Public Law Project (PLP) — Supplementary written evidence (NTL0059)
- Public Law Project (PLP) provides this note to supplement our response to the Justice and Home Affairs Committee’s call for evidence on new technologies and the application of the law (the Response). It concerns the role of human decision makers in the context of the sham marriages algorithm, described at paragraphs 7-11 of the Response.
- PLP has been investigating the sham marriages algorithm by making requests under the Freedom of Information Act 2000 (FOIA). Since submitting the Response, we have received new documents from the Home Office which bear on our understanding of the involvement of human decision makers in instigating sham marriage investigations.
- On 27 October 2021, we received a document entitled ‘Marriage Assessment AQA Report’, dated 16 January 2020, together with an Addendum, dated 11 May 2021.
- The original AQA Report described the sham marriages algorithm as a “fully automated end to end process”. Cases were categorised according to risk. ‘Green’ cases were described in the following terms: “Green – These cases pass the triage and are low risk. Letters should be auto generated and sent to these applicants advising them they can marry after the 28-day statutory period.” ‘Red’ cases were described in this way: “Red – These are those that have not passed the triage/assurance and will automatically extend to 70 days in order for the Home Office to investigate. Letters informing applicants of this to be sent automatically.”
- However, the Addendum paints a somewhat different picture. The context for the Addendum is that the AQA Report presented “the model’s use as envisioned at the project’s inception”. There were “some inconsistencies between what is described in the report and how the model has been implemented in practice”.
- In the Addendum, the sham marriages algorithm is described as “a semi-automated solution to support the end to end process”. The model identified two types of case: ‘pass’ and ‘fail’. Pass cases are described in identical terms to green cases (above). But the description of fail cases differs from the description of red cases: “These are those that have not passed the triage/assurance and will be considered for extension to 70 days in order for the Home Office to investigate. Letters informing applicants of this are generated and sent out by reviewing staff. The cases are then sent to staff that will task to ICE teams [emphasis added].” In short, the Addendum indicates that fail cases – unlike red cases, as described in the original AQA Report – are reviewed by a human being before an investigation is instigated.
- The detail of the human review stage is unclear. We do not know whether the human decision maker exercises meaningful discretion. Especially if the human decision maker knows that a case has failed triage, there is a risk of automation bias (see paragraphs 16(d)(v) and 18 of the Response). We have been told that cases that fail triage are “usually” investigated by the Home Office, but we do not have precise figures. It may be that the vast majority of fail cases are investigated.
02 February 2022