt that any disagreement with her was being mean and bullying, so you can't please everyone. If you have a really good command of the language, you can be very nasty without using anything that would be considered profanity by a filter, and one person's - oh, pointed but clever is another's being rude and insulting. Given the amount of traffic here, I think the moderators do a good job. they remove true spam quickly, and respond to abuse complaints rapidly.
I like your reply and agree particularly that overworked moderators do a good job on this excellent free learning site. I lingot you...L
Personally I find senseless comments more annoying than profanity although if bad language was used in a conversation with me, I would not be pleased. Just saying "Hi" on a discussion forum, supposedly to get attenntion, seems pretty pointless and should be stopped. It's all about respect, at the end of the day.
I like a profanity filter, there are some words that shouldn't be used on here in any cases (like the N word). The problem is that some people are more thick-skinned than others, which means that some people find less things offensive than others which can cause some controversy.
You never did talk about a spam filter, though. Seeing "pls lingots or ur ganna die XDDDDDD" every day in the forums is annoying, and people who just bang their heads against their keyboards and think that its a good idea to post it ("fhodsiafuahdsfoaslik"), should be removed because it doesn't contribute to anything at all.
As for upvotes/downvotes, could you restrict it to keep the up/downvoting to users who have actually finished one lesson and are not shadowbanned? That way people can't just create accounts to spam downvote others who they dislike, they would have to put a lot of effort into it, and which if they did they would need a lot of e-mail addresses, which brings me to my next point.
Could Duolingo verify email addresses? This would definitely slow people down who are spam creating accounts to spam someone. I never got a verification email asking if it was me who made my account, if this has been in place before it would be good to put it back, so spammers must create fake emails to make accounts to level up and then downvote, which then by this point i don't think anyone would go to that trouble(or at least most people wouldn't do this)
Lots of new people can't see the rules/guidelines when they first join, which would be super helpful because there are people who need help to basic questions that are in the guidelines/rules that aren't clearly visible on the site. (They are in the drop-down menu and the bottom of the page in small text)
Personally, I haven't seen any profanity that needs to be filtered, and -no offense- I don't think it's moderators' work.(Moderators are human beings, you know) So maybe a profanity filter is needed for titles, or if word frequency is too high in one post. But not a scythe-filter, I think the filtered discussion is better to be hidden until a moderator approves it. Besides, where was that email@example.com? I hadn't seen it. onthehype25 pointed out that Guidelines [and many other mandatory info] aren't easily visible to new users, even older users.
About spammers, Email verification(Both post-registration verification (making sure it's not a temporary address) and past-registration verification), IP observation, device registration and "complete something to have voting or posting privilege" have been suggested before. You can also restrict email addresses to famous ones only, like Outlook, Yahoo!, Hotmail, Gmail, Sina, etc. This way unknown providers of temporary emails will be blocked.(Edit: this method is very unpopular and irresponsible, and it's extremely discouraged. Refer to b05aplmun.ca's comment down here) Yahoo itself offers temporary mail, but there's always a certain code in the address. If there are too many addresses with the same code, it should be blocked.
"Unknown providers of temporary emails" might be blocked, but so would, for instance, universities and workplaces, which arguably do a much better job of vetting the people who have their email addresses than, say, Outlook does.
They tend to provide email addresses to people who they know actually exist and who have or have had some connection to them and the establishment of that connection, whether as student, staff, or faculty, often includes some fairly serious vetting.
I don't know about anyone else, but I would not have gone out of my way to establish a new email address just to use a language-learning site. I would have concluded that the site had weird and annoying rules that I didn't feel like complying with and moved on without trying out the site's services.
I wasn't thinking of that, I just wrote the solution I had seen somewhere else. Thanks for noting it. I'll edit my post.
Both IP-lock and device lock seem particularly inappropriate for a community whose target audience is likely to include a large number of travellers.
Actually, I wanted to edit this IP-lock two days ago and change it to "IP observation" or something like that since I myself use VPN most of the time.(Thanks for reminding me, a lingot for you!) But device lock... the "device registration" is better here. But Duolingo shouldn't necessarily block people out when they're on other devices but ask them to verify their ownership instead. This would also set their account to traveling mode to ensure any public device would be removed in a certain time.
Personally, I am adverse to typing personal details into any public machine. I don't mind logging into Duolingo precisely because my account does not hold my personal details. I would probably cease using Duolingo when away from home if they required me to type in such information. So how do you envisage then "verifying ownerhip"?
[Note: if they took the approach of requiring me to translate a certain number of sentences in the languages I have supposedly been studying, that would be another matter! :P ]
Lol, I'd prefer losing my streak! Like email or picture verification, when you attempt to login with a device not registered to your account. Since passwords are personal, coffee nets usually use Deep Freeze or a free security solution to eliminate petty viruses and logging softwares. However, one should use incognito mode, virtual keyboard and temporary security solutions on public computers ("be paranoid", I mean). If there's a picture verification (you have a picture nobody else has and can use it an proof) this would be easier.
If nothing else, the publicity damage would be prohibitive.
In any case, if we want to know what they'd be prepared to do, let's look at what they've done. They're short of money right now, and they haven't done any of that so far.
What is it you think Duolingo is going to do with your email address?
As an app for one general use, this invites people of all ages and backgrounds to learn a language. In this case, we need to consider that children could be learning on these sites, making profanity inappropriate. For this reason I feel that profanity should be discouraged.
The moderation quiz was pretty cool. Looks like I should work for NYC. (Just kidding Duo, I wouldn't leave ya ;)
I got a 100% ;)
The first one kind of tripped me because I thought the word was "Republican" simply mispelled, I didn't think it was an insult xD
It is impossible to download" The Wind in the Willows" without deleting one or more words in every chapter ! Do not the American or English children read this book ? Or do they have to buy an edition with blocked or deleted words? The English literature of the 19th century is banished from Duolingo because Dickens, Kipling, Trollope, etc. had written in their works words as profane as queer, ❤❤❤❤, jungle, gay.. DuoBot is a perfectly idiot machine ! And why the French people do not have the right to take "une douche" (shower). Well, you often say "bite", "bit" or "face" which sound very dirty to French ears. We forgive you all the same...
I don't like it when others swear at me on Duo : / The first time I've ever been cussed at was on this site.
Yeah, I hope they'll have to do something about that.
Wish I could add something constructive. However, I will support Duolingo whatever it decides, as it is a brilliant site run by good people. It is just so sad that there are others who rely on being offensive to seem cool and clever. When will they ever learn, as the song goes. Auguri to Duolingo!
It would be interesting if they did have that option later on. If they do, they should start with romance languages like spanish, portuguese, french, catalan and italian and then add other ones because, those are the easiest to learn.