- Who runs it
- MoJ
- Can it affect people?
- Yes. It may support a decision about a person, such as access to a service, place, payment, or permit.
- Possible rights impact: Direct decision or access
How much do we know?
A source is a published page or document behind a claim. The label measures what is documented, not whether a system is good or safe. Missing information does not prove that a safeguard is missing.
Sources and further reading
- MoJ: Data First (Splink) / GOV.UK Algorithmic Transparency Records / 2024-12-17
- GOV.UK Algorithmic Transparency Records - official inventory / GOV.UK Algorithmic Transparency Records
optional detailFull record and details
The interface and explanatory guides are translated. Record names, record descriptions, quotations, detailed editorial notes, and linked public documents stay in their published language so evidence is not silently changed.
What it does
Splink is an open-source tool for probabilistic data linkage developed for the Data First program. Data First provides academic researchers with datasets across the Justice system where individuals can be identified with a single, universal ID across domains.
If this affects someone
The Data First datasets have been created solely for use by accredited researchers to perform cross-cutting analysis in the justice space. The datasets provided are at an individual record (i.e. person) level and is anonymised for data protection purposes. This high level of granularity allows for more robust and flexible analysis. The datasets are intended for identifying trends across the justice system to inform government policy, not for operational decision-making. A summary of the latest research can be fo...
Given researchers are not provided with personally identifying information, it is not expected that they will be able to assess whether a link between individuals is correct or not. However, as the models are regularly being reviewed within the MoJ, researchers can provide feedback if they believe they have found an error or anomaly which will be addressed in the next model iteration cycle.
set into Splink. Splink identifies links within the dataset (i.e. pairs of records with a match probability above a specified threshold) and the linked records are grouped and given a deduplicated identifier. These identifiers are added to the domain-specific datasets. For cross-domain linkage, all datasets are fed into Splink together to identify links. Once these links have been created, all records with a sufficiently high match probability are grouped and given a new linked identifier. These identifiers are...
Sources and further reading
First indexed from GOV.UK Algorithmic Transparency Records on 2026-09-03. Only usable fields from the public record are shown; missing answers remain blank.
Fields were mapped from the linked public record by deterministic code. No generative model wrote this text.
Facts to check
- Who runs it
- MoJ
- Life area
- Public sector
- Proof level
- Source-backed
- Footnotes
- 2
- updated
- 23 months ago
The score orders work, not people.
Review priority rises for rights-sensitive uses, missing sources, unanswered accountability fields, and old evidence. Strong official documentation lowers it. The score is not a verdict on safety, fairness, or legality.
A missing field means the attached public sources did not answer it. Coverage is uneven, publication does not prove performance, and a certificate has a defined boundary.
Previous versions
- Imported2 months agoSubstantialImported from GOV.UK Algorithmic Transparency Records.
Help check this record.
Add a public source, challenge a claim, or point out a missing way to reach a person.
A submitted name, email, description, source, or correction is used only to verify the record and prevent abuse. It is not published automatically.