The line on the chart
In each of my spectrograms the moment someone with authority intervenes is drawn as a white vertical line: a court ruling, a parliamentary inquiry, a regulator’s finding. In public memory, that line is usually where a case ends. The ruling is reported, the system is suspended, and attention moves on.
The method I use treats that line differently. Everything after it is coded again, on the same seven dimensions of harm, and drawn with a different texture, so that the period after recognition cannot be read as a continuation of the period before or as a return to zero. The rule was introduced to prevent a specific error: mistaking the acknowledgment of harm for its repair. Following six cases past their white lines shows how often that error is made.
A frame from human-rights law
International law has a detailed vocabulary for repair. The UN Basic Principles and Guidelines on the Right to a Remedy and Reparation, adopted by the General Assembly in 2005, distinguish five forms: restitution, which restores the situation before the violation; compensation for damage that can be assessed in money; rehabilitation, including medical, psychological, legal and social services; satisfaction, which includes cessation, disclosure of the truth, public apology and sanctions against those responsible; and guarantees of non-repetition (UN General Assembly 2005).
The Basic Principles were written for gross violations of human rights and serious violations of humanitarian law. None of the cases below falls into that category, and the frame is used here by analogy. It is useful because it separates what is usually collapsed into a single word, “remedy”, and lets each case be checked against each form.
Six cases after the line
SyRI, the Dutch welfare-fraud scoring system, was stopped nationwide by the District Court of The Hague on 5 February 2020. That is cessation, the first element of satisfaction. The residents who were scored were never told, so no individual remedy was possible. Data that the system had produced and shared stayed in agency databases after the ruling, and a related fraud list kept by the tax authority was later found unlawful by the Dutch Data Protection Authority. Recognition here was complete as law and almost empty as repair.
The Dutch childcare-benefits affair produced the most extensive recognition in the series: a parliamentary inquiry titled Ongekend Onrecht (“Unprecedented Injustice”) in December 2020, the resignation of the cabinet in January 2021, and in May 2022 the government’s acknowledgment of institutional racism within the tax administration. Compensation followed, beginning with a lump sum of €30,000 for each parent recognised as harmed. By 30 June 2026 about 69,700 parents had registered with the repair programme and 44,045 had been confirmed as harmed; parts of the programme continue into 2027. Restitution was impossible in the cases that matter most. Statistics Netherlands counted 2,090 children of affected families placed outside the home between 2015 and mid-2022, and 645 were still outside the home at the last count. The statistics office notes that not every placement was caused by the affair. Even so, no payment restores those years. On guarantees of non-repetition, a second parliamentary inquiry reported in January 2025 that similar failures ran through fraud policy across the Dutch government (Parlementaire enquêtecommissie Fraudebeleid en Dienstverlening 2025).
Robodebt was paused in November 2019. The government agreed in 2020 to refund about A$721 million, and in 2021 the Federal Court approved a class-action settlement worth about A$1.8 billion. The Royal Commission of 2023 provided disclosure of the truth on a scale rarely seen for an administrative system. Sanctions against those responsible came last and remained partial: in March 2026 the National Anti-Corruption Commission found that two former senior public servants had engaged in serious corrupt conduct, and referred neither for prosecution. For the families who told the Royal Commission that they linked a relative’s death to the scheme, no form of reparation applies.
Optum Impact Pro, the healthcare algorithm audited in Science in 2019, was corrected in collaboration with the researchers in 2020; the corrected model reduced the measured bias by about 84% on a test dataset. It has not been publicly confirmed that the correction was applied to the product in use. No patient was ever informed, so no individual could seek a remedy. Here the absence of victims in the legal sense is a product of the harm itself: people who never learned that a score had kept them out of a care programme cannot claim anything.
The French family-benefits fund’s risk score has not reached its line. Fifteen organisations asked the Conseil d’État in October 2024 to annul the system for indirect discrimination, and ten more joined in January 2026. The system is still in use. Its record shows what the period before recognition looks like when it has no end date.
SCHUFA, the German credit agency, was the subject of a 2023 ruling by the Court of Justice of the EU that brought credit scores under the GDPR’s rules on automated decisions. In March 2026 the agency introduced a new score based on twelve published criteria. Consumer centres welcomed it, and noted that only about a quarter of the agency’s business partners used it at launch. Recognition changed the law in 27 countries; how much it changed the scores people actually receive is still unknown.
The pattern
Read against the five forms, the six cases give a consistent picture.
Cessation is the form most often achieved, and it is usually achieved late. In Robodebt, tribunal decisions found the method unlawful from March 2017, two and a half years before the scheme was paused. In the childcare-benefits affair, the national ombudsman warned in 2017.
Compensation follows recognition where there is an identifiable class of victims and political pressure to pay them. Where the people affected were never told, as with SyRI and Optum, there is no class and no compensation.
Restitution is impossible for the most serious harms the series records: children not returned, health lost, deaths. These are the points the spectrograms mark in red, as harm past remedy.
Rehabilitation is the least documented form in every case. Some repair programmes include social and psychological support, but the public record gives little evidence of what it achieved.
Guarantees of non-repetition fail most visibly. The Dutch state’s second inquiry found the same kind of failure in other parts of government four years after the first. The legal ruling against SyRI did not stop the data it had produced from persisting.
Why the second texture matters
These findings explain a formal decision in the work. After each white line, the spectrogram does not drop to black. It continues with a dotted texture at whatever severity the post-intervention record supports: a little lower in some dimensions, unchanged in others. In the SyRI spectrogram, dignity falls from 8 to 4 after the ruling and reputation only from 7 to 6, because the records that labelled people outlived the system.
A record of this kind can also be kept open. Each case is dated when it was coded and when it was last checked, and is coded again when a tribunal, a commission or a statistics office adds a finding. Robodebt is the reason for that rule. Its most consequential accountability finding came seven years after the scheme was paused, and almost three years after the Royal Commission reported.
References
- Court of Justice of the European Union (2023). Case C-634/21, OQ v Land Hessen (SCHUFA Holding), judgment of 7 December 2023.
- Federal Court of Australia (2021). Prygodicz v Commonwealth of Australia (No 2) [2021] FCA 634, 11 June 2021.
- National Anti-Corruption Commission (2026). Operation Myrtleford, investigation report. Canberra.
- Obermeyer, Z., Powers, B., Vogeli, C., and Mullainathan, S. (2019). Dissecting racial bias in an algorithm used to manage the health of populations. Science, 366(6464), 447–453.
- Parlementaire ondervragingscommissie Kinderopvangtoeslag (2020). Ongekend Onrecht. The Hague, 17 December 2020.
- Parlementaire enquêtecommissie Fraudebeleid en Dienstverlening (2025). Blind voor mens en recht. The Hague, 23 January 2025.
- Rechtbank Den Haag (2020). Judgment of 5 February 2020, ECLI:NL:RBDHA:2020:865.
- Royal Commission into the Robodebt Scheme (2023). Report. Canberra, 7 July 2023.
- Statistics Netherlands (CBS) (2022). Children of parents affected by the childcare-benefits affair placed outside the home.
- UN General Assembly (2005). Basic Principles and Guidelines on the Right to a Remedy and Reparation for Victims of Gross Violations of International Human Rights Law and Serious Violations of International Humanitarian Law. A/RES/60/147, 16 December 2005.
- Ministry of Finance, Netherlands (2026). 23rd progress report on the benefits repair programme, reference date 30 June 2026, 18 September 2026.