The Evolution of Etiquette: Why Language Matters in Conversational AI
Language is complex.
It is messy, nuanced and has the power to make or break relationships, declare love or war, create myth, religion and inspire masses. Throughout the course of human history, it emerges as the thread that has propelled us to survive and develop to where we are today. Most people have the ability to communicate, be it through speech or sign, yet it is one of the things we think the least about — at least consciously.
Language Evolution
It’s worth stepping back in time a moment to speculate on the origins of language. There are a plethora of theories, from language evolving as a means to share information about the world, i.e., ‘I saw a lion by the creek yesterday morning’; as well as share information about other people, otherwise known as gossip.
Throughout human evolution, language has been the edge that has enabled us to organise, share and develop ideas; from cave-painting and stone-axes to the awe-inspiring technologies we experience today.
As we traverse this new digital age, language and conversation emerge yet again at the forefront of change. In the efforts to replicate human conversation, Conversational AI is faced with the complexity of encoding human language and crafting natural conversations, while LLMs, (Large Language Models) are taking things to the next level by drawing on vast data sets to impressive effect.
So what is it that makes up human conversation?
Grice’s (1975) maxims of cooperation — quality, quantity, relation and manner — are the famous foundations of conversation that are often used as a cornerstone for Conversation Design. If we were to only follow these maxims, it would be a very robotic conversation indeed, as there is much more to conversation than being truthful, relevant, informative and clear.
Humans are simple creatures that are happiest when we feel understood and respected, and it’s the subtle cues of engagement and kindness that cultivate this sense of emotional connection.
Conversational Saviours
Politeness and empathy are the true quiet achievers of conversation. However, these conversational norms can vary from person to person and across cultures. Politeness strategies are one of the main differences that can cause conversational trouble, as we balance our conflicting needs for feeling involved as well as independent, and work to express our needs in a way that causes the least offence possible. A person who favours directness will often struggle to understand another’s indirectness; not to mention metamessages — not what is said, but how it’s said — that can convey something else entirely. Misunderstandings are one of the most common causes of arguments, and it is not surprising, considering the many factors at play in every interaction.
Unpacking these elements, we begin to see the complex and mysterious workings of conversation. When we apply these factors to Conversational AI, we are as Cathy Pearl says, ‘teaching robots to talk like humans, instead of the other way around’.
Keeping language at the forefront of design, (not as a mere afterthought) with awareness of the nuanced and multi-faceted nature of human conversation will ensure we are designing conversations that get people talking.
