Decision Readiness Tool

Role
UX Designer & System Architect
Timeline
2 days
Type
Solo sprint
Tools
Google AI Studio
Role
UX Designer & System Architect
Timeline
2 days
Type
Solo sprint
Tools
Google AI Studio

TL;DR

Problem
AI tools answer decisions quickly and with confidence, even when people want room to think.
Approach
Drew on decision-making psychology and the Magic 8 Ball toy, and built a prototype in 2 days.
Solution
A working prototype that surfaces what to consider in a decision instead of recommending an answer.

01.Defining AI decision support outside of recommendations

AI tools are increasingly used to help people make decisions, but most LLMs are built to resolve a question quickly, not to help someone think one through.

In conversations with students and people outside the design world, I noticed these tools suggest too much, sound too declarative, and erode trust. General online sentiment about asking LLMs instead of searching the web pointed the same way. What users need differs: some want a direct answer, others want support working through a decision themselves.

How might we build a conversational interface that helps people think through a decision instead of handing them an answer?

02.Drawing on psychology research & a familiar visual metaphor

To create AI decision support that makes room for reflection, I paired psychology research with Magic 8 Ball inspiration, hypothesizing it would make a new interaction paradigm feel familiar and low-stakes.

Why a Magic 8 Ball?

While considering the idea of externalizing a question, a Magic 8 Ball immediately came to mind. It felt like a natural fit for this interaction model, which used delayed answers. The question quickly became: how do I extend the ask-and-pause interaction to AI? And could I design AI that simply states the unknowns?

Research-based design requirements

Research-based design requirements: the theory, the human behavior it describes and the requirement it sets
TheoryHuman behaviorDesign requirement
01 Automation biasConfident AI answers invite over-relianceAvoid declaring solutions for the person
02 Cognitive offloadingHard thinking gets handed to AI to save effortHelp the person work out what to do themselves
03 MetacognitionEasy answers can feel like understandingSurface gaps for the person to check

03.Prototyping a slower, more reflective kind of AI decision support

To let people work through a decision in their own words, I built idk (yet), a working prototype that sorts what a person already knows from what is still unknown into separate cards, and keeps both apart from answers.

Conversational flow for thought

Turn 1: What seems true so far. Grounding. Restates what the person said.
Turn 2: What's still unclear. Uncertainty. Names the gaps in understanding.
Turn 3+: Framing the decision. Then stops, with no recommendations or predictions.

Visual and motion design

Visual
  • Few elements and no decoration, one screen at a time, keep attention on the person's own thinking instead of answers.
  • Near-black and blue-grey cards are calming, and the glowing accents call back to the Magic 8 Ball ritual.
Motion
  • After submit, the screen holds on "Reflecting…" before the card fades in, giving the person time to keep thinking.
  • The confirmed card drifts up and dims as the next takes its place, keeping attention on the current question while earlier cards stay above to revisit.

04.Reflection

In future explorations, I would start with discovery research: talking to people who want help understanding a decision without being advised, to learn what they need from it.

If I were continuing with this project, I’d test the prototype and visual language with people facing real decisions, observe how they move through grounding, uncertainty and framing, and how they respond to getting no recommendation. I’d also check how people perceive the visual language and the Magic 8 Ball metaphor.