The whole tech community is focused on chatbots right now. This is creating an opportunity for businesses of all sizes to enhance their relationships with their customers.
However, the technology is not yet sophisticated enough to have ‘truly human’ conversations that convey subtle nuances such as humour, wit and sarcasm.
But at Meekan we’ve developed ways to give an illusion of personality and here are my top tips to cheat the unofficial Turing test:
1. Get the basics right
First thing’s first, make sure the robot is doing its job properly. This is imperative. Once you have your robots core functionality up and running you can play around with the cosmetic and conversational touches.
2. Add in conversational touches
Humans, being human, don’t respond quite as predictably as your robot would like. You can prepare your chatbot for every eventuality relating to his task but, as we found, people tend to stray from your carefully thought through engine.
To counter this loop, which for the user will undoubtedly get boring, we added conversational intents to our already established work ones. These included “Greeting”, “Compliment” and “Insult”. Suddenly our robot is interacting with users on a ‘human’ level.
3. Build a response bank
Teaching your robot to give an on-topic response is not worth it if he/she/it will just keep repeating the same phrase over and over again.
This immediately breaks the illusion of sentience and reveals the mechanisms underneath. To counter this, for every intent the robot recognises you need to have prepared a whole list of responses for the robot to choose from. Often we will write 30 possible responses to common questions or intents. It’s a very quick and effective way for the robot to appear intelligent outside of its core functions.
4. Draw from the information around you
Use the basic information you have about your user to your advantage. Little touches like using the user's name, knowing what the weather is like where they are, or even knowing the local time, can go a long way to making your robot seem more knowledgeable.
5. Maintain the illusion of a physical being
Refer to your robot as a robot, not a bot, in conversation with users. Or even as a he or a she.
This helps to maintain the illusion that there is a real life humanoid robot sitting on the other side of the conversation typing away on their futuristic chrome keyboard.
Another little touch that goes a long way is answering long questions line by line instead of in one large body of text. Pacing your answer like this will make it seem like your robot has to obey the physical constraints that humans face (typing with fingers) and less automated.
6. Stay in character
Who is your robot? Develop a character and stick to it. Mixing in different styles of speaking will make your user feel like they are speaking to a random service rep rather than the humanoid robot helper you’ve spent a lot of time developing.