Why translated conversations feel awkward even when the translation is right
Better Translate Team ·
Ask somebody why they gave up on a translation app mid-conversation and they will
tell you it was inaccurate. Push a little and the story is almost always about something else.
Three things break before the translation does
The timing. Conversation minimises silence between turns in every language
studied, with cultures varying by only a couple of hundred milliseconds around the mean. Insert a
device handover and you are adding seconds to a system tuned that tightly. It reads as awkwardness
rather than as latency, because nobody experiences a conversation as a latency measurement.
The gaze. Research on interpreted consultations finds gaze and body orientation
doing much of the work of managing whose turn it is. Hand your phone over and both of you are
looking at a screen instead of at each other, so the signal that used to hand off the turn is gone.
The ownership. Giving a stranger your unlocked phone is a small social event
every single turn. People hold it awkwardly, angle it away, hand it back too fast. None of that is
about language.
What is actually left for the translator to get wrong
Plenty, and we have written about it: machine translation drops negations, invents content, and
quietly omits lines rather than mangling them, which is the
failure hardest for a person to catch.
But those are content errors, and content errors are not what makes an exchange feel
uncomfortable in the moment. Interaction errors are. They are also the cheaper ones to fix, and
almost nobody was fixing them.
What we changed
- Nobody hands anything over. The screen splits and the far half is rotated 180
degrees, so it faces the person opposite while your half faces you.
- Both halves are live. They are reading your last sentence while you are
starting the next one.
- The phone stays on the table. It sits between you like a menu, not like a
thing being passed.
- Nobody touches your phone. They read their half and answer on it.
None of that makes a translation more accurate. It removes the reasons people stop using one.