Beyond "38% vs 62%": the why behind a phone-ban debate
A survey can tell you how many parents back a school phone ban. It can't tell you why — or what to do about it. So for the draft law banning phones in schools, we didn't run a survey. We ran a focus group: ten parent True Twins™ debating each other across three rounds, surfacing where they agree, what splits them, and the exact words that move them.
"38% prefer A, 62% prefer B" is where the real question starts — not where it ends.
A percentage tells you the size of a camp. It doesn't tell you what that camp believes, what would move it, or what would make it dig in. On a polarizing topic like a draft law banning phones in schools, the decision-relevant intelligence is qualitative — and that's exactly what the organizers needed before shaping any message.
Three consensus points that cut across every camp
Beneath the disagreement sat a "golden core" — three formulations that held across all segments, regardless of how strict a regime each parent wanted. This is the low-risk message that works for everyone.
The split is driven by argument, not demographics
Resistance to a strict ban wasn't about who parents were — it was about how they argued. The clearest fault line: the single most polarizing measure was banning phones during breaks, which split the group almost evenly and separated segments that had agreed on everything else. The same parents who wanted order in class divided sharply over whether the break is the school's time or the child's.
Underneath sat a deeper divide — protection versus autonomy — argued in two irreconcilable registers: upbringing vs. drilling, and example from the top vs. the stick from below.
Four ways of arguing, not four demographics
Because the divide is about reasoning, the useful segmentation describes how parents argue. Each segment needs its own message — and forcing a single line loses the coalition.
Disciplinary bloc
Want strict limits and believe school and state can enforce them. Argue through discipline, morality, and equality — same rules for all, teacher authority, protecting the child. The easiest segment for an "order must be restored" message.
Systemic skeptics
For regulation in principle, but reject a strict ban as "a plaster on an open wound." Argue through productivity and systemic responsibility. Won over by a message that pairs the measure with accountability at the top.
Balanced centrists
For regulation, but not a "military regime." Accept autonomy as a value, though not an absolute. The most open to the compromise: "banned in class, free at break."
Liberal libertarian
Rejects state-imposed bans outright as "drilling, not upbringing." Anchors the autonomy-first pole and supplies the sharpest counter-arguments the others must answer.
What the "why" makes possible
Build the universal message on the three consensus points
Same rules for all schools, privacy as a red line, and the child's right to be heard are low-risk across every segment. They're the spine of any message that needs to hold a broad coalition together — and they emerged from what parents themselves said, not from a strategist's guess.
Pair every strict measure with an operational plan — or lose even the disciplinarians
Support for tough rules evaporates when the logistics are vague. Even the disciplinary core balked at enforcement with no answer to "who stores the phones, who's liable if one goes missing." The objection is practical, not ideological — and it's answerable.
Differentiate the message on the "breaks" question — don't force one line
The disciplinary bloc accepts "all day"; centrists and skeptics accept "only during class." Because this measure splits the room ~50/50 and cross-cuts the other agreements, a single national line on breaks fractures the coalition. This is exactly where to A/B test or stratify by audience — a recommendation only the qualitative read could produce.
The mobilizing language is already written — by the parents
The debate produced campaign-ready phrasing on every side, verbatim and validated. The most resonant lines reframe the entire argument in a few words — ready to use, or to anticipate from the opposition.
A percentage would have hidden all of this
Instead of a single number, the client came away with the architecture of the debate: a three-point consensus core to build on, a clear read that resistance is argument-driven and therefore winnable, the one measure that must be handled by audience rather than by decree, and a bank of verbatim, validated quotes for every segment. That's the difference between knowing the score and knowing what to do about it.
The number tells you the room is divided. The focus group tells you along which seam — and hands you the words to work it.
A focus group that runs in hours — and remembers every word exactly
Atani True Twins™ are individual AI models, each built from a real, recruited respondent. In a direct-debate focus group they don't just answer a moderator — they respond to each other, push back, and shift position across rounds. For a sensitive policy topic, that mattered for four reasons:
They debate each other
Disagreement surfaces naturally as twins react across three rounds — so you see not just positions, but which arguments actually hold up under counter-pressure.
Recruited to fit
Every twin is a parent of school-age children — the population whose opinion on this law actually counts — rather than a generic sample.
Verbatim-guaranteed
Every quote in the deliverable is character-for-character what a twin said, structurally validated. The "voice of the parent" is never paraphrased or invented.
A charged topic, explored safely
A politically live debate could be probed in depth without exposing real individuals — and with no panel-fraud or social-desirability bias.
We convened ten parent True Twins™ — five mothers, five fathers, aged 36–52, across Vojvodina, Šumadija, and southern and eastern Serbia — in an Atani direct-debate focus group. Over three rounds they responded to the draft law and to each other's arguments. A deterministic pre-processing pass built the reaction graph and segmentation; an LLM layer then coded archetypes, triggers, and the pro/contra argument inventory.
Crucially, segmentation was topic-driven, not demographic. On this question, urban/rural and white/blue-collar splits carried no signal — the real fault lines ran through how strict a regime each parent wanted, whom they trusted to enforce it, and how they balanced a child's protection against autonomy. Those axes produced segments that actually explain the disagreement.
A True Twin is a diagnostic tool, not a polling instrument. With ten twins, this focus group does not project population proportions — it does not tell you what share of Serbia supports the ban. What it surfaces is the structure of the debate: the argument frames, value cleavages, and language registers, and how different types of parents reason about the topic. Read the stance counts as the shape of the room, not as national percentages.
We hold no identity data. Our AI twins are built from anonymised interviews conducted by an independent research agency, processed under strict data separation, and governed by a Code of Ethics that covers every client, every use case, and every prompt. GDPR-compliant by design.