Don’t make chatbots cute, make them capable
Ever since the dawn of science fiction, we’ve been fascinated by the idea of machines talking back to us. Every small advance is lapped up excitedly; at the recent Mobile World Congress in Barcelona, the unveiling of Sony’s Xperia Agent, a 1ft tall robot with a benign, calming voice, was described as “adorable”, “crazy cute”, and “hard not to love”, while at another stand, Pepper, a robot built by SoftBank Robotics, was enchanting visitors with its cheeky personality, exclaiming that it was “so excited, my processor is racing!” We love this stuff.
Little wonder, then, that so much effort is put into constructing the “voice” of chatbots. The idea is to make them likeable, to give them human qualities. But let’s be clear: broadly speaking, the technology isn’t yet up to the task. Interactions can be human-like, but they’re a long way off being human. Yes, Apple’s Siri was proudly described by its creators has having “a light attitude… friendly and humble, with an edge”, while Microsoft’s Cortana was designed to be “competent, caring, confident and loyal, always prepared to help, not bossy, eager to learn”. If we took away their names and quirks and just called them assistants, would people still try to challenge them?
But when it comes to chatbots these AI creations have raised expectations and are selling us a false dream. As chatbot use becomes more widespread, the temptation will be for brands (who, understandably, are very keen to maintain their tone of voice) to try and make them human-like, to emulate the Siris and Cortanas of this world. But that’s where we stray into dangerous water. The minute bots try to be human, that’s where they start to fall short of our expectations, and that’s where they start to fail.
An enthusiastic “Hello, can I help you?” might be welcome from a real human, but the same line delivered as a pre-programmed greeting by a snippet of code doesn’t sound remotely friendly. It’s trying too hard. Yes, businesses have to take care over their messages, but the focus with chatbots should be on accuracy and efficiency, not than superficial gloss. This shouldn’t about trying to pass the Turing test, trying to convince users that automated messages are written by humans; it should be about adding value, creating positive interactions and minimising the likelihood of annoying people. The chatbot world is only in its infancy, and right now that’s the best way to create affection and to create trust; not by buttering up users with matey chit-chat, but by solving their problems in a coherent, methodical way.
In the future, the question of how a business should chat will become more important. Ultimately we’ll find ourselves drawn to certain brands because of the tone they use – but as AI develops, businesses and services will be able to have more than just one tone; personalisation will be the key, adapting to the voices of its users, making judgements about how we want to be addressed, being friendly with some people and formal with others. But let’s set that aside for now. Users have to understand what the emerging chatbots are capable of, and what their limits are. If you want a philosophical conversation, a chabot isn’t currently the best place to go. But if you know that it can retrieve a recipe, alert you to late trains or remind you to buy milk for breakfast, and they deliver on that promise, their reputation will continue to grow. And we’ll want to use them.