Is It OK to Be Mean to a Chatbot?
wsj.com
wsj.com
So, we already bring non-sentient things into our circle of empathy. You're right that someone out there is going to use our brain wiring against us.
Or, more accurately, no matter how much money is given to them, the same non-sentient video file is played on your TV?
Sure, the image backs to a real reality of some sort. It doesn't matter if that same kid is literally doing the same thing you see, it is certainly a reasonable representation of something that is true. But the video itself is in essence a non-sentient thing that is not responding to you in any way whatsoever as it is used to manipulate you and your beliefs about your agency in the world.
And that is just one example.
It is good to notice the manipulation, but it is also helpful to then leverage that into realizing that this is a continuation of existing trends and you've actually been getting manipulated this way all your life.
In most encounters a "Chatbot" is nothing more than an interface for a business.
If your cell provider is telling you, via a cute human-like chatbot, that it leaked your data or can't remove an unexplained fee, it is pitting your sense of empathy against your rage at the unfair corporation behind the scenes. If you lose your nerve and cuss out the chatbot, they can use that as justification to end the chat and escape all culpability, now leaving the customer feeling both guilty and upset.
So I think a more relevant question is: "Is it okay to use caring personas as an interface to something that doesn't really care?"
Just because you anthropomorphize software, doesn't mean the rest of us have to. All of these questions are ridiculous and the obvious answer is that there's nothing there to be mean to. They aren't things that can perceive and feel your attitude or treatment; there is no "mean" happening.
I had a relative with that issue. They told me the quickest way to get out of the loop and talk to a real human was to become insulting. I do not know if that was the only option but at least it was an efficient options. In that scenario at least you got rewarded for getting angry.
That goes double for chatbots.
I don't say hello, goodbye or thank you to furniture.
I do say those things to waiters.
I don't believe any of that, but I do think that a general attitude of gratitude can be very beneficial to one's happiness and general outlook. IMHO psychologically your approach has benefits.
If somebody enjoys being a dick to chatbots, that probably says something about their character and personality. But double? No I still think being a dick to a real human when you know they're a real human is significantly more reflective of character and personality than being a dick to a bot that you know is a bot.
This reads a bit like "videogames make people violent".
So you think that people who play videogames because it makes them feel superior to the NPCs and as such feel they're allowed to shoot, maim, kill or generally mistreat them are necessarily bad people?
Or are you specifically talking about people who interact with chatbots thinking there's a human operator instead of a bot?
If the latter, I'd argue that the person never knew whether they were talking to a bot or a person, so how they act will depend on what they perceive their interlocutor to be.
No, "bad people" is far too binary and ill-defined. Yes, I do think it says something about their character. Human history shows that we are innately that way. Our entire history is full of evidence from people being dicks to waiterse to wartime atrocities (rapes, pillages, etc) as evidence that it's not just a rare emergence. It's buried inside all of us. IMHO, it's a mark of high character when someone is able to recognize and suppress their animalistic urges.
I can't be either, anymore than I could be a dick to a slab of granite, or to 5 kilograms of oak wood shavings.
And given how most humans are of the opinion that apathy is dickishness, I'm pretty sure I can guess what most of you will think of me. But I'm empirically correct on this issue. You all are experiencing defective cognition. Your species has scaled technologically well past your ability to have sane responses.
Things are going to get bad soon. Then they're going to get worse. And most of you won't even understand why or how.
I'd rather not get used to being rude accidentally.
This includes being nice to animals, children and telesales people.
For the record I think being a dick to a slab of granite is quite possible - given that the being-a-dick-ness is inherent in the person being-a-dick more so than the slab of granite's ability to perceive it.
Why would I be nice to food?
> and telesales people.
Why would you encourage them? Decent people seek to punish them harshly by any legal means.
> For the record I think being a dick to a slab of granite is quite possible - given that the being-a-dick-ness is inherent in the person being-a-dick more so than the slab of granite's ability to perceive it.
I disagree. Dickishness only exists within the interaction of two people. There's no meaningful claim of dickishness for the man alone on the desert island. At best it is a prediction for when he is around other people, but it doesn't even seem like a very good prediction.
> That I could ever be a dick to a chatbot seems to suggest that the only other way of being is that I'm nice to the chatbot.
Why is that? I don't consider being a dick to be binary. You can be anywhere from extremely non-dickish to sort-of-dickish to 12-pound log.
Even if that is possible, it wouldn't change anything. The rest of you seem to have fixated on the idea that anything intelligent/sapient/sentient is what gives it moral standing.
I correctly adopted the position that "human" is what gives a thing its moral standing. I could meet intelligent aliens tomorrow, and they would be no more than bugs to me. I wouldn't try to stomp on them or anything (unwise), but until humanity as a whole negotiated or decided they had the same moral weight as humans, they're nothing to me.
Your confusion on this issue is noted, and I hope that, in time, those confused like yourself will grow up. The chatbot's not Commander Data. You liked him because he was still played by a human actor.
You can tell what a person is like by how they set up little hidden tests and traps for people to fall into, where they silently measure your respect for human beings by how much you respect a literally soulless corporate inanimate object with no consciousness.
You don't need to thank your compiler.
If I can indulge in a bit of what-aboutism to promote discussion, how would you classify animals? Do they deserve respect, and if so, what characteristic qualifies them?
If such a characteristic (e.g. the ability to feel fear/pain) could be programmed into a model, would that be ethical? Would it change the expectations for appropriate treatment of such a model?
I'm genuinely curious about HN's thoughts on this.
It certainly doesn't!
I would look...negatively...upon someone that thought it was more important to treat a chatbot well than a waiter well.
Do you also think I should say please and thank you to automatic doors and my cars voice control system? And if not, why not?
No.
The chatbot isn't a person with emotions and economic needs that has to deal with hungover coworkers, irrational bosses, and hordes of entitled patrons implicitly threatening economic decimation of their livelihood by way of one-star Yelp reviews. Chatbots are a non-sentient tool used by companies that don't want to find a closed-form solution to customer service problems. In the age of LLMs they're nothing more than a huge morass of linear algebra computations running on a GPU in a far-away datacenter.
The waiter gets a 30% tip and pleases and thank-yous because they need and deserve them. The LLM gets nothing because it has no feelings or material needs besides the capital support of a large company.
This isn't an episode of Star Trek. Hell, if you ask ChatGPT...
> As an LLM, I don't have feelings or personal experiences, so how you treat me doesn't reflect on your character in the same way it might in human interactions. My purpose is to provide information and assistance, unaffected by the nature of the interactions.
With scheduling systems especially, reminders might be automated but anything you send back goes straight to a receptionist. That person is going to get to look you in the eye later, so maybe you shouldn't be mean where they can see it.
In my own personal and likely unpopular opinion I think the right question to ask first is "Should one assign personification or anthropomorphism to a chat bot?" The answer I believe the answer is a resounding NO, especially if that bot is not hosted locally and prevented from ever talking to the internet. I will delve even further into conspiracy or spoilers and suggest that this act will put people of all ages and security clearances at risk of volunteering sensitive information to corporations that will self-profile their psychological profile, personal secrets, habits and other potentially dark patterns. Chat bots are the next evolution beyond social programming through social media algorithm manipulation and should be avoided at all costs unless one has the discipline to know consciously and continuously treat the bot as a data-feed into a corporation. What's more is that with time the bot will gain the trust of it's human target and slowly but progressively manipulate the targets behavior from algorithm changes by the bots operators.
So to expand on the question posited by WSJ, the answer depends on what one wants the corporation they are interactive with, the third-party data brokers purchasing the data, questionable organizations and ultimately their government to know about them. Once that data is captured and sold, erasing it would be as feasible as deleting a picture that was shared on anonymous chan boards regardless of statutes.
One's empathy for a chunk of smart sand is disguised apathy towards your fellow humans. If someone feels they must be kind to a bot a corporation is using to interact with them (while reducing costs) then we are opening society up to being emotionally manipulated by these bots. At the very least customers are abused in this situation by not having a real person to talk to and that should not be met with deference, it should be met with contempt.
https://www.thenationalnews.com/world/furious-with-customer-...