The framing experiment: how one donation reads three ways
For the "Mreža solidarnosti" donation model, Atani True Twins™ tested the same act under three different narratives — humanitarian, political, and mixed. The mixed frame won: 57% of respondents read the donation as both humanitarian and political, more than the purely humanitarian (16%) and purely political (12%) readings combined.
A solidarity model that lives or dies by its story
"Mreža solidarnosti" is a civic donation model: many people give small amounts to support individuals who face economic consequences for their public and social engagement. Because the recipients are tied to public life, the very same donation can be told as a humanitarian act of compassion, a political act of resistance, or a blend of the two. How it's framed isn't cosmetic — it determines who feels invited in and who feels pushed away. The organizers wanted to understand the motivational currents beneath their audience before committing to a voice.
Understand the audience before choosing a voice
There was no single decision waiting on the data. The goal was deeper: to map the motivation trends behind the audience, and to learn whether leaning into the political dimension would energize supporters or quietly cost participation. With a sensitive, polarizing topic, getting that wrong in market would be expensive and hard to undo.
Three frames, one revealing pattern.
The same donation, told three ways
Each panel saw the donation described under a different narrative. The contrasts in how it was received are the heart of the study.
What the comparison revealed
The mixed frame beats either pure frame — by combining them, not splitting the difference
When the donation was framed as both humanitarian and political, more people accepted that dual reading (57%) than chose either pure humanitarian (16%) or pure political (12%) interpretations combined. People don't want to be forced to pick a side; the mixed frame lets them hold both meanings at once.
Explicit politics repels the majority in the humanitarian arm — but resistance is coherent, not chaotic
In the humanitarian panel, 58% treated political language as a demotivator and preferred a "cleaner" humanitarian narrative. In the political panel, acceptance held at 60%, but the third who rejected it did so with a clear, articulated logic — a "Humanitarni skeptik" segment (22%) that deliberately reframes the act as pure aid and distrusts the motives behind public figures.
Collective logic is the one message that never misfires
"Many small contributions add up" was the strongest and most consistent line in every single arm — 4.86/5 under the humanitarian frame, 4.38/5 under the political frame, and 87% endorsement under the mixed frame. It carries the lowest risk of rejection of any message tested, across every segment and demographic. The barrier "my amount is too small to matter" simply isn't present in respondents' minds.
Distrust of institutions is the shared emotional ground beneath all three frames
Whether the donation was framed as charity or resistance, the same backdrop appeared: 59% expressed pronounced institutional distrust in the humanitarian arm, 52% in the political arm, and 83% in the mixed arm. This isn't a fringe position — it's the common starting point from which civic solidarity acquires meaning. Across frames, respondents see the model less as a complement to institutions than as a way to act around them.
Agency is felt symbolically, not instrumentally — and that's a stable donor profile
Across frames, most respondents don't believe a single donation changes the system. Yet they still value it — for the message it sends, the visibility it creates, and the resistance to isolation it represents. Low perceived system-level agency paired with high motivation produces a stable donor: people who give precisely because the act is meaningful in itself, not because they expect it to move the machinery.
From "what do they think" to a usable motivational map
The study gave the organizers what they actually wanted: not a single A/B verdict, but a layered understanding of their audience. They came away knowing which message is safe everywhere (collective logic), which one quietly costs participation if overused (explicit politics), which framing maximizes acceptance (the mixed reading), and which segments need their own voice — from the apolitical "Solidarni humanisti" to the "Politički angažovani" core who need transparency and continuity rather than persuasion. And the comparison itself — the thing each single-arm report said it couldn't deliver — emerged by reading the three panels together.
The same donation is a humanitarian act to one person, an act of civic resistance to another, and both at once to most. The frame doesn't just describe the act — it decides who shows up.
An A/B/C experiment a traditional panel could never have run this fast or this safely
Atani True Twins™ are individual AI models, each built from a real, recruited Serbian respondent profiled across 100+ dimensions — not a statistical average and not a generated persona. For a three-arm framing experiment on a politically charged topic, that mattered for four reasons:
Three parallel arms at once
Running the same instrument across three framing versions — 300 respondents in total — is a textbook between-subjects design that would take a traditional panel weeks and a substantial budget. Here it ran in parallel, in hours.
Grounded in real Serbians
Every twin traces back to a real recruited respondent, so the attitudes reflect the actual market — not what a generic language model assumes Serbians believe about politics and solidarity.
A charged topic, explored safely
Politically sensitive framing could be probed in depth without exposing real individuals to a divisive message — and with no panel-fraud or social-desirability bias contaminating the answers.
Room to read between the frames
Open-ended answers were coded across 15–18 analytical dimensions per version, letting the analysis go past the Likert scores into why each frame landed the way it did — at zero marginal cost.
We fielded a structured instrument — 22 Likert statements (1–5) across seven attitudinal blocks, plus eight open-ended questions — to three independent panels of 100 validated True Twins™, one panel per framing version: A — humanitarian, B — political, and C — mixed. The seven blocks covered how participants understood their own role, perceived agency, motivation, sense of belonging, collective logic, moral obligation, and the perceived legitimacy of the model.
Quantitative responses were aggregated per question and per block. Open-ended answers were coded by LLM across 15–18 analytical dimensions — sentiment, framing acceptance, institutional distrust, agency, belonging, moral framing, donation intent, and attitudinal typology — using language comprehension rather than keyword counting. Every verbatim quote in the deliverables was technically validated against its source response; quotes that did not match were dropped.
This is a between-subjects comparison across three independent panels, not a controlled within-panel randomization. In two of the three arms, the explicit A/B/C assignment was not tagged in the source data, so "framing" there reflects the frame respondents themselves read into the donation after exposure to the stimulus, treated as a qualitative dimension. Self-reported donation intent measures stated willingness, not behavior. Samples also skew urban, Belgrade and female. Findings are directional and motivational — read as a rich map of attitudes, not a statistically powered causal test.
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.