TR →
Table 1 / Engineering
1 / 20
arrow keys + F fullscreen + swipe on mobile

Making the Technical Understood by Everyone

Table 1 - Engineering. Not classic networking; the argument is at the center, introductions emerge from it.
01

Explaining the model ≠ making the problem visible

Engineers explain "how it works." Decision-makers need "why it matters, what could go wrong." The gap between these two languages is where policy dies.
Example: "The model has 70 billion parameters" is true but decides nothing. "Accuracy drops 12% on low-resource languages" changes a decision.
02

The jargon wall

"The model has bias because there's class imbalance for underrepresented groups in the training data." Clear to an engineer, a wall to a bureaucrat.
biasfine-tuninglatent spaceoverfittingattentionnot understood
Jargon is not bad by itself. The problem: speaking without knowing the other side doesn't know the language.
03

When an engineer says "the model has bias," how does one explain it to a bureaucrat?

Which analogy is enough, which one misleads? This is the table's focus question.
The bureaucrat's real question: "How does this change my decision?" If the answer reaches that question, the translation worked.
04

"Bias" in two languages

Failed
"The model has bias because there is a class imbalance for underrepresented groups in the training data."
Worked
"AI learns from past data. If some groups were underrepresented in that data, the system makes more mistakes for them."
Difference: the left explains a concept, the right makes the consequence visible.
05

Three rules

STOP
ASK
ELEVATE
Each rule has one goal: catch the moment translation breaks.
06

Jargon appears, stop

When a term is used, the facilitator steps in.
"Could you restate that in three sentences so someone without an engineering background could understand?"
STOP "With its attention mechanism, the transformer architecture..."
FIX "The model reads the whole text at once and focuses on the key words."
07

No monologues

If one person talks for 2+ minutes, the floor goes to someone else.
The goal is to measure the gap from everyone's language. One voice, one language means missing the table's purpose.
MONOLOGUE 5 minutes of uninterrupted technical talk
MOVE "And [other person], how does this look from your field?"
08

Elevate the naive question

Naive questions often sit exactly at the center of the interdisciplinary gap.
There is no "dumb" question. A question no one can answer is a map of the gap.
QUESTION "Wait, does the model 'know' things, or does it 'predict'?"
VALUE This question exposes the core difference between "knowledge" and "prediction."
09

You drive the translation

QUESTION 1 - "So what?"
"What consequences does this have?" Pulls the technical claim up to the decision layer.
QUESTION 2 - "Could you explain?"
"Could you explain this so someone without an engineering background understands?" Stops jargon on the spot.
You don't need to be an expert to use these two questions. Just curiosity.
10

The "So what?" test

Asked after every technical claim. If there's no answer, the claim doesn't belong at the table.
CUT "The model has 70B parameters." → So what? It changes no decision.
KEEP "Accuracy drops 12% on low-resource languages." → Turkish-speaking citizens get worse service. An equity problem.
KEEP "The system updates every two weeks." → A law written for v1 may be obsolete before it's published.
11

"Could you explain?"

When you hear jargon, ask without hesitation. That's exactly why you're at this table.
"Could you restate that term in three sentences so someone without an engineering background could understand?"
JARGON "Fine-tuning"
TRANSLATION "Taking someone with general knowledge and training them in one specific field."
12

Two questions, one goal

"So what?"
Aims for action, not understanding. Connects the claim to a decision. Filters empty claims.
"Could you explain?"
Demands understanding. Stops jargon from becoming a wall. Gives the speaker the responsibility to simplify.
One asks what it's for, the other what it means. Together they complete the translation.
13

Don't trust analogies, test them

Every analogy must pass three tests.
TEST 1 Does it survive two follow-up questions? "AI is like a student" → "Can it refuse? Does it know when it's wrong?"
TEST 2 Does it hide important complexity? "AI is like electricity" → electricity doesn't make decisions, AI does.
TEST 3 Does it work for this audience? "Gradient descent is like searching a valley" → not for a bureaucrat.
14

Three outputs from the table

1
The stuck concept
Where did the translation break? Which term hit which barrier?
2
The analogy that worked
Which comparison actually worked? We're writing it down.
3
The unanswerable question
The most concrete evidence of the interdisciplinary gap.
We write these where everyone can see them. They'll be the map for the next gathering.
15

A 60-minute path

Warm-up
Describe your profession without technical terms. 1 min each.
→
Cards
6 question cards. Each card is one translation practice.
→
Closing
Write the 3 outputs where everyone can see. Table done.
16

No prep required

But thinking about two questions before you arrive enriches the table.
QUESTION A concept from your field you could never explain to someone outside it?
QUESTION A question you've always wanted to ask another discipline but never could?
No answer? That's fine. The gap itself is the table's raw material.
17

This problem is bigger than one event

Strategy, question cards, jargon dictionary, participant guide, research notes - all open. So other tables, other cities can reuse it.
The jargon problem isn't unique to Turkey. Making it open source multiplies the solution.
18

Ready to use

For participants
5-minute guide, visual framework, 7 question cards, jargon dictionary.
For facilitators
Strategy, intervention cards, analogy tests, research notes.
github.com/murataslan1/bridging-tech-policy
"There is no classic networking at this table. We put a real problem at the center, and introductions emerge from the argument." TeknoPolitika Gathering - Table 1
Let's leave the table with one jargon, one question, and one question we couldn't answer.
Join: luma.com/kdyb73vf