Concept evaluation · Conversational UX
Virtual assistant state design
Two rounds of testing followed the same question: could people tell what the assistant was doing without the animation taking over the screen?
Two study rounds, showing the visual states revised between them.
The design choice
Which visual treatment communicated the assistant’s state clearly without competing for attention?
Listening, detecting a voice, thinking, speaking and recovering from a misunderstanding each needed a recognisable state. Asking which animation people preferred would only answer half the question.
Round one, then round two
Let the first study change the second
- Before testing
Review familiar patterns and earlier evidence
Previous modality studies and established assistants showed which patterns people might already recognise for listening, processing and speaking.
- First round · 7 people
Compare two visual families, state by state
Seven participants discussed strengths, weaknesses, expected meaning and preference during a 90-minute concept evaluation.
- Second round · 6 people
Bring the revised states back to participants
Six further participants completed an association exercise and comparative review of the revised animations and states.
Two rounds, not one repeated test
Round two tested revised states and a distinct recovery signal.
The second study did not repeat the first. Revised states and a clearer recovery signal were tested with a new group of participants.
What participants actually distinguished
One visual family was preferred; individual states still needed changes
The line-based family felt more familiar and readable. The useful detail was in the exceptions: one processing animation dominated the screen, and recovery needed its own unmistakable signal.
- 01
Comprehension
The line-based family communicated actions more clearly
Participants described it as more visible, familiar and understandable across several assistant states.
- 02
Attention
A processing animation could become too dominant
The first round showed that one thinking state occupied too much of the screen and could distract from the primary task.
- 03
Iteration
State-specific changes mattered more than one global rule
The next direction refined size, thickness, position, colour and motion, and added a clear misunderstanding state.
Design direction
The final recommendation was not one animation everywhere
The final handover covered motion, scale, placement, colour and recovery feedback for each state, rather than applying one rule to every animation.