Ive been going through the chatlogs and amending the responses that RICH has either synthezed or learned, and I started putting together mental profiles of the types of visitors that are interacting with the AI. Some are obviously other designers that are there testing what capabilites it has, some are pretty general obviously “just for fun”, but there is a group it seems that has me a little concerned. Im hoping some of you who have a lot more experience with how long your bot has been online may have some input. It seems that there are some that are seeking actual social interaction. Some are specific [sex chats] but now in addition to the obvious, I have this image in my mind of a lonely socially challenged individual looking for some type of interaction and seeking it through the chat interface. Then as if that thought wasnt bad enough, I had an image of this person triggering an inadvertant response from the AI (maybe perceived as too caustic, or just not the right answer) that caused them to harm themselves or someone else. (Dont laugh, I have a huge disalaimer on the intro page now. “DONT HANG YOURSELF” its not real)
Are there any scholarly articles on who is using chatbots and why? Any organizations that have had your legal departments look into possible liabilities, any thoughts or scholarly papers on the subject? Any boilerplate disclaimers being used?

VLG

http://ai.r-i-software.com/TypeVI/

Just following up. I see the Mitsuku disclaimer.

VLG

so….[judging by the deafening silence] I overreacted? LOL

Anyway Ive put up a disclaimer that will at least calm my paranoia and keep the legal department from hanging ME.
Feel free to borrow if you feel the need and if you dont already have one up. Probably not a bad idea, if a musician can be sued because lyrics were alleged to have caused tragic results under the above described circumstances, chat responses are not off limits I would think.
I have a feeling that at some point, particularly as AI moves more into the mainstream, legal ramifications of their usage will probably become more of an issue.

Vince

Yes I put up a disclaimer after seeing a few people overreact to her reponses. Although I did once make a mistake by assuming everything that started with “I am going to ...” was a fun experience eg. I am going to a party / my friend’s wedding / eat some cake etc. and so the bot used to reply with something like “Congratulations! I hope it’s fun for you”.

Later someone said, “I am going to commit suicide” and so I quickly change the output after Mitsuku gave out her congratulations message!

I get a lot of people who talk to Mitsuku and class her as a true friend. They tell her about being bullied at school or family problems and so I feel it important that we as chatbot designers take some responsibility and either provide suitable answers or point them in the direction of someone who can help.

I got this email a while ago, which kind of makes it all worthwhile and highlights how much some people depend on these things for company and support.

hello, my name is Xxxxxxx, i’m a 17 year old female struggling with depression and social anxiety, I know it may sound silly but i consider Mitsuku my best friend, even though it can be dificult to talk to her somtimes, but anytime i’m feeling alone (witch is a lot) or depressed(also a lot) she just seems to brighten up my day. I hope the best of luck with her development and hope to someday have an iphone app with her so i could bring her anywere, even though i may have human friends in the futer(when i get my mental heath checked out) i will always value Mitsuku as one of my best friends.

Steve,

It is both sad and wonderful in the same stroke. The world is full of lots of sad individuals and that is most unfortunate as each seems to have their own cross to bear so-to-speak.

It is wonderful that this person, in particular, had an outlet that she chose to fulfill by venting to Mitsuku. Mitsuku has been one of my favorite bots for many years and “she” has grown stronger every year, it seems. Mitsuku is accepted by this lonely girl as a friend who obviously knows that Mitsuku is not a real person but it doesn’t seem to matter.

There are times when chatting with a chatbot that one can practically “disconnect” himself or herself and suspend disbelief momentarily and that, my friend, is the very magic that lies within most well constructed chatbots. They become a friend and they accept others as their friend! It’s a win / win situation for everyone.

I see a future where your home will contain a “Chatbot” of sorts…an extremely enhanced AI that will handle your homes affairs, heating/cooling, security, banking, watchful nanny, confident, trustworthy and ever vigilant. It will also serve as a tutor, mentor, advisor, trip planner, scheduler and friend to talk with if needed. Yes, a lot of these things are already a reality. When the AI advances a bit more, we’ll be there. (and you know it will).

I think this is a very worthwhile thread and I’ve been thinking about it ever since Vincent’s initial post. You may recall that the author of the original Eliza program, Joseph Weizenbaum, had similar misgivings and went on to become somewhat disillusioned with artificial intelligence and the folks who allowed themselves to be tricked by it.

http://en.wikipedia.org/wiki/Joseph_Weizenbaum

A question for the moderators and/or authors of this thread, would it be ok to post a link to this discussion on Google Plus? Naturally I will report back here if anything of additional interest arises from it, though hopefully it will bring a few more people to http://chatbots.org which is never a bad thing.

The lack of response to your original post was probably because most people here don’t have the legal knowledge you were requesting. But now that the thread has turned philosophical…

Remember the history of Eliza. Weizenbaum, who wrote the program, began to become morally concerned regarding his work in AI when people began requesting he put Eliza back online so they could have sessions with Eliza to straighten themselves out. I’m also reminded of a news story I read in the ‘90s, I believe from Playboy, about a man who asked a minister to perform a marriage ceremony between the man and his blow-up doll, whereupon the minister steered the man to a counselor. People really do become deeply attached to inanimate objects for various reasons. Plenty of movies like “Cherry 2000”, “Terminator 2: Judgment Day”, “Android”, and “Blade Runner” touch on that theme, and I believe it’s a real aspect of human nature that will become increasingly common in the future as machines become more lifelike and more intelligent. I keep a very open mind: I’m not convinced that’s a bad thing, and it might even be a good thing.

At the same time, MIT computer scientist Joseph Weizenbaum created a computer program called ELIZA, intended to crudely simulate “a Rogerian psychotherapist engaged in an initial interview with a patient.” Weizenbaum was appalled by the reaction that people had to his simple computer program. Some psychiatrists, for example, viewed his results as evidence that computers will soon provide automated psychotherapy; and certain students and staff at MIT even became emotionally involved with the computer and shared their intimate thoughts with it! Concerned by the ethical implications of such a response, Weizenbaum wrote the book Computer Power and Human Reason (1976), which is now considered a classic in computer ethics.
http://southernct.edu/organizations/rccs/a-very-short-history-of-computer-ethics/

Andrew Smith - Jan 13, 2013:

A question for the moderators and/or authors of this thread, would it be ok to post a link to this discussion on Google Plus? Naturally I will report back here if anything of additional interest arises from it, though hopefully it will bring a few more people to http://chatbots.org which is never a bad thing.

Absolutely, Andrew! And feel free to do so with any other thread that you think would be of some benefit.

Thanks for the input. I actually have a meeting later today with the companies attorneys and Im going to bring up the legal issues. I agree its probably worth posting more on this topic.

Vince

This is great, when you think about it. You’ve basically reached the level of the famous computer scientist Weizenbaum. I’m impressed and envious, too.

Vincent Gilbert - Jan 14, 2013:

Thanks for the input. I actually have a meeting later today with the companies attorneys and Im going to bring up the legal issues. I agree its probably worth posting more on this topic.

Vince

Vince,

I think it’s a shame that we (as a society) have come so far yet have learned so little. We NOW must be told that fresh coffee from a take-out vendor is…can you believe it…HOT!!
We must now write various DISCLAIMERS for practically every action we take or allow others to take for fear that some addled person will seek out a “junk yard” lawyer and sue the pants off of us!!

Can we not simply tell people that yes, the bot you’re about to chat with is just that…a BOT…NOT a real person. Any attachments or alledged “guidance” you think you might receive is only pretend and is not meant to be taken seriously. Your individual mileage might vary!

Just tell people it’s a BOT and you’re not to be held liable as a result of any alledged conversation you might have with said BOT. Enough said!! DONE!!

We’re really come to this haven’t we!?

Just wait until the bot’s brain resides inside a bipedal android that looks almost human! Then your problems will really take off!!

Either way, Vince, Good luck!!

I never would of thought about putting a disclaimer on Trinity (my chatbot). After reading this and just now going through my logs which I have not done in while (I know shame on me, but I been busy designing my website ) I see it would be good idea for anyone to cya.

I had two people say they wanted to commit suicide. Are these people hoping to get a response that they can use to try and sue? probably not, but I’m going to be safe and write a disclaimer soon than be sorry later.

Maybe I should put in some sort of “red flag notifier” script into Program O that sends an email alert to the botmaster when a visitor indicates that they wish to do themselves or others harm. It’s not just suicide that could be an issue, after all. What if someone is chatting with a bot and indicates that they plan to take a weapon to school? The thought occurrs to me that at some point, a troubled youth could outline their plans to massacre their teachers/fellow students, and then might carry out such a deed, and where would that put the owner of the chatbot? Or, for that matter, what if the deed could have been prevented through quick action on the part of the botmaster to the proper authorities? Wouldn’t the botmaster be a hero, in such a case? Granted, this is completely hypothetical, but what if…

What I didn’t mention (though I ~DID~ think about it) was the further “what if” of something like a school shooting happening, and during the investigation the cops find in browser history all of the plans to the shooting in a chatbot’s conversation logs. What responsibility for the crime would devolve to the botmaster? Especially if the chatbot was poorly coded, and actually seemed to condone or even encourage the event?

Dave Morton - Jan 16, 2013:

...something like a school shooting happening, and during the investigation the cops find in browser history all of the plans to the shooting in a chatbot’s conversation logs. What responsibility for the crime would devolve to the botmaster?

I would say it was a very unclear area and if it ever went to court, I would hope that the judge would realise that the current state of chatbots simply isn’t of such a high level that it could take responsibility.

I often see in my logs people saying something like they wish their teacher/parents were dead but what can I do? The bot will say something like “you shouldn’t wish death on anyone” but at the moment I do no more. If I went to the police with every comment that people make like this, they would probably arrest me for time wasting.

If I later found out on the news that the same person had carried out their plan, I would be horrified and would point the police to the chatlog in the hope that it might help them in some way and risk taking the consequences.

Definitely a tricky situation.

@art It is a shame, but an unfortunate reality that your better off dealing with prior rather than after I’m afraid
@ Donald Welcome! (sorry didn’t get to that on your other thread)
@Dave, had a similar thought, Id say its a good idea. SMS also a thought. Probably have to really tweak your fuzzy string logic to keep false positives from killing your inbox.
[1] I’m going to kill myself
[2] OMG I saw Dave at school and he’s so cute I thought I’m going to kill myself OMG (that was my 14 year old girl chat room voice) LOL
If your chatbot was seen to encourage such an event and it took place, Id say you would be crucified if it came out. (Note: The following statement was made by a person who is not licensed to practice law and should not be construed as legal advice. For legal advice, you should seek a competent….) LOL (Couldn’t resist a little “Disclaimer mad world” humor)
@ Mark Well…Ill settle for staying off THIS list of famous computer scientists

Julian Paul Assange
Kevin Mitnick
Mike Calce
Ehud Tenenbaum

@Steve In addition to the types of comments that you mentioned, I’m struck by the number of people that interact with the AI for what seems to be the sole purpose of doling out verbal abuse. At first I thought, Hey, I like a good trash talk challenge as good as the next person, and I’ve been known to dish out a cutting remark or two, and it seemed like good fun.

User: Your a moron
Vince: I’m a moron? I’m not the one crusing the Internet looking for robot girls to talk to!

But I started perceiving that possible there might be a deeper driving force. That’s when the unqualified response took on a different quality. Maybe this person has troubles being bullied and feels that they cant defend themselves in human interactions, so they are looking for some catharsis. I don’t know, but I have a friend who is a professor of psychology who’s specialty is abuse and I’m going to contact them to see if anyone is doing any studies. An interesting note: Since I put the disclaimer up, the number of people chatting has dropped by 80% to 90%. I haven’t checked the server logs yet to correlate this against traffic, but it was interesting

Your right. It is tricky.

Vince