How each model solves the same problems
Instead of “democracy versus anarchism”, compare the concrete ways collective problems get solved.
| Question | ✕ | ✕ | ✕ | |
|---|---|---|---|---|
| How are decisions made? | Participants submitted and voted on statements in Polis; researchers kept statements with cross-group agreement and converted them into training principles. | The Prime Minister is formally responsible for Diella; the Constitutional Court states AI cannot change who exercises public power or bears accountability. | A stratified random sample of 523 voters deliberated in moderated groups and plenaries and answered questionnaires before and after; no decisions were taken.Sortition / citizens' assembly | |
| How are conflicts resolved? | Polis maps disagreement into opinion groups; statements lacking consensus across groups were excluded rather than adjudicated. | Opposition challenged the decrees before the Constitutional Court, which dismissed the request on 8 October 2026; its full reasoning is pending. | Balanced briefing materials, trained neutral moderators and competing experts structured disagreement. | |
| How are public goods funded? | A published public constitution, methodology and evaluation results reusable by other developers and researchers. | Aims to improve public-service access and reduce discretion in public procurement. | Privately funded; produced published results, media coverage and a peer-reviewed study. | |
| How is AI used? | Constitutional AI: the model applies the written principles to critique its own outputs and generate training feedback. | Chatbot and voice assistant on e-Albania; a procurement system tender was still being prepared in August 2026. | No AI role. | |
| What are the main risks? | Organiser discretion in translation, unrepresentative or small samples, no ongoing accountability and voluntary uptake by the developer. | Unclear accountability for an AI 'minister'; no AI law yet; no published technical detail; AKSHI's leaders face a tender-rigging case.CLM-0359 | Self-selection into attendance; artificial setting; analysis by the method's designers; no institutional uptake. |
This illustrates differences in design; it is not a universal claim about outcomes. Each cell links to mechanisms and the claims behind them.