5 Insights Watching Humans Interact with Robots Can Reveal About Chatbots
I recently attended CogX 2018, a conference in London ran by CognitionX — a news platform for all things Artificial Intelligence. At this conference, I got the opportunity to speak to a lot of people and, more importantly, a lot of robots! As a geeky girl from Northern Ireland who’s never had the chance to see such cool things, I was immediately enthralled.
I’m an R&D developer so I work a lot with machine learning and recently my focus has been on chatbots. What I found interesting was observing how people reacted to a robot, including my own behaviour. What I observed in the interactions of people with real physical bolts and screws robots can give some insights into how we can spark a connection between a human and a chatbot.
Natural Reactions 🤝
One thing all of these robots had in common was their ability to track your face with their eyes, tricking you into feeling a physical connection with a lifeless being. Robothespian followed me automatically with its eyes, Pepper kept an eye on me when it moved out of my way, Sophia intently watched my face as I spoke while also looking to the side when she was speaking — a very natural human behaviour.
Being able to adapt and mirror a person in a conversation is easy for most humans to do, it helps us connect with each other, and imitating this in a robot seems obvious.
It’s hard to do this in the form of a chatbot, but Assurion have managed it with their Advanced Virtual Assistant aka Ava.
Ava is able to read your facial expressions and tone of voice and react accordingly using a device’s webcam.
A more simple solution might be to use sentiment analysis to gauge the tone of the conversation and adapt a chatbot’s responses accordingly. However, this is definitely going to be a challenge chatbots will face for some time as there are just too many social nuances to consider.
Humour 😂
Who doesn’t love a good joke? It’s a great ice breaker and can even help spark a connection. Robothespian was programmed to sing songs and give famous movie quotes — from C3P0 to Terminator. Introducing a joke at an appropriate time to your chatbot could ease the conversation and leave a more positive experience better than any website could if done well.
Not to mention the fact that I overheard a number of people at the conference telling their friends and co-workers about this one joke that a robot had just told them. A great way to get people to come back and bring their friends with them!
Uncanny Valley 🤖👩🏻
It’s got to be mentioned that Sophia is creepy. Her developers have clearly worked really hard to implement even micro-movements in her facial expressions but we can sense there’s something just a tad robotic and off but because she looks so human, it’s impossible to shake off those slightly off aspects. Meanwhile, MiRo and Robothespian look delightfully robotic and it adds to their charm — people aren’t creeped out by these robots despite their cartoonish features and whacky movements.
After witnessing the uncanny valley in action, I’m inclined to believe that an overtly non-human chatbot would be much easier for a human to connect with than a chatbot that’s trying too hard to sound and appear as a human.
Novelty 🤩
There’s a lot of value in novelty — it’s a chance to win people over! Every single person I saw interacting with a robot was excited by the experience. It may seem crazy to suggest that a customer service chatbot replacing a phone call could be novel and fun, but you could certainly help by changing how you frame it, for example as a means of speeding up a service rather than the only option.
Humans are Naive and Easily Fooled 🤯
I was amazed by Robothespian, thinking it must be the best robot ever created — it had a response to every single thing said and politely waited for a person to finish before saying its part, waving its arms about like a real human does while speaking. I was so enthralled and entertained, I was easily fooled into thinking it was managing all this on the fly, never considering there was a person 300 miles away controlling its speech. Which there was, of course.
I saw a robot, an entertaining robot at that, and I believed what I saw. Importantly, when a person sees a message from a chatbot all they see is the instant messaging channel and the chatbot’s messages — a channel used by all their real human friends and family. A channel strongly associated with human connection. It’s going to be their first instinct to think that who is speaking to them is a human because that’s what they typically see in this “environment”. And it’s not a fun experience to be fooled and feel like you’ve been tricked. The controversy around Google Duplex is a fine example of this.
Do not use their human traits against your users. Be upfront and you’ll create a more positive experience.
