Understanding chatbot conversation design


Understanding chatbot conversation design


Text-based messaging has quickly become one of our most popular methods of communication. In the UK, adults receive an average of 93 messages or notifications every day, with 18-24 year olds receiving three times that. Text-based conversations offer similar immersion and emotional feedback to traditional conversations, and there’s an added element of always feeling connected. As a result, we’re motivated to keep using messaging services. It’s logical for businesses to take advantage of this by incorporating chat features into their customer service. For it to work well, it’s important for chatbots to simulate natural language. Conversations are fluid and unpredictable, with subtle layers of emotion and context. An obvious oxymoron arises – how can we design a natural, free-flowing conversation? The goal of artificial intelligence (AI) is to replicate human interaction – not only the language itself but also the nuance, subtlety and implied emotions. This is where chatbot conversation design comes in.

Three Pillars of Conversation Design

We have identified three key pillars of conversation that need to be considered when developing chatbots.

Cooperative Principle

A term coined in 1975 by philosopher Paul Grice, the cooperative principle dictates that effective communication requires cooperation between all parties involved. Not only must we cooperate in the exchange, but we also expect cooperation from the other parties. It’s a concept that’s rarely articulated but underlies virtually all communication. A conversation requires more than literally answering questions; it also involves understanding a person’s intent and ‘meeting in the middle’. For example, let’s say you are asked, “do you know what the weather will be like tomorrow?” and you simply reply with “yes”. While the exchange is linguistically sound, it misses the question’s underlying subtext. The implication is that the asker wants the responder to explain the weather forecast, rather than answer the question in a literal ‘yes-no’ manner. This concept is known as ‘conversational implicature’. Grice also posited four maxims that must be followed in order for a conversation to feel like a conversation. - The Maxim of Quantity dictates that a person only gives the amount of information required to further the purpose of the conversation. Providing too much or too little information can halt a conversation. - The Maxim of Quality says that participants in a successful conversation are honest and provide accurate information. - The Maxim of Relevance states that contributions to a conversation must be relevant to the topic. - The Maxim of Manner says that successful conversations happen when participants are clear with their words and phrasing. It’s common for some chatbots to break these maxims, which adds to the feeling of an unnatural or unhelpful conversation. A chatbot may overload a user with information, provide outdated and irrelevant information, or give ambiguous responses. These maxims must be respected when designing a chatbot conversation in order for it to be natural and free-flowing. People will subconsciously expect these unwritten dynamics to be followed, even if they are fully aware that they are talking to a machine.

Turn Taking

It should come as no surprise that conversations require participants to take turns ‘speaking’. Face-to-face communication involves a number of subtle cues and signals that one party has finished speaking and the other party should respond. These cues ensure a sense of balance that is key to a successful conversation. Chatbots cannot pick up on these cues, so it’s important that they are programmed to await responses and not overload the user. A common way that chatbots can signal the end of their turn is to ask questions, which naturally prompt the user to respond. This allows the participants to take turns organically, similar to a traditional conversation.

Context

Context is vital for all conversations, and chatbots must respect the context in which they are being used. Certain conversation styles will suit some contexts more than others, and this needs to be taken into consideration at the design stage. For instance, a jovial, upbeat conversation style will suit a party-planning company chatbot but would be highly inappropriate for a funeral director. The device someone is using must be taken into consideration too; it’s appropriate for chatbots to be more direct with a user on a mobile device since they are much more likely to be on the move. Also, for life-and-death subjects such as medicine or finance, being clear is of the utmost importance, and a more factual approach is preferred to using superfluous fillers.

AI chatbot development

Creating convincing, natural-sounding chatbots is a complex undertaking, but artificial intelligence allows chatbots to achieve a level of sophistication that cannot be reached by rule-based chatbots. Natural Language Processing (NLP) and machine learning enable chatbots to offer users more freedom in conversations, thus creating a more authentic experience. NLP is the field that focuses on the interactions between humans and machines via language. A machine that communicates with a human using natural language must not only recognise and understand what a human is saying, but also be able to generate appropriate responses in real time. By exposing a system to vast amounts of language data, it can learn the meanings and intentions behind our conversations. AI chatbots recognise a wider variety of words and phrasing options, whereas previously, bots would be hindered by more rigid communication styles. For example, a common practice with rule-based chatbots is to have the user select answers to multiple-choice questions at each stage of a conversation, in comparison to AI chatbots that can accommodate more open-ended questions and answers. The combination of chatbot conversation design with artificial intelligence is the key to creating useful, valuable chatbots that constantly improve. Here at We Build Bots, our AI customer service chatbots offer the very best service while simultaneously freeing up your time. Invest in the future of your business and use the power of AI to take your customer service to the next level. View our other resources for more information.