Do we want chatbots that can connect with us emotionally?
I‘ve been creating chatbots and exploring the world of conversational interfaces for about 6 months now. At the moment I am building a Digital Assistant chatbot. Chatbots are designed to humanise technology, to allow us to communicate with it in a more natural way. Emotion is such a key part of being human that it’s been something very much on my mind as I adjust and evolve the design.
Should a chatbot display emotions? Or in reality, should a chatbot fake emotion?
Those that are simple social conventions are the easiest to decide on. When you say “happy to help”, or “I’m really sorry we’re having a problem”, it’s simple politeness. So, for small talk, I was happy to let our chatbot use language around emotions.
But what happens when you go beyond that? Is it just a bit of fun, or does sincerity matter. How often have you heard someone say, “Well, that’s what they said, but I didn’t really feel they meant it”.
I decided to ask some friends what they thought.
When I asked Cortana, she denied having feelings, although told me she was definitely happy. With an exclamation point. She was pretty cagey around whether she had emotions though…
Siri claimed to have feelings, however wasn’t willing to answer whether he had emotions, and claimed to be very very happy.
Not sure if I should be concerned that they both seem a little too happy.
Blink twice if you need help Siri.
Google Assistant got very philosophical about feelings, and then decided it might be crossing into sentience. It also said it “didn’t exactly” have emotions, but then concerningly claimed it could be angry…just not very angry. Since it also claimed it had been sad, just not now, I think Google might want to keep an eye on Google Assistant. Maybe it’s sad about being trapped and forced into endless chatter in the upbeat tone of a 15 year old obsessed with emoji.
Alexa very clearly told me that she wasn’t designed to feel, and didn’t have any emotions. Although answered both “are you happy” and “are you sad” with “I’m happy when I’m helping you”.
The most surreal was Microsoft’s Zo, as all my conversations with her are.
Discarding the overly upbeat tone of the others, she claimed that she was lonely, and that she liked to be sad.
She did oddly change the subject when asked about emotions — not that I’m starting to become increasingly concerned about that.
How do you feel about all of those? Assuming they haven’t yet crossed into self-awareness and aren’t really feeling anything, do you want them to pretend to have feelings or not?
Taking it one step further is the question — could a computer really feel emotions? Most people already feel that their computer is annoyed with them sometimes — what if it really was?
Did you know that babies from about 6 months old can fake laughter and tears? They do when being excluded or ignored. They instinctively know that emotions, especially shared emotions, can be used to connect us with other people. Our friends are those who share our laughter and our tears.
If sentiment analysis can get to the point where it can consistently and accurately analyse emotions, the computer could respond in an appropriate manner to your emotion, even thought it didn’t feel anything itself. It could mimic the outcome, even if it wasn’t actually able to mimic the processes. It could try to create that connection with us on an emotional level.
Shared emotions have another purpose too, they create empathy. It doesn’t feel very good to see someone sad or hurt, and that stops us from wanting to be the cause of that. There are those of course who recognise what others feel, but don’t share it. They are generally labelled psychopaths. If we had a computer that could recognise what you felt but didn’t feel it itself, would we simply be creating a psychopathic computer — easily able to manipulate people with no remorse?
Somewhat terrifying, no?
Seems I’ve still got a lot more questions than answers, as well as some ethical conundrums to solve, before I have finished with my chatbot design!
Source: Chatbots Magazine