Appendix C: National Conversation on Live Facial Recognition – Summary of Written Submissions

Introduction

In addition to the five focus groups held with representative organisations and stakeholders and a national survey, written submissions were received as part of the national conversation.

Details of submissions

The majority of written submissions provided further detail on why these organisations were not supportive of the implementation of live facial recognition (LFR) by Police Scotland. A summary of key points includes
the following:

The need for primary legislation and that the current legislative framework does not provide sufficient legal basis for LFR. This also includes how LFR should be used, for what purposes, and how it would be evaluated and scrutinised.

Overall impact on human rights, including Articles 5, 6, 8, 10, 11,14 of the ECHR and Articles 2 and 3 of the Universal Declaration of Human Rights, and concerns that the impact on these rights cannot be balanced with necessity and proportionality.

Incompatibility with Public Sector Equality Duties.

Potential ‘chilling effect’ impacting individuals’ choice in exercising their freedom of right to assembly and expression.

Concerns around the overall accuracy of LFR technology, and that accuracy and bias disproportionately impacts certain communities. This specifically relates to research indicating potential racial and gender biases in the technology, as well as potential inaccuracies in relation to trans and non-binary persons. There were also concerns on the impact of low lighting and face coverings on the accuracy of LFR. This was specifically relating to disabled or immunocompromised people who may wear face masks for their safety or those who may wear face coverings for faith-related or cultural reasons.

Potential misidentification of women and individuals from Black and Minority Ethnic (BME) backgrounds. Concerns were also raised in relation to people who are transgender who may not carry identification that corresponds with how they identify.

Negative impacts on relationships between Police Scotland and certain communities, particularly individuals with intersecting negative relationships with the police. The potentially negative impact in relationships and trust in policing was specifically reiterated with regards to LGBTIA+ and black and minority ethnic
communities.

Impact on children and young people.

That human-in-the loop decision making is not a sufficient guardrail.

Limited evidence for the use of LFR as a deterrent of crime, particularly in relation to the proposed use cases. In addition, how LFR will function as a deterrent or preventative measure when not all individuals potentially likely to cause harm may be on watchlists.

Clarity on how watchlists are defined and managed, and potential for racial disparities of watchlists.

Information governance, data protection and consent to data processing.

Admissibility of LFR evidence in court trials.

Scope creep beyond identified use cases.

An overreliance on technology, reducing police presence at protests with a preference for alternative policing approaches.

References to LFR use by other police forces in the UK and concerns around evidence of its efficacy.

Clarity of cost vs benefits for the use of LFR.

Concerns that public polling referenced in the paper may present an inaccurate public opinion due to a lack of understanding of LFR by the general public.

Comment on the discussion paper produced by the Authority and Police Scotland, relating to a misquote that has since been corrected, and comments on the overall paper.

Whilst most submissions were clear in their stance that LFR should not be implemented, some safeguards were suggested. Safeguards suggested across two written submissions included:

Authorisation: Independent authorisation, including judiciary sign off or a two-stage process by public bodies. Have a clear authorisation environment (application based, assessment/approval by senior officer of lawfulness, proportionality, and necessity).
Development of detailed authorisation records.

Guardrails and transparency: Ensuring that independent safeguards are in place (e.g. Information Commissioner’s Office, Scottish Biometrics Commissioner, Investigatory Powers Commissioner's Office) and that any decision is always made by a human. There should be a level of public reporting of the outcomes

Technology limitations: Ensuring that any solution has been independently certified as non-discriminatory.

Independent evaluation: Any implementation should not be ‘self-proclaimed’ as a success. Suggestion that LFR is independently evaluated 12 months after any introduction.

Watchlists: Individuals should not be added to watchlists on suspicion alone, and there should be continual review and oversight of watchlists. Having clear policy and processes around how watchlists are populated and approved, alongside image quality standards.

Right to be informed: Data subjects must be informed of the processing of their data. Requirement to consider a strategy to convey this (e.g. example signage at deployments).

Public and political acceptability thresholds: Any policy should be based on the principle of policing by consent and should mirror acceptability thresholds. For example, if the public were opposed to the use of LFR on fixed town centre CCTV. If any further work or conversation is undertaken, meaningful and extensive public consultation with stakeholders, civil society organisations and the general public. Co-design and two-way feedback with respondents and the production of more detailed evidence to support respondents.

Right to redress: A right of redress recognising that errors may be made by the technology or those in a decision-making capacity.