Try to prove ETHICA wrong.
A system that only publishes arguments for its own excellence has already failed. Objections, risks, evidence requirements and corrections must be first-class information.
An objection should become visible — not disappear into an inbox.
When the external review ledger opens, every substantive objection should receive a status, response and traceable link to any resulting change. Personal data, security-sensitive detail and clearly abusive content will not be published.
- 01Register
Claim, affected component, evidence and proposed test.
- 02Triage
Substance, duplication, security and publishability are assessed.
- 03Test openly
Support, counter-evidence and remaining uncertainty are recorded.
- 04Version
An accepted correction is linked to change and history.
No external objection ledger is published in this owner-restricted review version.
What should be published
- The substantive core of the objection
- Current status and accountable assessment
- ETHICA response and evidence state
- Correction, rejection or open uncertainty
What can change the model
- A disproved empirical claim
- A constitutional conflict
- An unacceptable security or power risk
- Outcomes that fail in pilot or test
Risk, control and evidence should be read together.
Technocratic power
Complex technology can turn experts and system owners into a new political class.
Role separation
No actor may alone propose, analyse, decide, implement and approve the same measure.
Open evidence chain
Norm → requirement → architecture → data → model → code → test → outcome.
Majority power
More direct mandates can increase pressure on minorities and fundamental rights.
Constitutional floor
Rights and appeal sit outside ordinary package majorities.
Rights review
Every package must state which rights are affected and on what legal basis.
Knowledge must not dress up as certainty.
Verifiable base.
Support with remaining uncertainty.
Interpretation or synthesis.
Construction that can be rejected.
Possible future.
Red-team questions
What if the AI is wrong?
It must not be the sole source of norms, judge or executor. Independent models, reproducible evidence, human review and explicit use boundaries are required.
What if a future government abuses the infrastructure?
Data minimisation, federated storage, separate cryptographic domains, independent control and technically viable recovery should reduce the damage a less benevolent government can do.
What if citizens do not participate?
ETHICA must not require everyone to become a full-time administrator. Major value choices are grouped into a few decision windows, while randomly selected citizen councils receive paid time to review the system.
Who reviews the citizen councils?
Every cohort starts with a public outcome contract and ends with an outcome card, open issues, handover and an independent assurance statement.
Isn't “non-ideological” democracy itself an ideology?
ETHICA does not claim to be value-neutral. Rights, priorities and trade-offs are normative and chosen by people. Non-ideological describes the governing method: it is not bound to pre-packaged left, right, green or religious answers, and every intervention faces the same rights and evidence tests.
What if ETHICA does not improve outcomes?
Then the model should not be protected by its name. Pilots and institutions must be reducible, changeable or terminable when the evidence does not hold.