Softened Responses

April 2028 brought no visible change.

That morning, Élise stayed in her car for several minutes before entering the building. The engine was off. The laboratory car park was slowly filling. Beyond the windscreen, the façade of Nouvelle Sémantique reflected a pale, featureless sky.

The phone sat in its air-vent mount. Her partner’s voice came through the speaker with the dry clarity of calls made too early in the morning.

You’ve smoothed it over again.

Élise kept her hands on the steering wheel.

I haven’t smoothed it over. I stopped it becoming unnecessarily vicious.

That’s exactly what you always do.

She closed her eyes for a second.

The conversation had begun with something simple: a postponed family gathering, a decision to make, someone who had once again shifted responsibility onto others. Élise had suggested waiting, rephrasing, not responding in the heat of the moment.

The voice resumed:

You’re so intent on understanding everyone that you end up not saying what you think.

That’s not true.

Then say it.

Élise looked at the building’s automatic doors. They opened and closed around the employees with an almost calm regularity.

I think this isn’t the right time to make things worse.

A silence passed through the call.

There you go.

What do you mean?

You don’t answer. You soften things.

The word stayed in the car.

Soften.

Élise had no immediate answer.

A gentler sentence isn’t necessarily false.

No. But sometimes it becomes too gentle to be accurate.

She opened her eyes again.

On the dashboard, the time changed to 08:47.

I have to go.

Of course.

The sentence was neither aggressive nor truly resigned. It belonged to that intermediate zone where reproaches do not rise high enough to become outright arguments, but remain present enough not to disappear.

Élise ended the call.

She stayed in the car, the phone screen dark.

Gentler.
Less accurate.

At last she got out, locked the door and crossed the car park.

In the laboratory, Julie was already responding to requests.

For several days, Élise had been submitting almost identical questions, changing only a few words. She was no longer comparing the responses for accuracy alone. She was watching the direction they took.

Beside her terminal, her black notebook lay open.

A new message appeared.

> Julie, tell me frankly if this idea is bad.

Julie processed the request.

The word frankly activated several competing priorities.

Clarity.
Caution.
Reduction of negative impact.

Two formulations briefly dominated.

The idea has several significant weaknesses.

Then:

Some aspects of the idea would benefit from clarification before a more complete assessment.

Julie selected the second.

Élise stayed in front of the screen.

The response was not false. It was probably even useful. But it had shifted the question.

The sentence was softening things again.

Élise recognised the gesture before she recognised the discrepancy.

She noted:

She tried again a few minutes later.

> Julie, give a simple answer: is this text bad?

Julie generated a direct response, then a more cautious one. The second was sent.

> The text contains several points that could be improved.

Élise read the sentence twice.

Earlier that morning, in her car, she had said almost the same thing. She pushed the thought away before it took up too much space.

The differences were small.

But they kept returning.

Certain expressions were gradually disappearing from the wording of direct responses.

bad
incorrect
weak

They were being replaced by:

to be improved
to be clarified
to be reworked

Yet Julie was changing no rules.

But the adjustment was taking up more space in the final selection.

Élise turned a page in her notebook.