Speaking to systems.
WHERE VOICE SUITS
Hands and eyes occupied Accessibility Devices with no screen Short, well-defined tasks
WHERE IT SUITS LESS
Noisy environments Public places, where speaking aloud is awkward Complex tasks with many options Anything requiring privacy
WHAT CONSTRAINS VOICE INTERFACES
No visible affordances, so users do not know what is possible Errors compounding through a conversation Slower than reading, for presenting options
WHAT TO DO ABOUT DISCOVERABILITY
Tell users what they can do, briefly, and let them ask.
WHAT TO KEEP SHORT
Responses.
WHY
Long spoken responses are not retained.
WHAT TO CONFIRM
Anything consequential, before acting.
WHAT TO PROVIDE
A way to cancel or undo.
WHAT TO CONSIDER LOCALLY
Accent recognition, which is frequently poor for local speech Background noise, which is common Whether users are comfortable speaking to a device
WHAT TO TEST
With actual users, in actual conditions.
WHAT TO ALWAYS OFFER
A non-voice alternative.
