Have you ever wondered whether communication between humans is based on some principles? Do we just talk because we learn how to and are meant to, or do we follow – subconsciously – the same old patterns that are species-typical behavior?
If we want AI to have more human conversations, we have to understand how human conversations work, and that is why in conversation design, we have turned to Paul Grice’s cooperative principle.
“…The cooperative principle describes how people achieve effective conversational communication in common social situations—that is, how listeners and speakers act cooperatively and mutually accept one another to be understood in a particular way.” (Source: Wikipedia)
The cooperative principle, like the wording suggests, is based on the assumption that humans will be cooperative in a conversation: a person will infer that the other person will cooperate and that each will make their own conversation contribution.
This principle can guide us to understand how we usually behave in conversations so that we can design AI-based conversations that sound more natural.
It is important to understand that we don’t want to trick people into believing they are talking to another person. This is not what we learn in conversation design and it’s not what we do. Your goal, if you want to add a chatbot to your website, should not be to make your clients believe they are chatting with a human. Your goal should be to help your customers achieve whatever it is they want to achieve, with no frustration, no friction, and actually enjoying the process.
If a conversation sounds natural, the person will enjoy it, even if she knows she’s talking to a chatbot.
The cooperative principle is composed of four maxims: the maxim of quality, the maxim of quantity, the maxim of relevance and the maxim of manner.
As Paul Griced said, “Make your contribution such as is required, at the stage at which it occurs, by the accepted purpose or direction of the talk exchange in which you are engaged.”
Before diving into the 4 maxims and some examples, I invite you to ponder on your conversations and even find examples of conversations that have been frustrating or a complete disaster. You will see that the reason is that one or several maxims have been violated.
The Maxim of Quality
Try to make your contribution one that is true.
Don’t say what you believe to be false.
Don’t say things for which you lack evidence.
This maxim applies to the quality of the information we are sharing in a conversation. I think it’s pretty self-explanatory and if you are looking for an example, think about all the fake news around: sadly, nowadays this principle is being constantly violated on the Internet.
The Maxim of Quantity
Say no less than the conversation requires.
Say no more than the conversation requires.
I think we all know that person that when asked a simple question like “Have you seen Charles today?“, starts telling the Odyssey. Maybe sometimes we are that person. Well, every time we overcharge our story with useless details, that don’t contribute to the conversation, we are violating the maxim of quantity and being uncooperative.
For instance, if you are shopping and the assistant asks you “In which color would you like the shirt?” and you answer “Pink, because pink is my favorite color. I used to like red more, but then one day…” I think you understood: every word that follows the word pink is not necessary to the conversation and is violating the maxim of quantity.
In conversation design, it is important to teach AI that humans will likely violate this maxim and add more information than required, and design accordingly.
The Maxim of Relevance
Be relevant.
This maxim refers to the relevance of the relation between two parts of a conversation. In particular, how relevant it’s an answer to a question.
Here’s an example:
John: Where is my phone?
Cathy: It’s in your room.
OR
John: Where is my phone?
Cathy: The maiden has cleaned your room this morning.
In the first example, Cathy’s answer is relevant and direct, therefore follows the maxim of relevance. In the second example, Cathy’s answer is apparently irrelevant. However, John might assume that Cathy is being cooperative, and will, therefore, infer that she is implying something that is relevant to the conversation. For instance, the implicatures could be that the maiden might have placed to phone somewhere else, or that he should ask the maiden, or, worst-case scenario, that she might have stolen it.
The Maxim of Manner
Don’t be obscure.
Don’t be ambiguous.
Be brief.
Be orderly.
In other words, speak clearly, avoiding jargon and ambiguity. When applied to a conversation with a chatbot, this maxim refers to when the prompt is too “robotic” or uses wording such as ” Unable to complete the transaction. Please try again”. A cooperative person would never speak like this. A natural alternative would be “Sorry, there was an issue with your card. Do you want to try again or use another card?”
Learning more about the cooperative principle has been life-changing for me. I’ve realized how many times I was uncooperative in a conversation, especially by adding too many unrequired and unnecessary details.
Have you pondered on your conversations? Do you have examples of conversations that violated the 4 maxims? Please share them with us, we might all learn something new!
Linguist and conversation designer, with a background of translator and content writer. An entrepreneur at heart, I founded and closed a virtual assistance agency and I’m now devoted 100% to conversation design.
I’m an empath, I’m passionate about communication and behavioral psychology. My greatest ambition is to contribute to improving the interaction between humans and computers so that machines can be at our service and make our life better.
If you need help with the design of your chatbot, just contact me!