There is no need to remind you how the chatbot business has been boosted recently and how much money has been spent in this field. There are so many articles highlighting the benefits of using chatbot in customer relationship services, human resource management, marketing, etc.
I have been working in the chatbot industry for more than one year and a half. I had access to the history of all interactions between users and chatbots in numerous fields: insurance, tourism, helpdesk service, etc. And every time, I was impressed by the way the bot understands perfectly the language used by the users and answers in a “human” way. Sometimes it even has a humanized appearance and a synthetically voice very close to human voices.
However, one question remains unanswered: why users prefer to switch to a human agent (when it is possible) rather than continue using chatbots specially when their request is urgent?
Indeed, a French survey published in http://www.mediaschool.eu/ showed that 40% of bots users think that the bot’s technology is not yet mature.
In my opinion, one of the predominant reasons is that users don’t really trust bots. They consider that understanding natural language is not enough. The bot has to capture some other “human” aspects in the interaction such as “emergency”, “fear” or “dissatisfaction”.
In the common-sense, those notions are defined as human mental states. Sometimes, bots’ users don’t even express it explicitly. In the case of vocal or facial interaction, there are tools which are able to capture those emotional information from prosody or facial expression and transmit it to the bot . But in textual interaction, it is more challenging.
In our team, we wanted to focus on this issue and prove that bots are capable to adapt themselves to the interaction’s context and the user’s states.
Our approach is based on psychological findings. We wanted our bot to be empathetic, expressive and socially acceptable. Let’s make all this clear:
An empathetic bot should be able to “feel” what the user is feeling. Thus, it is necessary to build a model of the user: his/her desires, fears, beliefs and social norms. Those components are inferred from the interaction context (commercial context, tutoring context, etc.) and dynamically from the interaction. The natural language processing tools make it possible to extract those information from the utterance.
A human-bot interaction has also an emotional aspect. Bots should be able to express their own beliefs, desires and feelings in addition to empathetic feelings. To make users feel in a real conversation, it is important to see real emotional responses of their interlocutor and feel responses’ intensity. In a textual interaction, it can be illustrated as follow:
Positive emotion with low intensity: This pizza is not bad.
Positive emotion with high intensity: I really recommend this pizza! All comments about it are positive. You should try as it is spicy as your favourites!
In the case of humanized appearance, it is necessary to add facial expressions and to display it with the adequate intensity.
Social acceptability is one notion that we learn through our life. It can be defined as the regulation of emotional responses to fit into social norms.
As the bot is talking in natural language and performing a task, we can consider the interaction as a social one. Thus bots should also be acceptable during the conversation. They should choose their emotional responses in accordance to their social context. If the bot is an insurance expert, its responses should be polite, serious and precise.
Talking with an empathetic, expressive and socially acceptable bot make the user feel comfortable when talking to an artificial intelligence. Indeed, intelligence can not only be defined as understanding words but also linking words to the dialog context.
Having an emotional and quick response for emergency / unhapiness request is the key to gain customer’s loyalty.
How To Make Your Customers Comfortable With Chatbots was originally published in Chatbots Magazine on Medium, where people are continuing the conversation by highlighting and responding to this story.
Source: Chatbots Magazine