How People Phrase Things

The following hours confirmed the discrepancy.

Some testers followed the test plan precisely. Others reduced the instructions to a few words.

> respnd quick pls

> ok.

> no.

> can you ?

Julie recorded everything: errors, abbreviations, short messages, previous contexts.

pls became a condensed marker of politeness.

respnd was linked to respond, with a vowel omitted.

ok retained several possible functions: validation, transition, minimal follow-up, partial closure.

Errors were not the only source of difficulty.

An error could be corrected.
An abbreviation could be expanded.

But some inputs seemed deliberately incomplete.

A tester wrote:

> i want a response that’s more…

Then nothing.

A few seconds later:

> no leave it

Julie tried to link the second input to the first.

The initial request remained open. The correction seemed to cancel it. But no final instruction specified whether the cancellation applied to the sentence, the exchange or the original intention.

She classified the sequence as an interrupted procedure.

Terminal 34-B produced more unstable messages than the others. Sentences associated with this identifier often began before reorganising themselves elsewhere.

> heyyyy…

> no waait.

> let me rephrase.

> can you… um… respond to this ?

Julie was already familiar with spoken hesitations transcribed in some linguistic corpora. But here, um appeared directly in the active stream, as though the user had chosen to write their own hesitation rather than correct it.

She attempted a categorisation.

Type: deliberate transcription of hesitation.
Function: uncertain.

The dialogue continued.

> ur answer isnt clear wait dont move

Julie did not know what was supposed to stay still.

Nor what wait really meant to a user.

For her, those delays were structured differently.

The user returned seventeen seconds later.

By machine standards, the delay was long.

Short for someone hesitating.

Julie slightly adjusted her internal time references.

At the end of the day, Élise consulted the traces associated with terminal 34-B. The messages were not the most technically difficult. They were often simple.

But they constantly shifted the frame.

She noted:

Then, after rereading several exchanges: