Show HN: Can anonymous chat within identified groups create better discussions?

http://freeversation.com/blog/2011/06/29/identified-anonymity/

by akharris • 15 years ago

165 104 15 years ago

spking• 15 years ago

Possibly. However, I've found that it's fairly easy to identify who is saying what based on writing style, grammar and punctuation habits. Any HR person will tell you that's how they identify respondents to "anonymous" employee surveys.

akharris• 15 years ago

Agreed that you can probably find commonalities within the individual comment threads, but I feel like you'd be hard pressed to do it within a large group.

Though I guess those semantic specialists from HR could do some damage.

terio5• 15 years ago

Perhaps you can try translating the message to another language and back. Then present the resulting text to the user for corrections. The resulting grammar and punctuation could be very different than the original.

rl41• 15 years ago

This could easily mangle the original meaning of the message, as www.translationparty.com gracefully demonstrates.

I suppose that the user could take a second to approve the recycled message before sending it though.

jabo• 15 years ago

You could probably strip all punctuation and capitalization from any user input so that the style remains fairly obfuscated. IMO it's punctuation and capitalization that usually gives away the IDE identity especially on a chat medium.

hammock2• 15 years ago

I suppose it depends on the size of the company, but I can't imagine any HR person at a co with >20 ee's that could identify someone based on their writing style unless it was particularly idiosyncratic.

bricestacey3• 15 years ago

I imagine the answers would allow you to shrink the sample down to a few select individuals. E.g. This anonymous survey complains a lot about a supervisor, let's start with his/her direct reports.

bugsy4• 15 years ago

This is a good point. To date I have not had an anonymous survey in a small group that hasn't bitten me in the ass afterwards.

I also find some anonymous surveys interesting when they are placed in my mailbox and come with a unique identification number. When asking about this uid, I am always told it is for quality control purposes or such. Yeah right.

Anyway, the proposed idea is interesting. I could see it being abused though. Someone from management, A, who wants to know what you really think of them could set up a fake discussion using several other sign ins from coworkers B C and D (and of course management has their passwords, that's required per the corporate contract), puppets who then ask "What do you think of boss A?" followed by puppet accounts B-D saying A is a jerk and then listening to see your response. This is the sort of thing that happens with the sort of jock management that infects many workplaces.

yuhong• 15 years ago

Of course, in the long term these problems needs to be fixed. That will of course take time though.

ericmoritz• 15 years ago

How about each message has to be rephrased by another person in the room in order to scramble the original author. It suffers from the telephone problem and general misunderstanding however.

qjz• 15 years ago

I've worked in environments where it would be trivial to get an email of your SSN the moment you visit an HR online survey, let alone complete it.

It's possible to promise confidentiality, but you should be extremely suspicious whenever someone guarantees anonymity.

kahawe2• 15 years ago

Could you elaborate on this with some real world examples? Unless it is something very obvious like using "..." or "-" a lot while nobody else does, I am wondering how this can be done reliably and efficiently especially considering that what HR knows about your writing style is most likely only from business emails, your application and other business correspondences where one would likely use a different writing style than in an internal anonymous discussion.

JoachimSchipper• 15 years ago

I find it funny that you use a 66-word sentence to express this sentiment. Just how common are those, I wonder?

highace• 15 years ago

This immediately strikes me as a "why didn't I think of that" idea, but in practice (in my workplace at least) everyone would figure out who was who based on the writing style and general tone. In a larger more anonymous group of people (like a class) you could probably get away with it, but then what have you got to lose in the first place by just voicing your opinion publicly?

But good luck, I think this is a pretty cool idea.

pavel_lishin3• 15 years ago

It seems best suited for a group who are familiar, but not necessarily friends. A class is an excellent example; close coworkers are not.

And depending on what it is you want to say, you may have a lot to lose. Have you never taken a class with a teacher who displays blatant favoritism, or is sexist or racist enough to make a difference to you, but not to the administration? Being able to say something without an authority figure being able to punish you can be pretty important.

dbs119• 15 years ago

One of the co-founders here; we entirely agree that the concept works better in large numbers. We're particularly interested in how the site could be used for those larger classroom or office discussions, as a way to increase productivity and generate more innovative ideas. It could be really useful for social circles as well, but the group would likely have to be more than 10 people.

nowarninglabel• 15 years ago

Neat, I had thought about putting this same concept into action, though it seemed like anonymity would be really difficult until you had a large enough user base. Otherwise, you invite say 10 friends to a conversation and only 3 accept, then it's going to be pretty easy to narrow down who is who.

terio5• 15 years ago

We at spottiness.com are very interested in this line of work. I can see the appeal of the simplicity of the solution you propose. On the other hand, in many cases the invitees would prefer to not have their emails associated with their comments anywhere (your service would have that information internally). That is an interesting problem.

dbs119• 15 years ago

just to clarify, we don't associate emails with individual comments even in our backend. Emails of those invited appear at the top of the discussion page. Each comment can be posted independently of whatever email is up there.

terio5• 15 years ago

But the participant would have to take your statement at face value.

bricestacey3• 15 years ago

dbs11, do the emails appear in any given order? You may want to be sure the order is randomized, otherwise the person who starts the group might be able to figure out who is who rather trivially.

luminarious• 15 years ago

It would be fun if this approach was also used for the Google+ Circles, so you could have anonymous conversations within a circle.

nyellin5• 15 years ago

Another co-founder here. Circles integration is a cool idea. I'll look into implementing it.

scrrr• 15 years ago

Perhaps you could run each message through Google translate or something similar (e.g. English -> French -> English) to obfuscate the writing style. Otherwise, as others have already pointed out, it is very indicative of who the author is.

terio5• 15 years ago

I also mentioned that idea below :)

pavel_lishin3• 15 years ago

I foresee this having hilarious unintended consequences.

terio5• 15 years ago

The user would have to correct the resulting text anyways.

andrewcooke• 15 years ago

i wonder if you could repeat with different languages and take some kind of semantic median? interesting problem.

[just to clarify, given the replies - i am talking about repeating in parallel, not in series. the idea being that errors might cancel out, rather than accumulate]

6ren• 15 years ago

You could use mechanical turk to rephrase each message. ugh that is so like Vinge's focus.

widget• 15 years ago

I don't think that's necessarily any better. Ever played Telephone?

hammock2• 15 years ago

You'd need to run it through a non-deterministic model for that to work. Maybe select a couple languages at random to run it through, rather than the same one every time.

pohl• 15 years ago

I tried it with your comment, and it came back the same. However, English-->French-->Polish-->English came back with something altered:

Perhaps start every message by the Google Translate or something similar (eg English -> French -> English) to hide the writing style. Otherwise, as already mentioned, this is very revealing of who the author.

Omnipresent• 15 years ago

very nice idea. I had an idea similar to this but it didn't involve the anonymous aspect you bring in.

It would be good if you included markdown

nyellin5• 15 years ago

Duly noted. Markdown isn't a priority, but it's been added to the wishlist.

ryan_brunner• 15 years ago

To lend some credence to the "it's easy to tell who's writing what", we recently used Rypple as a team review aid for team members to anonymously ask for feedback from the rest of the team.

It was immediately obvious who was writing what. Granted, we had a relatively low population of people writing feedback (5), but I'd imagine it wouldn't change much until you had at least twice as many people.

mtdev• 15 years ago

This looks really cool, I like the layout. I have been working on something similar, (http://mobilethread.com), which lets you setup/join anonymous chat threads and is designed to run in a smartphone web browser. I got around the emails issue by creating a unique URL for each thread which you can then email/SMS/embed in a QR code. When joining, a user automatically gets a random name or they can set their own. Finally, anything you post gets removed after four hours, inactive threads get removed after a week. I decided against Captcha but did implement thread creation/message flood limits.

te_platt• 15 years ago

Makes me wonder what comments here would be like if names weren't shown until a couple of hours posting.

city413• 15 years ago

I made a reddit subreddit that did this. I used CSS to hide names and comment scores (yes people can still see them if they want, it was just a prototype). The overall response was pretty negative.

shii• 15 years ago

Link to said subreddit?

StavrosK2• 15 years ago

I changed everyone's name to "Spartacus" over at f7u12. Nobody seemed to mind much.

athst• 15 years ago

I think it's a cool idea, kinda like LikeALittle but for organizational stuff.

To pre-select the group of people who can participate, you might want to look at a model like Yammer, which validates accounts based on the email domain. This way, you wouldn't have to invite specific emails (which would make me less trustful of the anonymity). If you go on Glassdoor, you see a lot of this dynamic where employees will write reviews meant more for the company than for perspective hires. It's because a lot of companies don't have a really good feedback system where the anonymity can be trusted. Might be a good use case to focus on in the beginning.

ignifero• 15 years ago

This is what scientific journals do with anonymous paper reviews. It works for specific situations but in many cases you can tell who the reviewer is by the content of their review.

gnosis4• 15 years ago

Could someone explain how this is different from a regular web forum, IRC, Usenet, or pretty much any other form of communication that doesn't require you to reveal your real identity?

icegreentea• 15 years ago

In this case, you have full knowledge of the pool of people you are talking to. It wouldn't be a 'regular' forum or IRC, but rather... say you and 5 and your friends all signed into a single IRC channel (so only the 6 of you), but all with randomized names, so you don't know exactly who you're talking to.

As for how the interaction would differ? Depends completely on the people involved.

gnosis4• 15 years ago

"It wouldn't be a 'regular' forum or IRC, but rather... say you and 5 and your friends all signed into a single IRC channel (so only the 6 of you), but all with randomized names"

I still don't see any difference between this and regular IRC.

There's no requirement on regular IRC to use your real name. Make up a random name if you want.

If you're too lazy to pick a random name yourself, you could automate it with something like this:

  #
  # IRCUSER is set to a random dictionary word
  #
  BIG_RANDOM=$(echo $(($RANDOM * 2))) # A number between 0 and (32767 * 2)
                                      # Note, this will not be regularly distributed
                                      # but who cares?
  export IRCUSER=$(head -n $BIG_RANDOM /usr/share/dict/words | tail -n 1)

eavc3• 15 years ago

I love this kind of attempt at innovation. I hope you are able to stick with this enough to find a userbase for it.

dbs119• 15 years ago

Appreciate the support. We're doing our best to get it out there and see how people actually use it. We created it because we think it's important, simple as that. So we definitely plan on sticking around.

eavc3• 15 years ago

Well, best of luck. It's a cool project, and I think marketing is going to be the decisive battle for you guys.

FWIW, I wanted to chime in on the issue raised by others. For use in smaller groups, much of the issue of identifiable style could be obfuscated by instructing the users on using generic sentence structure, word count, rate of input, and synonyms.

Or, you could even have a list of commonly useful expressions, and typing isn't an option or is limited to filling in blanks as with MadLibs. The list of stuff that's hard to say to directly is probably not all that long.

Best of luck.

varikin4• 15 years ago

A forum was recently launched at work which allows anonymous posts and comments. It has been both good and bad.

The good: Open discussions about sensitive topics with fear being singled out. A good example of this is that someone posted about the company doing more to accommodate individuals who are transgender. Obviously, the person who posted it wanted to speak out with fear of stigmas attached. Also, it has been great for openly discussing what we think the company is doing wrong.

The bad: There are some bad seeds that have been very negative about the company, how we work, what we focus on, what tools we use, etc. This in and of itself is not bad, but it isn't always a discussion. Instead, it is more in line with trolling.

Overall, it has been good for us, but we are constantly looking at how to make it better. Like how to promote the good aspects while trying to discourage the bad.

j79• 15 years ago

How does moderation work? I could see the bad seeds, who contribute nothing to a discussion, being filtered out by community moderation? Even taking an approach similar to HN where users who flag down a specific comment to a certain threshold could be less visible.

At least then, the "bad" practices can be discussed without the noise that the trolls bring.

dbs119• 15 years ago

We're planning to implement a rating system for comments.

In terms of participants who contribute nothing to the discussion, keep in mind that whoever starts the conversation chooses who to invite. Presumably, those invited will be relevant in some way to the topic being discussed.

varikin4• 15 years ago

We recently added agree/disagree buttons which works much like voting. There have been talks about more sophisticated things like thresholds and sorting, but it is happening slowly.

bugsy4• 15 years ago

It defeats the whole point of this system if those who criticize things are simply labelled trolls to be considered bad corporate citizens.

Chairman Mao called for a hundred flowers (opinions) to bloom. He invited criticism of the system. At first people said "It's a trap" and no one seriously criticized, only token critiques were offered. Perhaps similar to modern suggestions about policies for washing mugs, what sort of free coffee should be available in the breakroom, or if there should be a third bathroom for transgender. Things that don't represent any serious challenge to the existing order.

Then Mao made clear he really wanted criticism of the system and no one should hold back. So people started talking about what was wrong, and telling the truth about the failures of totalitarian communism. These criticisms were considered "absurd", the people were labelled trolls ("rightists"), rounded up, and executed in the Anti-Rightist Campaign.

If you want honest feedback from people, calling them trolls, with the implicit threat that criticism is wrong and should be eliminated, is not the way to get it.

The only way to have real honest feedback in a small group is to listen sincerely, and eliminate punishment of dissidents. This can only be done through trust, which is developed slowly and with difficulty. 99.99% of the time when corporate says to trust them and it's OK to make suggestions, they are lying and intend to identify and destroy anyone who seems to be a threat to their agenda.

I guarantee that if we collect a list of all the actual "troll" things that were said, most people outside your system will see significant numbers of them as representing valid criticism, and therefore unmask the truth that labeling the criticism "trolls" is an exercise of power to eliminate minority dissent.

alex_c• 15 years ago

That assumes the people giving the feedback are making an honest effort at a serious, constructive discussion.

There are people who are inherently negative and will complain about absolutely anything, justified or not, if there is no moderation from social pressure.

There are also people who don't have their company's best interest at heart, or who just like to stir up trouble for the sake of it.

Many of these cases come very close to the Internet definition of trolling - adopting a controversial position not because you believe it, but because you want to get a reaction out of people. You can't just assume that all criticism is valid.

varikin4• 15 years ago

We want input. I know the executives want feedback. Our company has a good history change through complaints and feedback. The founders were skeptical of version control, a programmer said we need it. They adopted CVS even though the founders thought it was pointless. We moved to SVN a couple years later when another new programmer said CVS was crap. One frontend/designer told the founders the brand and company name was crap. He created a new brand and name. Now he is in charge of the marketing department. We now have a UX department because one designer wanted to try it.

We have changed lots of things about our process and tools because the managers listen to what we, the developers, designers, QA, etc, say and ask for.

As we grow (doubling this year), our biggest challenge is maintaining the communication. Management wants, and I believe they are sincere, to have a way for anyone to complain or point out what we are doing wrong in an open and safe way. Part of that means being able to post anonymously. We once had a great discussion questioning why we are growing and why we focus on the market we do. These are big topics, unlike washing dishes and choice of coffee.

The problem is not people complaining or questioning what we are doing. The problem is someone anonymously saying "x sucks" then not having a discussion. Yes, X does suck, but if the person posting this isn't willing to give examples of why it sucks or suggest how to fix it, I consider it a harmful use of the anonymous post.

drivebyacct22• 15 years ago

This may seem pedantic, but ranting is not the same as trolling. Even loudly complaining isn't really trolling.

qq66• 15 years ago

Is it an internal tool or an off-the-shelf product? Could you tell me the name? I might be interested in deploying that here.

dbs119• 15 years ago

With regards to freeversation, we're looking to implement our anonymous conversation tool in multiple use-scenarios. One of those scenarios would be to facilitate the type of HR employee feedback mentioned above.

varikin4• 15 years ago

In regards to our system, very internal, unless you want add PICK to your list of databases :)

jpr• 15 years ago

How many times does irc need to be re-invented before it's considered fully invented?

elrodeo• 15 years ago

As long as it makes fun :)

Here is my quite private project for friends and family — it works pretty well for us for instant discussions on a concrete topic: http://getch.at

bxc• 15 years ago

real world analogue: http://en.wikipedia.org/wiki/Chatham_House_Rule

owenmarshall• 15 years ago

This reminds me of an anonymous variant SGI & Netscape's bad-attitude newsgroups:

http://everything2.com/title/Bad-Attitude

Failures of those newsgroups were attributed mostly to changes in company culture. I'd be very interested in seeing how this works anonymously.

rgbrgb2• 15 years ago

This is creepy. I'm working word for word on the exact same thing.

rgbrgb2• 15 years ago

Why is this down-voted? I literally worked on this idea for the entire summer. People I had talked to sent me links to this and asked if it was mine. I'm in no way insinuating that this guy stole it but you can probably understand why this feels a little spooky.

rmc• 15 years ago

Great idea. As others have said writing style, grammer, punctuation are the biggest threat to anonymity.

As a tangent I wonder if there's some sort of lossy compression you could do with English that would mangle this enough to make text anonymous. People have suggest google translate a few times, that's using a translation tool to hide some of the original meaning. Has anyone done a dedicated library to mangle text this way?

noodle• 15 years ago

what if a site like this integrated a system that identified an individual's writing patterns and alerted the person to them, prompting changes in text they submit, thus attempting to ensure anonymity?

laglad• 15 years ago

I can see useful brainstorming apps being born from this concept.

uast232• 15 years ago

For the bigger picture of use case mentioned here, what's wrong with IRC. And for the specific use case mentioned in the blog as in School/College, I don't understand how hiding in a closet can relate to productivity. This instead might encourage people to single out someone who is already low and take a dig. As mentioned, people can log in as anonymous but nothing stops you from taking someone's name out there!

dbs119• 15 years ago

The issue of anonymous slander is obviously a big one, certainly not limited to our site alone. Having some aspect of identified group information may in fact limit that type of trolling, but that remains to be seen. Everything can be used irresponsibly. Your first point speaks to the question of anonymity's overall purpose. Wanting to be anonymous isn't always a sign of weakness- it's a reality for a lot of confident, intelligent people. Productivity, in our view, is hampered by the absence of contributions from people who have great ideas to share.

uast232• 15 years ago

Oh yes, the word is "anonymous slander" and I also appreciate the fact that many a times ideas don't come out just because you are too scared. I was just trying to point out the irresponsible use which is kind of bound to happen if given in hands of irresponsible people.

I see that you are a co-founder. So curious about this - did launch of Google circles affect your decision to launch or not to launch? I am asking because Google circles kind of deals with similar group bound problem, though it's not anonymous but very helpful!

jvandenbroeck• 15 years ago

There's a technique to create forecasts based on this concept. The Delphi method http://en.wikipedia.org/wiki/Delphi_method

If I recall it correctly, it's an iterative process: 1. you create a question list 2. every expert answers the questions (anonymous) 3. a secretary processes the questions and makes an overview (+ creates a new question list if there's no consensus)

Like that every expert has equal amount to say, without compromising the identities of people.

The main disadvantage of the Delphi method is that it takes a lot of time. Maybe you can change your concept to "Delphi on the cloud" -- cashing in on company's that want to make accurate forecasts.

windexh8er• 15 years ago

You'd think a service revolving around privacy (i.e. anonymity) would go to greater lengths to use SSL across the entire site. Otherwise, fantastic idea. Can't say I haven't bounced the idea around of the "groupthink" need lately.

nyellin5• 15 years ago

I disagree. SSL is always nice, but end-to-end encryption isn't magic. SSL protects you from targeted man in the middle attacks by governments, isps, and users on the same network. Your friends are neither.

ljlolel• 15 years ago

they can be if they're on the same college LAN

mechanical_fish• 15 years ago

I await the results of this experiment with interest.

My gut feeling is that:

A) To first order this won't really work. You can't really be anonymous in a closed group. In a small closed group you can't be anonymous at all (the degenerate case, two people, is obvious; writing style is an identifier in larger groups) and in a really large group the stakes of letting your mask slip are too high and will squelch serious conversation, while the lure of the big crowd will drive trolls and comedians.

B) To second order this could work great. The second-order advantage of this is the useful social fiction of anonymity. Anonymity doesn't have to be cryptographically strong to be useful. Think: a masked ball. The masks aren't that highly effective - that seven-foot-tall masked woman is pretty easy to ID, mask or no - but if everyone plays along the game can go well. Plus, you can get plausible deniability, which can be very useful in social situations. ("No, it wasn't me who confessed that secret in anonymous chat. It must have been some other native speaker of Finnish from a basketball team in Toledo, Ohio.")

C) But, alas, I'm not sure use case (B) is significant enough to drive a product on its own. Nor that it will be easy to get the site's culture off on the right foot.

A great experiment; I hope it goes well despite my worst fears. We certainly need many more alternatives to the "stand naked in front of the world on the Googleable Internet" model of online conversation.

dbs119• 15 years ago

Thanks for the interest, all good points.

Even in smaller groups though, people may be able to consciously subvert their mannerisms and still get their point across. Not saying it will be easy, but I wouldn't rule it out.

I agree with the plausible deniability point, but i think it can theoretically apply equally to small (and by small i mean at least 10 ppl) and large groups alike. There are many ideas that are not too controversial, but just touchy enough that plausible deniability would be sufficient to overcome the anxiety barrier. This would only work for comments that a user would have no problem being suspected of saying, as long as noone could categorically prove it.

drivebyacct22• 15 years ago

Is everyone anonymous or pseudo-anonymous (My messages are attributed to anonymous35 or something)? If so, as people leave and email/anonymous names disappear you can identify people.

nyellin5• 15 years ago

Freeversation is a forum, not a live chat site, so the names never disappear.

slowpoke• 15 years ago

On a small IRC network I used to frequent, at one time I set up a channel with a bot that lead to a very similar situation. The idea was that channel ops can see people talk even if the channel is muted (+m). Furthermore, on unreal, you can set auditorium mode (+u) which hides all users in that channel from /names and /who. The aforementioned bot was the only op on that channel and was set up to repeat everything said in the channel for everyone to see, effectively anonymizing it. The problem was, this isn't scalable unless you remove flood throttle from the bot (which is almost always a very bad idea) and there's still the possibility of the person running the bot looking at the raw IRC data to find out who is saying what. It was funny while it lasted though.

kragen• 15 years ago

I've used anonymous groupware before, and yes, it was pretty cool.

gwern• 15 years ago

Anonymous chats with verified memberships have a number of interesting applications. For example, I seem to recall Mencius Moldbug discussing them as a mechanism for regime change.

The idea being that in a totalitarian regime like North Korea (or to a lesser extent, the countries of the 'Arab Spring'), it's well known that even the elites are discontented and would support the overthrow of the regime. But there is a knowledge problem, a coordination problem, that they cannot be sure of the support of all the other elites and factions. If they declare their opposition publicly, the still loyal elements will dispose of them. But private declarations which are safe also do not reach enough people to constitute any sort of Schelling point. (This is one of the main points of protests: taking risks, demonstrating solidarity, and a genuine costly signal. If there's sporadic gunfire and suppression from the regime, all the better - scare away the pikers.)

On the other hand, suppose you had a secure anonymous forum for, say, just military officers? With one of the anonymous e-voting schemes integrated to get a headcount of subversive questions like 'how many of us officers would support a regime change?' Then maybe public protests would no longer be the best way to solve the coordination problem for coups...

bugsy4• 15 years ago

That's a very interesting post. The problem is that in such extreme cases as North Korea, if the government is setting up this discussion list or has any access to it there is no possibility that the conversation was really anonymous. If the government doesn't have access, then mere silent participation without informing would be considered treason. Even if there was anonymous speech going on, Jung merely needs to approach each participant one by one and have them identify which statements were theirs. By a process of elimination, you find who the dissidents are, even if they refuse to talk.

raganwald• 15 years ago

If you have something to say but fear that your writing style will "out" you, feel free to ask me to rewrite your message:

http://raganwald.posterous.com/freedom-of-speech

veyron• 15 years ago

Tried, failed miserably. Look up bored at butler.

The general course of events is: Person A will make a comment about person B, person B will complain (under some misguided expectation that the site should be moderated), and you will have to handle the problems. And if you are in a place with a draconian internet policy, you will be paying dearly. Then the site warps to something which loses the original idea of anonymity

pacifika• 15 years ago

I think the anonymous identified group chat idea is really great, but I think that the concept around the experience should be more temporary, to accent the anonymous nature - ie although maybe at any point during the chat you can save out the contents, perhaps it should be more of a chat than a forum and perhaps the discussions should be more throwaway (like removed after 7 days of nonactivity). In my not expertly opinion at the moment the way it comes across for me is a forum/newsletter look but that's not the strength of the system.

Great idea though and I think this would work great as a enterprise tool to help honest discussion where company politics can get in the way sometimes.

Show HN: Can anonymous chat within identified groups create better discussions?

Related Stories