So you’re seeing the benefits of introducing a chatbot in your customer support, but have you thought about how human your chatbot should be? Should you introduce a human-like avatar to your bot or keep it to just plain text? These seemingly trivial points can make a big difference to the user experience.
Back in 1970, the Japanese robotics professor, Masahiro Mori coined a term known as the ’uncanny valley.’ The theory works like this; the more human you make your robot, the more people like it – until they don’t. There’s a point at which the robot becomes so humanlike that we think it’s downright creepy. So much so that we’ve seen incidents in recent years with teenagers and middle-aged parents alike smashing robots to pieces in a way that would frighten the fiercest 19th century Luddite.
This rage against the machine even occurs with chatbots. If the chatbot is ’too human’, we try our best to destroy it or make it do things it shouldn’t be doing. In Warsaw’s Kozminski University, the roboticist Aleksandra Przegalinska conducted a test* with her students on two chatbots – one that used an animated human face and also a voice and one that went with just plain text.
The single text chatbot was much less engaging for the user. Frankly, it was dull and straightforward. The heart rate of the people speaking with this chatbot was much lower and stayed consistently low but the chatbot with the animated face and a human-sounding voice created much stronger engagement and even a higher pulse.
So – go for the chatbot with the animated face and voice then. Right? Well. No. Przegalinska’s team did a survey of the students after the test and it turns out that the students hated the animated chatbot and perceived it to be both more stupid and incompetent than the single text bot. The answers given by both chatbots were just the same but the perception was radically different.
The advice from the study then is to think very carefully about a few factors before creating a chatbot for your services. If you’re using voice, then you need to think about tonality. If the chatbot always sounds monotone and uninterested then you, the user, will be wondering why your bot is so bored when you have something exciting to tell.
Another one is flexibility. Should the chatbot agree with you no matter what in a bid to create trust and empathy or should it have some kind of integrity and never bend when it comes to certain values? Do you feel like the chatbot understands you and your issue or is it just saying things it believes you’d want to hear? For a chatbot in customer service, there’s also the balance of making the customer happy without over-committing.
Then there’s the level of transparency. Do we know we’re speaking to a human or a chatbot and are we being clear about it? Most big companies want to hide their chatbots for as long as possible but this may not be advisable in the long term.
And then there’s also the level of anthropomorphism – how human-like it is. We do want it to be human-like because we’re used to interacting with other humans, but again… when it’s too much of a human we’ll end up in this ‘uncanny valley’ squirming with discomfort at our humanoid friend.
The goal is to get your chatbot to cause a positive influence on people and not to be classified as a zombie. If you do it right, you can reduce costs as the bot tackles basic queries and fields the first questions of any discussion. But make it too human and you risk facing a losing battle for credibility.