Image by By Alex Fate
From Anti-CCTV video animation
In January 2020 the Metropolitan Police announced that they would begin "operational use" of Live Facial Recognition (LFR) cameras in London [1]. The Metropolitan Police (the Met) said that what they were doing was perfectly legal and they even published a legal mandate document [2] to prove it. The legality, they claimed, was decreed by the High Court in the far off land of Wales four moons earlier (September 2019), where the South Wales Police Force’s use of facial recognition cameras was rubber stamped as having "a clear and sufficient legal framework" [3].
Just to be clear the police and the High Court judges stated that in the absence of any new act of parliament, statutory instrument or any meaningful public debate the police can increase their powers of interference of the public and scan the faces of people suspected of no wrong-doing whatsoever and run said faces through facial recognition software and compare the faces to the police’s watchlists. This is not to suggest that an act of parliament would make such an interference acceptable – but merely to highlight the sad state of affairs when the police believe they can increase their powers with no recourse to any sort of established legal procedure.
In the aftermath of the Met’s announcement, there were the usual calls for more regulation and for a code of practice specifically in relation to the police use of facial recognition cameras – but as with all such calls for regulation, they merely want to formalise "proper use" and thus remove any discussion of outright ban [4].
In 1985 the Met stated that:
"a respect for citizens' individual rights and freedoms and the avoidance of arbitrary or unlawful action are fundamental to the constitutional meaning of the Rule of Law and thus to the whole meaning and purpose of police duty" [5]
How did we get to a place where the police do not consider subjecting members of the public to automated police line-ups as an arbitrary action and do not consider this disrespectful of a person's individual rights and freedoms?
The Case - Proportionality and Necessity
The court case held in September 2019 looked into whether live facial recognition cameras are a breach of the right to privacy as defined by the European Convention on Human Rights/the Human Rights Act. The case like so many others hinged on the legal concepts of "necessity" and "proportionality." [In August 2020 the case was appealed but the supposed victory looked a lot like defeat as it really just called for stricter guidelines whilst still rubber stamping police use of facial recognition [6].]
The September 2019 judgment stated that South Wales Police (SWP) used facial recognition for a "legitimate aim" and that this aim was important enough to justify interference with Article 8 (privacy) rights (namely the "democratic necessity" defence). It goes on to state that the police use of facial recognition "struck a fair balance" between the rights of the individual and the interests of the community and "was not disproportionate" (the "proportionality" bit). The judgment also states that CCTV cameras without facial recognition could not have achieved the aim of apprehension of suspects wanted on warrant and suspects in the South Wales area (some more "proportionality" for good measure).
Watering "rights" down
The judgment is a turgid sixty-nine page tome which takes an awfully long time to get to the point, but ultimately it all hinges on a proportionality legal formula from a 2013 case relating to sanctions against an Iranian bank, Bank Mellat [[2013] UKSC 38] [7].
The Bank Mellat formula takes the German Administrative Law style three pronged proportionality test [8], adds an additional balance test - "the need to balance the interests of society with those of individuals and groups", and then drops in a margin of appreciation in favour of the state/public body in the necessity test prong to avoid a strict application of a "least restrictive means" test. The reasoning behind this weakening of the test is:
"To allow the legislature a margin of appreciation is also essential if a federal system such as that of Canada, or a devolved system such as that of the United Kingdom, is to work, since a strict application of a "least restrictive means" test would allow only one legislative response to an objective that involved limiting a protected right."
[ Bank Mellat judgment paragraph 75]
In other words because the UK has a devolved legislature, with some powers devolved to England, Northern Ireland, Scotland and Wales, the application of a key part of the proportionality test should be weakened to allow for regional differences.
So how did that work out?
The case was heard in Wales, in relation to the actions of a Welsh police force but the judgment applies to the whole of the UK. It was the Met police in England who were the prime beneficiaries and launched the "operational use" of facial recognition on the back of it, trumpeting to the Mayor of London that "the case confirmed that there is a lawful basis for the police to use facial recognition technology." And in a September 2019 fact sheet the UK Home Office said: "The judgment in the South Wales Police case confirms that there is a clear and sufficient legal framework for the use of LFR in the UK, which includes ensuring that it is necessary and proportionate." [9]
So the "proportionality" test was weakened to account for distinct legal jurisdictions which do not in the event exist. It seems like the Bank Mellat test was sold to us by a used car salesman, we clearly need to open the metaphorical bonnet and see what’s inside.
Well, the Bank Mellat test was derived from a three pronged test in the 1998 De Freitas case [10] with the fourth prong from the 2007 Huang case [11] which in turn was developed from a 1986 Canadian Supreme Court Case R v Oakes [12] (in relation to the Canadian Charter of Rights and Freedoms).
Oh dear. This looks rather like several cars stuck together and not the vintage original we hoped for.
The three prong test of the De Freitas case asks whether:
1. the legislative objective is sufficiently important to justify limiting a fundamental right;
2. the measures designed to meet the legislative objective are rationally connected to it;
3. the means used to impair the right or freedom are no more than is necessary to accomplish the objective."
The second prong of the De Freitas test above comes across as dry and formulaic, so it is worth noting that the Canadian Supreme Court from whom it is borrowed, originally states that the measure: "must not be arbitrary, unfair or based on irrational considerations. In short, they must be rationally connected to the objective". The key phrase thus removed is that a measure must not be "arbitrary", i.e they took "in short" to mean "just take this bit"!
The third prong of the De Freitas test in its original Canadian formulation states that: "there must be a proportionality between the effects of the measures which are responsible for limiting the Charter right or freedom, and the objective which has been identified as of 'sufficient importance’". So in fact the third prong is a type of sub-proportionality of the proportionality – thus adding a proportionality test to a proportionality test!
And now the Bank Mellat four prong test considers if an interference is justified by coldly asking:
1. whether the objective of the measure pursued is sufficiently important to justify the limitation of a fundamental right;
2. whether it is rationally connected to the objective;
3. whether a less intrusive measure could have been used without unacceptably compromising the objective; and
4. whether, having regard to these matters and to the severity of the consequences, a fair balance has been struck between the rights of the individual and the interests of the community.
In applying this test, which became the crux of the entire case, the High Court in Wales started by stating: "It is common ground that there is no issue as regards the first two criteria", a strange and very handy way to reduce the test down to two prongs and to ignore the fundamental issues.
So let’s look at the two prongs they left to test the "proportionality" of facial recognition, starting with number 3, the so-called "least restrictive means" or "least worst" option test.
The court focussed primarily on one of the two South Wales police deployments of facial recognition cameras, at an arms fair in March 2018. The judgment states that "CCTV alone could not have achieved these aims: CCTV could not have identified whether those at the event were on the watchlist". The court has decided that the least restrictive means with which to compare facial recognition cameras is CCTV cameras without facial recognition. This is bizarre as the classic formulation of the least worst test is to ask whether milder means could have been used to achieve the aim. The evident aim here was identifying campaigners at the arms fair who were on a "red watchlist" of apparently six people, some of whom had previously been arrested at the same event. Therefore the least restrictive means should have been a human being looking out for these so-called "subjects of interest". The facial recognition camera is said to have matched one campaigner, no arrest was made, the presence of this campaigner was passed on to the Event Commander and "no further action was taken". Leaving aside the absence of any discussion about whether it is appropriate for the police to operate watchlists of campaigners and surveil said campaigners at all, the job of spotting a known campaigner could have been done by a police officer. In fact controversially this has been police practice for some time via the police’s Forward Intelligence Team (FIT) [13], where police use "spotter cards" to identify campaigners. Surely they should have compared the use of facial recognition cameras to the means that would have been used had such cameras not been used? Why did the court choose to ignore the police use of FIT?
One reason appears to be the way that the police constructed their watchlists. As well as the red watchlist the police also threw in two other watchlists for good measure – an amber watchlist containing 347 persons wanted on warrants and a purple watchlist comprising 161 suspects linked to crimes in the South Wales Police area – and so by padding the facial recognition system with these additional lists (which surely amounts to a fishing expedition) they can create the illusion that a human could not perform the task of spotting this many individuals. Of course the facial recognition system also didn’t spot anyone from the additional watchlists, and it is clear by the categorisation of a watchlist of campaigners as the "red" watchlist, that the system was deployed to perform the task usually undertaken by the FIT. The court then did not consider these "milder", though still controversial, means. Another reason is the deference of courts in the UK to the police, helped along by the margin of appreciation which is afforded the police, precisely as feared by critics such as Cedric Thornberry [14] and as specifically spelt out in the Bank Mellat judgment.
What about the fourth prong of the Bank Mellat proportionality test I hear you cry? Here the court is supposed to consider whether "a fair balance has been struck between the rights of the individual and the interests of the community" - and so they do in their own special way:
Nobody complained as to their treatment (save for the Claimant on a point of principle). Any interference with the Claimant’s Article 8 rights would have been very limited. The interference would be limited to the near instantaneous algorithmic processing and discarding of the Claimant’s biometric data. No personal information relating to the Claimant would have been available to any police officer, or to any human agent. No data would be retained. There was no attempt to identify the Claimant. He was not spoken to by any police officer.
For the other side of the scales, the court said that facial recognition was used for "the specific and limited purpose of seeking to identify particular individuals (not including the Claimant) who may have been in the area and whose presence was of justifiable interest to the police." Further on the extended watchlists they added: "In fact, by including all those who were wanted on warrant there was, potentially, a considerable additional benefit to the public interest, without any impact on the Claimant."
It should be noted the cold almost mathematical analysis performed by the court, suggesting that as computer algorithms work fast any interference with the right to privacy is "near instantaneous" - so interference with an individual’s rights becomes about quantity not quality, representing a major step change in policing tactics. Secondly they only consider the very specific interference of the right of the claimant, an individual, rather than the rights of "the individual" - which is surely what is at stake here because the courts decision rubber stamps the use of facial recognition cameras on all members of the public in the UK. Against this they are happy to weigh the theoretical "public interest" that could "potentially" be of benefit to the community.
This balancing of hypothetical public interest on one side against a specific instance of one person’s experience exposes the woolly nature of proportionality, as pointed out by Stavros Tsakyrakis in a 2008 critique of balancing:
"The most effective critique of balancing concerns the assumption of a common metric in the weighing process. The metaphor says nothing about how various interests are supposed to be weighted and this silence reflects the impossibility of measuring incommensurable values by introducing a mechanistic, quantitative common metric." [15]
Thus the court was able to award all of the Bank Mellat prongs to the police and declare that facial recognition did not breach the right to privacy.
And the two prongs that were not considered? As detailed above prong number two was derived from another case where it stated that an interference "must not be arbitrary, unfair or based on irrational considerations". One would think it worthy of discussion whether subjecting people suspected of no wrong doing whatsoever to what amounts to an automated police line up is an arbitrary action. Likewise surely the question in the first prong, whether the "objective of the measure pursued is sufficiently important" to subject people suspected of no wrong doing whatsoever to such a police line up, should have been a major focus of the case. The absence of such a discussion is at the heart of why the courts are such bad custodians of our freedoms. They seek to apply cold dry formulas to address issues that demand rigorous debate. They avoid philosophical discussion of right and wrong, the supposed foundations of the legal system, preferring to consider the duration of algorithmic processing of a facial image over the arbitrary nature of subjecting people not suspected of any wrong-doing to arbitrary searches to check if they are on a watchlist. And finally they favour the person attacking the freedoms, the police, who are afforded discretion, margin of appreciation and the benefit of the doubt – because under the Human Rights construct it is the police or the state who are seen as the party who should benefit from the principles of self-defence and it is the individual’s rights that are seen as the attacker that can be repelled when the state or its proxy decides it is "necessary" and "proportionate".
As Stavros Tsakyrakis puts it:
"The problem with the rhetoric of balancing in the context of proportionality is that it obscures the moral considerations that are at the heart of human rights issues and thus deprives society of a moral discourse that is indispensable."
Society has been deprived of a moral discourse on the use of facial recognition cameras. Furthermore, in Europe and now in the UK, Human Rights cases have no jury and so a fundamental part of the constitutional protection of freedoms is switched off to once again favour the state over the individual.
Endnotes:
- [ 1] https://web.archive.org/web/20200124212135/http://news.met.police.uk/news/met-begins-operational-use-of-live- facial-recognition-lfr-technology-392451
- [ 2] Met Police Legal Mandate
https://www.met.police.uk/SysSiteAssets/media/downloads/force-content/met/advice/lfr/mps-lfr-legal-mandate-v1- 1.pdf - [ 3] Bridges, R (On Application of) v The Chief Constable of South Wales Police [2019] EWHC 2341
https://www.bailii.org/ew/cases/EWHC/Admin/2019/2341.html - [ 4] https://www.theyworkforyou.com/whall/?id=2019-11-05a.229.0#g247.1
- [ 5] Respect for Individual Rights, The Principles of Policing and Guidance for Professional Behaviour. Metropolitan
Police, 1985
https://law-journals-books.vlex.com/vid/policing-principles-metropolitan-police-51759325 - [ 6] https://www.thejusticegap.com/automatic-facial-recognition-secures-a-victory-in-all-but-name/
- [ 7] Bank Mellat v Her Majesty's Treasury (No. 1) [2013] UKSC 38 (19 June 2013)
https://www.bailii.org/uk/cases/UKSC/2013/38.html - [ 8] "Three Aspects of Proportionality", Margit Cohn, paper presented at VIII Congress of the International Association
of Constitutional Law, Mexico City, December 2010
'Proportionality Balancing and Global Constitutionalism', Alec Stone Sweet and Jud Mathews, 2008
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=1569344 - [ 9] 'Fact Sheet on live facial recognition used by police’, Home Office
https://homeofficemedia.blog.gov.uk/2019/09/04/fact-sheet-on-live-facial-recognition-used-by-police/ - [10] de Freitas case, [1998] 3 WLR 675
https://www.bailii.org/uk/cases/UKPC/1998/30.html - [11] Huang case, [2007] 2 AC 167
https://www.bailii.org/uk/cases/UKHL/2007/11.html - [12] R. v. Oakes
https://scc-csc.lexum.com/scc-csc/scc-csc/en/item/117/index.do - [13] https://web.archive.org/web/20110524062304/http://www.fitwatch.org.uk/forward-intelligence-teams/fit-the-role/
- [14] See Memorandum of evidence by Cedric Thornberry, Governor of the British Institute of Human Rights, paragraph 21,Minutes of evidence taken before the Select Committee on a Bill of Rights (p224-239), HMSO 1977
- [15] 'Proportionality: An Assault on Human Rights?’, Stavros Tsakyrakis
http://jeanmonnetprogram.org/wp-content/uploads/2014/12/080901.pdf
