3. Generative Research

Workshop with Chiyu (Engineer working on autonomous vehicles)

The first generative workshop we conducted was with Chiyu, an engineer from CMU.


  1. Collage — Describe your ideal partner
  2. Task flow — What do you do when you drive an autonomous vehicle
  3. Flexible modelling — Imagining a possible dashboard for an autonomous vehicle


100 concepts

Feb 27

We faced a short lull in activity while trying to schedule workshops with EMS professionals. We decided to do a quick brainstorming session. The goal was to come up with 30–35 ideas each in about 30 minutes. Through this exercise we were able to identify some ideas and generate questions we wanted to answer through upcoming workshops.


Questions that arose post-ideation:

Feedback modalities
Ambient light – flashing lights and sound – function allocation
Q- Validate it

Window tint – to show different level of control
Wearable that captures your personal data and changes the ambience of the ambulance accordingly

Car empathy – Safety
“Stay safe, 3 shootings reported here in the last week”
“What a jerk” about another driver
Petting/cleaning an ambulance as a ritual

Display of information – switching between navigation and vitals
Q- Validate it

Steering wheel
Q- Prototype form

Personality – changing over time, changing over time per instance?

Notes that can be filled out with a CUI – low hanging fruit.

Thoughts/Questions/Things we don’t know/Things we want to investigate

Rituals – We should study people’s car stating rituals

Design decisions- Input/CUI
Do people want to see the V2V, V2I conversations?
Distractions – Sounds
What are the most common cases – to build scenarios
Allowances- Make decisions until such time as…


Workshop with Michael – Volunteer EMS

Feb 28

Our first workshop with an EMS participant was helpful in confirming what we found out during our exploratory research.

The collage exercise (Ideal partner) led to surprisingly similar results as the one we did with Chiyu. Perhaps that’s something that’s universal anyway.


The task flow exercise made much more sense with this participant as well. We could probe around specific points using the visual map.


Our final flexible modelling exercise was also successful. People didn’t have a hard time  imagining themselves in an ambulance and were able to give us feedback and suggestions. This felt like the most ‘generative’ part of the workshop and helped the participant understand what we were trying to design. Using this to debrief was helpful as well.




Ambulance Tour

Feb 29

We were able to arrange a tour of one of the ambulances currently part of the Pittsburgh EMS network. Actually being in the physical space occupied by our main stakeholders was enlightening. It gave us a better sense of the current ambulance-driving experience, as well as a clearer understanding of the physical context.


We found that the current interface installed in the ambulance is messy and unergonomic. The constrained space and inefficient layout of the controls makes it hard for them to do their job well.


Workshop with Paramedics

Feb 29

Following the ambulance tour, we conducted a workshop with EMS professionals — 2 crew chiefs and a paramedic.

We conducted two of the exercises from the previous workshop with them. After the workshop we found that the task flow exercise might be more fruitful if combined with flexible modeling. We retained the “ideal partner” collage exercise, and added a component of representing a “nightmare partner”.

The collage exercises again confirmed some of what we already suspected; Ideas of a teammate who’s reliable, confident in their skills, and is also a friend you can have fun with came up across all the participants.


We then asked them to role play scenarios so that we could see how they interact as a team and what they expect from each other. Using post-it notes, drawing on the acrylic ‘windscreen’ and using wooden blocks they showed us how they go about receiving a call, navigating, making decisions, coordinating with other teams and logging incidents.


Insights — Implications

Framework for initial synthesis of generative research 
  • overload of information — contextualize info; multimodal feedback
  • high levels of aural input — good noise and bad noise
  • trust is an important aspect of the jab — often trust is built through non-verbal cues and proactive assistance
  • self-confidence, know what you’re doing — but gap in experience and training among partners
  • ambulance can break rules — but this makes the job harder because of stress and liability
  • miscommunication is common
  • High stress and requires a great deal of adaptability and self-assurance — but those levels of focus and energy are not sustainable for long; need to decompress and capitalize of calmer moments


These insights and implications largely fall into three major themes. These could also be related to the different personalities the CUI could take up as informed by the collage exercise.

  • Safe tranportaation and navigation — CUI as co-pilot
  • Communication, planning and coordiantion — CUI as teammate/mediator
  • Stress relief — CUI as friend/therapist




While thinking about concepts, we identified three main questions that would need to be answered:

  •  Does the CUI have a form, or is it disembodied?
  • Does it have a personality/agency?
  • Does it provide a customised experience to each EMS professional, or a standard one for all?

While we tried to imagine what the CUI could be like depending on the various combinations of answers to these questions, it felt too soon to lock these answers down. However, we were able to think about the form in which the CUI would manifest. We narrowed it down to: around the user, with the user, or on the user. 

Presentation 3

Presentation 3 – Phase3_GenResearchv5

Debrief of Presentation 3


  • Need to narrow (Kevin)
  • What about if it fails? What’s the backup? (Irina) High reliability organizations – therefore need redundancy and recovery.
  • Where does the CUI help? (Kevin)
  • Are we thinking more software or hardware? (Kevin)
  • What about management infrastructure issues? How can make this from triage to optimal to more? Pushing the imaginary vs working with the current imaginary (Kakee)


  • Liked showing possible form factors (Kevin)
  • Idea of ambient data collection

Next steps

  • Talk about conversation framing – Personality and dialogue?
  • Talk about applications for generic car experiences
  • Create persona
  • Ways to prototype and whom to prototype with – can we put CMU EMS people in stressful situations?

Post presentation ideation

After debriefing, we decided to map out the entire task flow and look at specific sub tasks and opportunities within each. Surprisingly, we came up with a ubi-comp concept! We’re going to use this going forward as we derive our features list.

IMG_5252 IMG_5257