Prototype v0.1 - Failure-Mode Explorer

Probationary System Simulator

Compare how permanent-record, time-decay, probationary, and mediator-review models handle the same case in a never-forgetting system.

This is not a repentance calculator. It is a research sandbox for testing assumptions about restoration, evidence, gaming risk, appeal, and bounded probation.

What this is

A side-by-side model comparison tool for credit, moderation, hiring, fraud, and related high-stakes systems where a record persists and restoration is contested.

What this is not

A claim that repentance, dignity, or justice collapse into one number. The output is provisional by design and should provoke critique, not compliance.

Case inputs

Adjust the case. The outputs update immediately.

12
68
62
45
58

Current read

Probation continues

The current case shows meaningful change, but the evidence is not yet strong enough for automatic restoration.

Probation target

18 months

Restoration signal

61 / 100

Gaming risk

Medium

Recommended path

Human review

Permanent record

Truth preserved, no restoration path.

Penalty remains

The violation stays active because this model never relaxes consequences.

Time decay

Time matters more than evidence.

Clock still running

Restoration waits for enough time to pass, even if the behavior evidence improves earlier.

Probationary metric

Evidence and trajectory both matter.

Probation continues

The system sees improvement, but restoration stays conditional until the case becomes harder to game.

Mediator review

A human or oversight layer absorbs uncertainty.

Human review

A mediator layer can compare the record to context instead of pretending the score is the whole case.

Why this result

  • The record shows change, but not enough verified time for automatic restoration.
  • The cost of repair is visible, but still modest relative to the violation.
  • The appeal evidence helps, but it does not eliminate uncertainty.
  • The safer path is to keep probation active and require review.

What this simulator does not know

  • It does not know intent, coercion, trauma, or structural constraints behind the behavior.
  • It does not know whether the available evidence is biased, incomplete, or legally unusable.
  • It does not know whether the scenario should allow probation at all under domain-specific law or policy.
  • It does not replace human accountability for deciding what restoration should mean.

How to critique it usefully

  • Change one assumption and say what failure mode it creates or resolves.
  • Point to a real domain rule, paper, or case where the model breaks down.
  • Explain what a mediator or oversight layer should be allowed to override and why.
  • Focus on measurable harms: false restoration, permanent lockout, opacity, or gaming incentives.

Get project updates without sending a full critique

If the simulator is useful but you are not ready to challenge the model yet, join the mailing list for research updates, new essays, and prototype notes.

We use basic site analytics and store your email, masked IP, and source page for project updates. We do not sell personal data. See terms and participation policy.

Ready to challenge assumptions, add constraints, or propose a pilot? Use the contributor form. Want to support the work? Leave a tip on Ko-fi.

Disagree with the output?

Good. That is the point.

Algodai is trying to turn an abstract justice-and-mercy argument into something testable. If the result feels wrong, say which assumption failed, which evidence type is missing, or which review path should exist instead.