Emobot

Emobot is a customer-care bot that reacts to human emotions and shows sensitive feedback through subtle colors and animations in order to create a climate of trust. It was built during the AI HACK DT, a 2-day hackathon organized by Deustche Telekom in Berlin. (event website )
My classmate Maëva and I worked on the challenge called "Autonomously Target Users with Emotional Responses for Extra Customer Care".

Role : Concept, UI/UX design, animation, pitch preparation

Our team

Before the hackathon, Maëva and I worked on the challenge and we came up with the idea of Emobot. We then presented it to our multidipliscinary team composed of Arne (software architect), Daniel (software engineer), and Andy (project manager), and they agreed to build the AI-powered conversation program that would bring Emobot to life.

Colors to convey emotions

During the take off of our plane to Berlin the day before, Maëva and I noticed that the lights were dimmed and their colors went through various shades of blue. It instanly set a calm atmosphere. We thought it was an interesting way to reassure the passengers and instil them with a feeling of safety and confidence. We were inspired by that to solve one of AI’s main issue : how humans react to it.

Some of us may be scared by the idea of sharing payment credentials or personal details when talking with a customer care chatbot, or others may be doubtful about how actually helpful and understanding the bot could be. We tend to be more confident when we reach out to real people because we know that they understand our emotions and how to react to them.

On the other hand, bots behaving almost but not exactly like real human beings can be confusing for the user. (cf: Theory of Uncanny Valley ).
We needed to find a balance. Rather than declaring anthropomorphism, the bot could just suggest it with colors.

Our solution

We started studying various color theories and we created a system that would work with the program built by our teammates. We assigned colors to different emotions, and we started thinking about subtle animations that could convey a feeling of understanding and empathy.

Then, I designed the interfaces and animated them on After Effects, while Maëva prototyped them in HTML and CSS.

Soon after writing our pitch, me and Maëva presented Emobot to a jury of various AI specialists and researchers such as Felix Burkhardt .

Conclusion

The analysis part of the project was the most stimulating. Reflecting on our own behaviors as humans is an interesting task : what are the dynamics of a dialogue between two humans, in what ways could we transcribe them into a chatbot window? We really wanted to create an experience similar to a conversation we could have with another human (according to our reactions, our interlocutor will choose a different tone to keep the conversation calm and cordial).

The UI and UX of a chatbot was a complex challenge to tackle since it had to rely on an already existing interface (Facebook chat, Slack, etc.). We had to keep in mind a lot of constraints (eg: keeping the chat clear and visible, not distracting the user from the conversation).

Although we didn't win the first prize because of technical difficulties, the jury really liked our idea and congratulated us. Overall, it was a great experience to work with different people coming mostly from dev/engineering backgrounds. The pitch was also a stressful but enriching exercise.

Read a more detailed article we wrote about the hackathon [FR] .