[Paper Review] Understanding bias in facial recognition technologies

David Leslie|arXiv (Cornell University)|Sep 26, 2020

Ethics and Social Impacts of AI21 citations

TL;DR

This paper examines how historical patterns of discrimination have embedded bias into facial recognition technologies (FDRTs), leading to distributional and recognitional injustices. It analyzes the sociotechnical origins of bias in FDRTs, evaluates ethical implications of pervasive surveillance, and proposes governance frameworks for more responsible development and deployment of these systems.

ABSTRACT

Over the past couple of years, the growing debate around automated facial recognition has reached a boiling point. As developers have continued to swiftly expand the scope of these kinds of technologies into an almost unbounded range of applications, an increasingly strident chorus of critical voices has sounded concerns about the injurious effects of the proliferation of such systems. Opponents argue that the irresponsible design and use of facial detection and recognition technologies (FDRTs) threatens to violate civil liberties, infringe on basic human rights and further entrench structural racism and systemic marginalisation. They also caution that the gradual creep of face surveillance infrastructures into every domain of lived experience may eventually eradicate the modern democratic forms of life that have long provided cherished means to individual flourishing, social solidarity and human self-creation. Defenders, by contrast, emphasise the gains in public safety, security and efficiency that digitally streamlined capacities for facial identification, identity verification and trait characterisation may bring. In this explainer, I focus on one central aspect of this debate: the role that dynamics of bias and discrimination play in the development and deployment of FDRTs. I examine how historical patterns of discrimination have made inroads into the design and implementation of FDRTs from their very earliest moments. And, I explain the ways in which the use of biased FDRTs can lead distributional and recognitional injustices. The explainer concludes with an exploration of broader ethical questions around the potential proliferation of pervasive face-based surveillance infrastructures and makes some recommendations for cultivating more responsible approaches to the development and governance of these technologies.

Motivation & Objective

• To investigate how systemic racism and historical discrimination have influenced the design and deployment of facial recognition technologies.
• To analyze the consequences of biased FDRTs in terms of distributional and recognitional injustices.
• To examine the ethical implications of expanding face surveillance infrastructures in democratic societies.
• To propose actionable recommendations for responsible development and governance of facial recognition technologies.

Proposed method

• Conducts a critical socio-technical analysis of facial recognition systems, tracing bias from historical and structural inequalities.
• Examines the technical pipeline of FDRTs, from data collection to model deployment, to identify points of bias introduction.
• Uses ethical frameworks to assess the impact of biased systems on marginalized communities and civil liberties.
• Evaluates real-world case studies of facial recognition misuse to illustrate systemic risks.
• Proposes governance principles for ethical development, including transparency, accountability, and oversight mechanisms.
• Integrates insights from computer science, sociology, and ethics to advocate for human-centered design in FDRTs.

Experimental results

Research questions

• RQ1 How have historical patterns of discrimination influenced the development of facial recognition technologies?
• RQ2 In what ways do biased FDRTs lead to distributional and recognitional injustices?
• RQ3 What are the societal risks of expanding pervasive face-based surveillance infrastructures?
• RQ4 How can ethical governance frameworks be designed to mitigate harm in FDRT deployment?
• RQ5 What role do data collection practices and algorithmic design play in perpetuating bias?

Key findings

• Bias in facial recognition technologies is not incidental but rooted in historical and structural inequalities, particularly in data collection and labeling processes.
• FDRTs disproportionately misidentify women and people of color, leading to tangible harms such as wrongful arrests and surveillance overreach.
• The expansion of face surveillance threatens to erode democratic norms by enabling unchecked monitoring and undermining privacy and autonomy.
• Recognitional injustice occurs when individuals are misclassified or denied recognition due to algorithmic bias, affecting access to services and rights.
• Current systems often lack transparency and accountability, making it difficult to challenge or correct biased outcomes.
• Ethical governance requires interdisciplinary collaboration, inclusive design, and enforceable oversight to prevent harm and ensure equity.