Machines have feelings too: how to design human conversations for your brand

Auteur
Matthijs Waaijer
Datum

Together with students of the University of Amsterdam we demonstrated that organisations should begin designing their bot conversations by assessing their brand. The brand personality gives us guidelines on how to design a brand’s messages to make them more satisfying for customers. In this blog post, I describe how emotion, arguably the most important human element of all, can be used by brands to create highly satisfying conversations.

For ages, humans have been talking to other humans. We’ve become pretty good at understanding what other people mean. Our conversations happen in a very natural manner - most of the time so naturally, that we don’t even have to think about it. We are also very good at understanding the conversational cues, emotions and personality of the person we are talking to. We just know if someone is excited, in a bad mood, telling a joke or being sarcastic. We also know how to react to these different emotions. For us humans, all this happens effortlessly.

Nowadays, technology enables us to have conversations with brands through the use of Conversational Artificial Intelligence (CAI). When talking to Google Home or any similar chatbot, the bot normally has a decent understanding of what the human is saying. Short requests such as turning on a light or asking for a weather forecast are fairly intuitive for most people. While the CAI technology is already smart, it is not yet at a human level. This, in turn, can lead to a greater potential for misunderstandings and miscommunications. As human-machine conversations become increasingly complex, it becomes harder to discern the intents of users. Consequently, users may indicate that these conversations can feel a bit unnatural or in some cases even frustrating; not exactly the effect you’re looking for as a brand. Therefore, conversational context plays a more important role.

Thus, to ensure more satisfying bot conversations, three main aspects must be considered.

  1. Expect CAI technology to rapidly improve its understanding of human interaction and its context
  2. Close interaction between humans and machines requires good UX designs. Poorly designed conversation flows and content ruin the user experience, no matter how ideal its underlying technology is
  3. The human element of conversations. When people talk, they use tone, pace, emotion and circumstance to make conversations engaging, personal and efficient. Likewise, brands also express their own identity. Currently, bots do not adequately replicate these advanced human elements.

With emotion, machines come to life

Dealing with emotions is very complex. Current bots are not able to experience emotions as human beings do. However, like humans, state-of-the-art technology is now capable of both recognizing the sentiment of the user they are talking to and expressing emotions in the responses they give. We can program bots in such ways they react with the emotions that fit a specific situation. For example, a bot can now give known users a warm welcome and could change its responses based on the emotions it detects from its user. In this article, we will focus on how bots can express emotions in writing and through emoticons.

To investigate the use of emotions in bots, we teamed up with students of the University of Amsterdam. Together, we conducted experiments aimed at understanding how the use of emotion in conversations influence the user experience. We developed mock-up human-machine conversations that exhibited different degrees of emotion and asked over 230 Dutch participants how satisfying those conversations were.

The results reveal that bots expressing emotions in their conversations outperform bots who do not. There are three explanations, which could account for this effect.

  1. Users are more engaged when bots display emotions. We found that bots, which express feelings, are just more interesting to talk to and their expressions are more similar to human-human conversations.
  2. Bots that use emotions nurture a more trustworthy connection, because users interpret such conversations as more welcoming, realistic and honest.
  3. Emotional expressions can also act as a feedback mechanism for users. For example, emotion can show that the bot has fully understood the user and that their goals are aligned. It turns out that emotions make bots ‘come to life’, resulting in more satisfying conversations.

Conversations that match your brand

Our research also shows that the positive effect of emotion differs between formal and informal brands. For formal brands, the positive effect measured in the study is only marginal. Users talking to a formal brand expect a formal, colder and focused conversation. The brand itself is sober and its conversations should reflect that. Hence, the benefits of expressing emotion are limited. Conversation designers should not overly exaggerate the amount of emotions for formal brands. For informal brands, the story is different. When talking to an informal brand, people expect a warmer, more playful and interactive conversation. This gives conversation designers for informal brands more room to play with emotion. Displaying a wider range of emotions and doing so more frequently, throughout the conversation, creates an experience that customers of informal brands are looking for.

cai1

So, what to do next?

We demonstrated that organisations should start with assessing their brand. Is it formal or informal? Thereafter, organisations can design conversations in such a way that the bots reflect the ‘personality’ of the brand. The most successful bots will be those whose use of emotions logically match the brand. With this information in mind, we can now compose bot conversations that truly reflect the brand and make conversations more human and satisfying.

CAI2

Interested in how we design natural conversations and build intelligent bots? Shoot us a message!


Special thanks to Robin van Hilst, Lisa Kal, Ilse Lambert, Wendy Meijerink, Ninah Rumkorf.

Tags

Conversational Interfaces Natural Language Processing Voice Cognitive Services Empathetic design Innovation