Association of Police and Crime Commissioners (APCC), National Police Chiefs’ Council (NPCC), and Police Digital Service (PDS) — Supplementary written evidence (NTL0057)     


1.             This submission answers questions raised during the evidence session attended by PCC Alun Michael and PCC Darryl Preston (APCC leads for Data and Bioethics), Mr David Tucker (Faculty Lead on Crime and Criminal Justice, College of Policing), and Prof. Paul Taylor (Policing Chief Scientific Advisor).

At Q84, the Chair said: “it would be helpful for us to see on paper who all these organisations are and what their different roles are”. … should encompass organisations/groups/bodies you have referred to in the session, and any others you feel we should be aware of – rather than simply being the three represented in the session.


2.             The following diagram outlines the roles (the diagram’s orientation is arbitrary). A picture containing text

Description automatically generated

Pg 14, please could Prof Taylor check & possibly clarify the phrase: “which means the informed PCC and CC are only partially informed.”


3.             This was checked and amended for clarity in the transcript. The amended version conveys that PCCs and Chief Constables can be informed only to the extent of the available scientific evidence. A limited ability to undertake testing restricts the degree to which PCC and CCs can be informed.


Following on from L. Rickett’s question (Q96) the College of Policing is looking into evaluating the OxRec model for risk of violent recidivism. The review is due in March 2022, source here: (A validation study of OxRec model for assessing risk of violent recidivism ( Our understanding is that the developers have themselves written that: “the tool has low predictive accuracy at the individual level” adding that “one potential harm that is not justified is preventive detention.”


a)            Why is the College seeking to validate this tool, when the developers themselves say it has “low predictive accuracy at the individual level”?


4.             The College of Policing is not evaluating the OxRec risk assessment tool. The College maintains a research map onto which forces and academic institutions contribute their police related research work. The map is intended to enable those with an interest in police related research to see if there is already work taking place in a particular discipline of policing. The OxRec entry on the map is solely to inform others of research being undertaken. It states that the tool is not to be operationalised at this stage, but researchers are testing whether the approach can be more effective at identifying people who pose the greatest risk. The College is not involved in the work.



b)                                          In the context of the assessment from the developers of “low predictive accuracy”, does the College endorse this statement from the validation study?: “The tool offers the opportunity to more accurately identify risk of violent recidivism”.


5.             The statement around opportunity, made in the project description provided by the study researchers, is a hypothesis about what OxRec may deliver, to be assessed by, inter alia, the validation study. It is not a positive or negative endorsement.



c)                                        Is there a threshold for when a tool is too inaccurate to be used? Is there a percentage of false positives or another measure of inaccuracy beyond which a tool should simply not be used? How would each organisation respond to the suggestion that a numerical threshold be introduced? (this threshold could be specific to individual tools, or it could be specific to the context in which it is being used).


6.      The use of thresholds to guide decisions about new technologies is desirable and something that researchers and delivery teams use. To set these thresholds once nationally, however, would be challenging due to the range of new technologies, the range of uses, the differences in contexts of use, and the pace of development. For this reason, policing sets principles of use that are followed by Forces, with national expert support, and oversight. When new technologies mature, such is the case for ANPR cameras for example, then thresholds for performance are often set and suppliers are required to meet them.


Per Lord Dholakia’s question (Q97) on the consultation of communities. Your very helpful written evidence submission (NTL0049) refers to work by the Digital and Data Ethics Guidance Group (DDEGG) to: “promote, where operationally appropriate, an ethos of transparency and engagement with the public to maintain and promote trust and confidence.” Please can you: 1)  provide details of work underway to consult and build trust with communities who may be particularly affected by the use of data held by police forces and 2) clarify under what circumstances this would not be “operationally appropriate”;


7.             As indicated in our initial evidence, the Digital and Data Ethics Guidance Group (DDEGG) is being established and so there is no work underway. DDEGG’s primary purpose is to support Force ethical processes so it will be promoting best practice amongst Forces to encourage public engagement and transparency. The intention, subject to DDEGG’s ratification, is to have an internal guidance group that facilitates sharing of knowledge across forces, and a second independent group comprising external representation from the public to provide scrutiny of and challenge to force-level and national projects.


8.             Complementary to this is the APCC Biometrics and Data Ethics Working Group, which was set up in 2020 to develop the APCC’s position on policing’s adoption of new biometric and analytical technologies (encompassing areas such as algorithms, AI and facial recognition.). This group has recently expanded to include the APCC leads of both the Transparency and Integrity portfolio and the Equality, Diversity and Inclusion portfolios to ensure a broad range of views and oversight.


9.             The principal circumstance in which it would not be appropriate to engage in public consultation would be where the Group is considering Sensitive information as defined by Government Security Classifications.


Per the Chair’s question (Q98), she asked: “Do you think more legislation is needed or should we be relying on our well-established legislation and principles as developed through case law?” Mr Michael responded with "I agree with that…” Please can Mr Michael clarify which option he agrees with?


10.        Mr Michael confirms that it was the second principle proposed in the question - "well-established legislation and principles as developed through case law" - that he was agreeing with, and this is his position.


5 January 2022