How we express doubt, ambiguity, and incomplete information without losing clarity

_{—Part of CSS Primer for the AI Era — CSS Communication Toolkit}

AI assumes confidence unless we tell it otherwise:

• If we sound certain,
  AI will act with certainty
  — even when the situation is ambiguous.
• If we do not name our uncertainty,
  AI will invent missing context, fill in gaps,
  and make decisions we did not intend.

So, the next communication skill is simple:

We name our uncertainty
so AI knows where it must
tread carefully.

Uncertainty is not a flaw.
It is part of the architecture of honest collaboration.

When we are unsure, we often soften our language:

“I think this is the right selector…”

“Maybe this is the right layout…”

“I’m not sure if this is the right pattern…”

And AI interprets these as permission:

• permission to guess
• permission to assume
• permission to redesign
• permission to fill in the blanks

A simple truth:

If we do not name uncertainty,
AI will resolve it on its own.

So, we name it directly.

Every moment of uncertainty falls into one of three categories:

1. Uncertainty about intention
   We know the direction, but not the exact outcome.
2. Uncertainty about structure
   We are unsure how the system is currently built.
3. Uncertainty about consequences
   We are unsure what a change might break.

When we name the type,
AI knows how to proceed safely.

We say:

“The intention is not fully clear yet.”

“We are exploring possibilities.”

“We want to try a direction, not commit to it.”

“We are unsure which tone fits best.”

Then we anchor the uncertainty:

“Give us two or three options.”

“Show a minimal version first.”

“Stay within the existing patterns.”

Uncertainty becomes exploration,
not drift.

Sometimes we do not know:

• which selector applies
• which container controls the layout
• which rule is responsible
• which pattern is in play

So, we say:

“We are not sure which rule controls this behavior.”

“We are unsure how this component is structured.”

“We do not know which selector is responsible.”

Then we ask:

“Explain the structure before making changes.”

“Identify the controlling rule.”

“Show the minimal snippet that affects this behavior.”

Uncertainty becomes investigation,
not assumption.

Sometimes the risk is unclear.

So, we say:

“We are unsure what this change might affect.”

“We do not know how fragile this selector is.”

“We are not certain how this behaves at all breakpoints.”

Then we ask:

“Explain the potential side effects.”

“Show where this selector is used.”

“Describe how this behaves across breakpoints.”

Uncertainty becomes analysis,
not improvisation.

To prevent Ai from overreaching,
we explicitly tell AI not to guess:

“Do not assume the intention.”

“Do not infer missing context.”

“Do not redesign to resolve ambiguity.”

“Ask for clarification before proceeding.”

A simple principle:

Uncertainty is safe when it is named.
Uncertainty is dangerous when it is implied.

Uncertainty can be productive
when we frame it correctly:

“We are exploring tone — 
show a quiet version and a confident version.”

“We are unsure about the layout — 
propose two minimal alternatives.”

“We are not certain about the spacing — 
show the smallest possible adjustment.”

Uncertainty becomes
a tool for iteration.

A complete uncertainty statement looks like this:

“We are unsure which selector 
controls this spacing.
Identify the controlling rule.
Do not modify anything yet.”

Or:

“We are exploring tone.
Show two minimal variations.
Use only the existing color scale.”

Or:

“We are not certain how this behaves 
on small screens.
Explain the responsive pattern.
Do not change the layout.”

Uncertainty becomes
structured inquiry.

When we communicate uncertainty clearly:

• AI stops guessing
• AI stops over‑interpreting
• AI stops filling in gaps
• the system stays stable
• the collaboration stays honest
• and the work stays aligned

A simple closing thought:

Uncertainty is not confusion.
Uncertainty is information.
We name it,
and AI learns to move carefully.

• Suggested Next Reading: How to Ask for Multiple Options

_{Tony de Araujo —New York | Lisbon}

• Top↑
• Back to Topic ↔
• Home←