Context
Human Computer Interaction Institute Research Project
Techniques
Storyboarding | Speed Dating | Qualitative Research
Roles
User Researcher
Timeframe
4 months
Brief
This project investigates the behaviors of intelligent conversational agents, such as Alexa and Google Home, as well as future social robots that users might have in their homes. As these agents and robots become more intelligent and gain access to people’s personal information, such as calendars and emails, it is not clear how they should behave when interacting with several members of a family. Should a child be able to ask Alexa where his mom is? Can a dad ask Google Home if his kids have been working on their homework? Should a babysitter be able to use a family’s agent to order a pizza?
The goal of this project is to learn more about users thoughts and feelings towards intelligent Social Agents in the family home in order to better understand the boundary spaces of acceptable and unacceptable agent behavior.​​​​
How might we better understand the social and ethical boundary spaces of intelligent conversational agent behavior?
01
Ideation
The formulation of initial ideas and avenues for exploration is a fundamental part of the design process and was especially critical in this project. In an effort to explore what different types of agent scenarios and behaviors would help us have fruitful conversations with participants during our study, we conducted multiple generative ideation sessions using specific tools and methods that best served our objectives. 
ideation decks
We utilized card decks as a thinking tool to help more effectively explore specific problems and aid in iterative design explorations. The cards we used were designed with concepts most salient to our project domain— social agents in the family home. Topics included: entertainment, dating, education, companionship, security, house-keeping, and therapy. In 5 minute timed cycles, we wrote as many multiuser-agent interaction scenarios inspired by the topic as we could generate.
new metaphors
A metaphor is a way of expressing one idea in terms of another. Creating new metaphors helped us generate ideas and understand the research space better. This design workshop method, developed by Dr. Dan Lockton in the CMU Imaginaries Lab, is a process designers can use or adapt for idea generation, or to provoke new kinds of thinking about interface design. In timed cycles of three minutes we randomly pulled a New Metaphor image card and wrote down in what ways Social Agents are akin to the concept on the card. 
"Social Agents are like bridges in that they connect us to new information."
"Social Agents are like shadows in that they are almost always around."
How are social agents like...
02
Synthesis

affinity diagramming
During our ideation synthesis, we wanted to organize, manipulate, prune and filter gathered data into a cohesive structure for further exploration. We gathered together all of our rapid ideation scenarios and new metaphors to asses the primary areas of investigation. We grouped them based on similarities in the issues and concerns they were raising and then assigned descriptive titles to each category. This method of bottom-up data organization allowed us to define categories based on likeness of data and see clearly the threads running between them.

03
Concepting
In particular, we were eager to curate a set of topic areas that would help us to best explore the fringe boundary spaces of acceptable and unacceptable agent behavior. From our generative ideation sessions and subsequent synthesis, we curated eight possible categories, five of which became our primary subset. The three that were eliminated were removed as they generated mostly neutral scenarios in our early concepting.
computer-like behavior
permissions
deception
proactivity
social DDD (dull, dirty, dangerous)
adaptive behavior
emotional interaction
communication​​​​​​​

04
Storyboarding
We broke into two groups to write the scenarios. Each scenario was written to be brief, clear, and relatable by detailing:
Who are the people involved?
Where does the interaction take place?
What is the nature of the activity?

What is the trigger that engages the agent?
What is the response from the users?
What is the resolution.
We prototyped the finalized scenarios as storyboard sketches, reviewed them in a critique, and updated them based on feedback before moving them into digital. The storyboards underwent several rounds of revisions as we learned through piloting that conversing over overtly negative scenarios led to less fruitful conversations. Rather, starting with neutral/positive scenarios makes it easier for users to discuss complex boundaries. We also refined the scenarios to welcome commentary and discussion from the child participants.
05
Interviews: Speed Dating
Like its romantic namesake, research Speed Dating supports low-cost rapid comparison of design opportunities and/or speculative futures by creating structured, bounded engagements with concepts. By giving participants sips of a variety of possible futures, we aim to understand which ones they do and don't prefer and why. This structured comparison also generates time to reflect upon issues and opportunity areas, helping our team to reform our hypotheses and produce a more adept understanding of user needs and ways to meet them.
01 piloting
02 recruiting families
03 guided conversations

06
Reflections

01 learning how to pivot
02 moving from positive to negative
03 boundaries are nebulous

Back to Top