Koala is an exploration into conversation between human relationships in public transportation systems. The results were translated into principles and applied into a service CUI for pedestrians.

The full project consisted of an Insights phase (4 weeks) and Development phase (8 weeks). It was developed at Umeå Institute of Design, teamed up with Christoph Zobl, Melissa Hellmund and Shigeo Katsura-Gordon.

It took part in the Microsoft 2016 Design Challenge: Achieving Symbiosis and the Conversational User Interface.


It is a service Conversational User Interface for public transportation systems. It interacts with pedestrians giving information about the bus schedule and assists in ticket purchase.


Design a product, service or solution that demonstrates the value and differentiation of the CUI.

Are there insights we can gain by looking at the relationship between people?


  • Ethnographic Research
  • Concept Development
  • Prototyping

Nearly all activities were performed by all members of the group.


Our exploration started in the public transport system of Umeå, Sweden. Every day thousands of passengers take buses and along the way they interact with different people, spaces and interfaces of the system.

Therefore, we have at least 3 different kinds of interactions in this system:

  • People-people: Passengers among themselves, Bus drivers and the company’s representatives;
  • People-technology: Phones, vehicles, bus ticket machines;
  • People-spaces: Bus stops, information kiosks, vehicle garage;

All these relationships impact the system and we went on field to explore more about them.


Exploring system agents

It started by talking and interacting with the most consumer-facing agents in the system: bus drivers, passengers and public transport management representatives.

We conducted semi-structured interviews with bus drivers in the city.

1. Bus drivers

Apart from driving, these professionals usually sell bus tickets onboard, solve doubts from passengers and some of them even welcome tourists giving tips about the city.
We quickly saw how the experience changed when they helped with activities that were outside their responsibilities, conscious about their impact in the system.

The method used with passengers were semi-structured interviews supported by user journey maps.

2. Passengers

Taking the bus is a daily activity that can impact on how your day starts. Passengers are highly dependent on the transportation system, especially the ones that live far away from work/school.
The vehicles and bus stops are spaces where different interactions happen: they meet friends and get to know new people.

Showing a video with user cases in the bus system (design probe) + interviews were the methods used.

3. Management representatives

The managers were very helpful in clarifying how the system works. For instance, my group was aware of 3 institutions responsible for the functioning of the system, but actually there are at least 6.
Their biggest concerns are costs. In Umeå case, they were able to match cut of costs with politics for leveraging passengers satisfaction with the service.

Research overview

The ethnographic research helped us to understand more about working relationship dynamics and friendship dynamics. I’d like to highlight 3 relationship paradigms found in the research:

When two agents communicate, there is a need to catch the message prior to analyzing it. Being aware of when an important event/message is going to happen is key for a seamless conversation.

Especially concerning symbiosis, the level of power an agent holds shapes the conversation, impacting the relationship.

Personal Space
The distance for what is called a “personal space” is dictated by cultural norms, culture, age and gender. Even an inanimate body like a bus stop has a space of its own.

These paradigms are communicated in visual and non-visual (body language, eye contact) manners.

Presenting key findings in a workshop

After 4 weeks of research, we needed to present the outcomes to our mentors at Umeå Institute of Design. The idea was to introduce them to the lives of the 5 individuals we focused on our research. A presentation happened in the format of a role play where the mentors would interact with 3 different stories found during the research. After each story, an explanation would be given on how it fit our findings.
A highlight was that the full-body role play with scenarios and props immersed the mentors into the situations, giving them not only an overview of the context but also of details within quick interactions.


This stage started with the following question:

What would happen if an agent of the system would be replaced by a CUI?

We wanted to assess the users’ expectations, feelings and assumptions on a new subject that would be as useful as the bus drivers we found on research.

Setting the right expectations: considerations on CUI for services

When talking about Conversational User Interfaces, in the realm of voice interactions, examples like Siri, Google Now, Alexa and Cortana come into mind. Other examples such as HAL 9000 (2001: A Space Odyssey), Samantha (Her), and Jarvis (Iron Man series) are examples of characters in popular culture. Still, they are personal assistants, different from what a public service needs.

Personal Assistant

A personal assistant focus on knowing its master in both practical and emotional manners. There is a usual expectation on them for the understanding of context and feelings.

Service CUI

Service assistants focus on efficiency and helpfulness. They deal with different people everyday, solving recurrent issues most of the times.

Human vs Machine

The purpose of initial prototypes was to understand the differences between how users interact with human entities agains robots. For that, 2 experiments were conducted with rough prototypes:

Experiment 1: HUMAN VOICE

A totem would stay at the bus stop to help passengers, answering questions and assisting tickets purchase. The communication was made through audio and when visual support was needed, a tablet screen could appear to the user.

For that we used cardboard, a tablet with WiFi and Skype.


No visual support was offered in this experiment. A light was added to indicate status: when the user should talk or listen.

Tools used: cardboard, a smartphone, a speaker with built-in bluetooth connected to a computer. The computer operator had a text-file with different sentences. For the robotic voice, we used Google Translate.

Guiding principles

In the end, the learnings from the prototyping sessions were translated into 5 guiding principles that were used to develop the final version of Koala:


Create a sense of physical presence for the conversational user interface and acknowledge human physical presence.

System limits

Use language to guide and limit the system’s offerings.

Not a human

Make it clear the user is talking to a robot with artificial intelligence.

Unknown problems

Enable the CUI to deal with problems outside the scope of the system.

Turn taking

Define and communicate when it’s time to talk and listen.

Final Solution


Let’s chat and find out how we can help each other.