Experts Blog

Chatbots need personality

  • Bruno Pedro Mettler

In times of digital increase, everyone is confronted with a diversity of communication tools such as iMessage, Whatsapp, Threema, Facebook-Messenger or Slack. Consequently, the idea of a chatbot as a “Conversational User Interface” between a human and a machine has become a greater focus of attention.

Chatbots gain in importance

With smartphones taking over as the primary means of communication, we have learned to chat with others. In fact, today we chat even more often than talk on the phone. Messaging and chat apps such as iMessage, WhatsApp, Threema, Facebook Messenger and Slack are among the most popular mobile apps and have changed the way we communicate. We are used to having digital conversations and feel anxious if the other person does not answer within seconds.

This leads us to another idea: using these “conversational user interfaces” not only for communication between two individuals, but also for communication between human beings and machine in the form of chatbots. Not an easy task, because the user expects to talk to a human and not a “stupid” machine.

What is a good chatbot?

In human-to-human chat communication, common context plays a crucial role. We rarely chat with complete strangers, we normally have common experiences that serve as a basis for the dialogue. A machine does not know or understand the context (at least not yet). But we can design the “stupid” bot so as to make it an acceptable conversation partner. This can be achieved by following a few rules.

A chatbot needs a personality based on a persona with a unique voice and tone, which is necessary to ensure a smooth conversation. Humans are able to communicate one thing using a huge variety of expressions. A good chatbot is able to play with these options to better imitate human communication. However, the user should always be informed that they are talking to a bot and not a real person.

An intelligent chatbot anticipates the user’s feedback and guides the course of the conversation. This is best done by suggesting possible answers to the user. This is not the case in human-to-human communication, since the other person does not need any suggestions to react appropriately. A bot, on the other hand, uses suggestions to guide the user and ensure that the conversation does not get stuck in a dead end.

Chatbots should be designed as learning systems: data can be collected and used for learning based on the content entered by the user. Typos, different words and phrases can all be used as input in the learning process.

The future development of chatbots will be greatly influenced by two driving forces. Firstly, machine learning, which enables the expansion of chatbots’ skills. The aim is for the machines to understand our communication and its context (cognition). To be able do this, they collect, aggregate and analyse a lot of personal information. This development is still in the early phase, but individual use cases can already be implemented today.

Secondly, the user interface can be developed further by integrating voice communication into the chatbot. If we combine a bot with voice instead of chat capability, we end up with digital voice assistants such as Siri, Alexa, Cortana and Google Assistant. It is a field in which the big tech giants constantly challenge each other. Speech recognition, however, brings an additional level of complexity into play.

Experience is important

The bottom line is that chatbots and conversational user interfaces have become part of our reality. Now it is time to learn how to design them correctly, use them and constantly improve them. Chatbots have a lot of potential applications. The biggest challenge is that the complexity increases very quickly.

As a result, it is better to launch a small project, implement it in a clearly defined application area and gain first experiences rather than to aim to develop the perfect bot right from the beginning.