THE MACHINE INVESTIGATING THE MACHINE
A human-AI experiment in public accountability
For years, I have challenged administrative decisions through telephone calls, emails, complaints and requests for government records.
Now I am conducting a different experiment.
What happens when the citizen uses artificial intelligence to audit the administrative systems that make decisions about the citizen?
The experiment has three participants: the human witness, the AI-assisted investigator and the organisation holding the original records.
The human supplies experience, correspondence and evidence. The AI helps reconstruct chronology, identify contradictions and formulate testable questions. The organisation is asked to provide the original decision records.
The AI does not decide whether the organisation is guilty. The records must establish what happened.
Our working principle is simple:
RESPONSE IS NOT REMEDY.
A complaint reference is not an investigation. A reversed demerit is not an explanation. A missing email is not proof that no email existed.
And an AI-generated allegation is not evidence merely because it sounds authoritative.
The investigation itself will be documented, including the mistakes, corrections, unanswered questions and eventual agency responses.
This is the Truth Sequences experiment: can a transparent human-AI audit make administrative decisions more accountable?
