Public concern about the relationship between the police and BAME communities has not diminished in the years since the Stephen Lawrence Inquiry Report. The reason why the PSED is so important is that it requires a public authority to give thought to the potential impact of a new policy which may appear to it to be neutral but which may turn out in fact to have a disproportionate impact on certain sections of the population.
The importance of the PSED was emphasised in R (Elias) v Secretary of State for Defence [2006] EWCA Civ 1293 , [2006] 1 WLR 3213, at [274], where Arden LJ (as she then was) said:
“It is the clear purpose of section 71 [the predecessor to section 149 ] to require public bodies … to give advance consideration to issues of race discrimination before making any policy decision that may be affected by them. This is a salutary requirement, and this provision must be seen as an integral and important part of the mechanisms for ensuring the fulfilment of the aims of anti-discrimination legislation. …”
We acknowledge that what is required by the PSED is dependent on the context and does not require the impossible. It requires the taking of reasonable steps to make enquiries about what may not yet be known to a public authority about the potential impact of a proposed decision or policy on people with the relevant characteristics, in particular for present purposes race and sex.
We also acknowledge that, as the Divisional Court found, there was no evidence before it that there is any reason to think that the particular AFR technology used in this case did have any bias on racial or gender grounds. That, however, it seems to us, was to put the cart before the horse. The whole purpose of the positive duty (as opposed to the negative duties in the Equality Act 2010 ) is to ensure that a public authority does not inadvertently overlook information which it should take into account.
Against that background of principle we turn to examine the reasons given by the Divisional Court for rejecting this ground of challenge. There are two aspects of the present case which particularly impressed the Divisional Court.
The first was the fact that there is a “human failsafe” component in the way in which AFR Locate is used. This means that a positive match made by the automated system will never by itself lead to human intervention. Before such an intervention (for example a stop of a member of the public for a conversation with a police officer) can take place, there must be two human beings, including at least one police officer, who have decided to act on the positive match.
We do not consider the “human failsafe” is sufficient to discharge the PSED. As a matter of principle, it is not material to the PSED, which as we have observed, is a duty as to the process which needs to be followed, not what the substance of the decision should be. Secondly, as was acknowledged at the hearing before us, human beings can also make mistakes. This is particularly acknowledged in the context of identification. We would note the well-known warnings which need to be given to juries in criminal trials about how identification can be mistaken, in particular where a person has never seen the person being identified before: see R v Turnbull [1977] QB 224. Further, and in any event, this feature of the present case does not seem to us to go to the heart of the Appellant’s complaint under Ground 5, which is that SWP have not obtained information for themselves about the possible bias which the software they use may have.
The second matter which impressed the Divisional Court was the witness statement of PC Dominic Edgell, who found that there was virtually no difference in the statistics as to race or gender. We have considered that evidence.
Mr Edgell reviewed the AFR Locate deployments from after the UEFA Champions League Final of 2017 through to June 2018. During those deployments 290 alerts were generated. 82 were true positives and 208 were false positives. He says that it is important to note that these statistics are only of the persons who have generated an alert. The identity of those who passed the camera without generating an alert is unknown.
188 of the alerts were males (65%). Of the 188 male alerts, 64 (34%) were true positives and 124 (66%) were false positives. In relation to females, of 102 alerts, 18
(18%) were true positives and 84 (82%) were false positives. A number of the female false alerts were matched against primarily two individuals who the AFR software provider would refer to as a “lamb”. A lamb is a person whose face has such generic features that may match much more frequently.
Mr Edgell also reviewed the ethnicity of those who were the subject of an alert. Of the true positives (82) 98% were “white north European”. Of the false positives (208) 98.5% were “white north European”.
Mr Edgell therefore concludes, at para. 26:
“From my experience and the information available to me, I have seen no bias based on either gender or ethnicity. …”
In our view, this does not constitute a sufficient answer to the challenge based on the PSED. As Mr Squires submitted, Mr Edgell was dealing with a different set of statistics. He did not know, for obvious reasons, the racial or gender profiles of the total number of people who were captured by the AFR technology but whose data was then almost immediately deleted. In order to check the racial or gender bias in the technology, that information would have to be known. We accept Mr Beer’s submission that it is impossible to have that information, precisely because a safeguard in the present arrangements is that that data is deleted in the vast majority of cases. That does not mean, however, that the software may not have an inbuilt bias, which needs to be tested. In any event, with respect to Mr Edgell, he is not an expert who can deal with the technical aspects of the software in this context.
We should address another submission which was made to us by Mr Beer for SWP although it did not feature in the reasoning of the Divisional Court. This submission arises from the scientific evidence which has been filed in these proceedings.