Our team was handed a new set of constraints—that the design be speculative, and incorporate a hot mic. We went to task devising a testable solution to a problem revealed by our research.

Between each concept, we considered tackling problems including anxiety, social isolation, and trust.

Eventually, we settled on Epi, a conceptual Alexa-like AI created to help partners keep promises, uphold rituals, and be self reflective.

We concocted a smoke-and-mirrors prototype and ran user tests to see how users would react to a science-fiction, fact checking, relationship monitoring AI.

Participants were told that our device syncs with their phone in order to get to know them and better navigate the conversation between them and the actor.

A friend of each participants sat in a separate room with another researcher and a laptop connected to EPI. These two used a text-to-speech program to attempt to moderate the conversation, check facts, and feed anecdotes to EPI.

While a lot of fun, the critical nature of EPI’s design forsook several design principals. We needed a pivot.

During testing, EPI was jarring and intrusive.

Furthermore, an overemphasis on things like fact checking and promise-holding were causing participants to tread on eggshells, which contradicted our principles of nurturing and assurance.

We pivoted. Preparing our final deliverables, we imagined a softer spoken EPI closer to a therapeutic AI than a "relationship cop."

The new EPI focused on allowing one or two people to achieve long term emotional goals and track their progress, while gently holding them to their promises.

She was something a little less out of the dystopia of Minority Report, and a little more at home in the summery future of Spike Jonez's Her.

Principles exist for many good reasons.

Just generating a set of feel-good principles isn't enough. Principles are to be consulted along every phase of the design process.

Not just a moral compass, principles help navigate through the pressure of deadlines when there isn't enough time to test an ideal amount of iterations. When properly heeded, they allow us to iterate with confidence.

Remember the emotional nuances of user testing and the importance of context.

Just as important as presenting an accurate mockup of your product's user flow is creating an accurate representation of the emotional surroundings you want your product to be used in.

I would have rather designed top-down, focusing on the emotional issue of insecurity, rather than building bottom up from the idea of promise-keeping.

It's good to design through specific problems over abstract ones, especially under deadlines. However, by beginning focusing on only a certain symptom (promise-keeping and arguments) of a broader issue (insecurity), we forgot the bigger picture, and our initial result slipped away from the issues at the core of our project.

Listen to your designer's intuition.

I discovered my first deep design flow state during the research synthesis stage of the project. The principles I generated were close to my heart, and when we began to drift, I felt it. Speaking up sooner would have meant hitting the ground running with a more solid, principled concept, saving us valuable time.