面部识别技术之负责任使用框架：流程管理用例（英文版）_20页_260kb

Summary of "A Framework for Responsible Limits on Facial Recognition"

Core Content

This document outlines a policy framework aimed at ensuring the responsible use of facial recognition technology, with a specific focus on flow management as a use case. The World Economic Forum (WEF) is leading a multistakeholder, evidence-based policy project in France to test and refine this framework. The goal is to create a governance mechanism that balances innovation with ethical and legal considerations.

Main Objectives

• Develop a policy framework for the responsible use of facial recognition technology.
• Test the framework in a real-world use case (flow management).
• Encourage multistakeholder collaboration and third-party audits to ensure compliance.
• Support public debate and policy development at national, European and international levels.

Key Use Cases

The document identifies several key use cases for facial recognition technology, including:

• Face access: Used for access to services, such as public transport, buildings, and digital platforms.
• Safety and security of public spaces: Includes law enforcement, border control, and surveillance in public areas.
• Marketing and customer services: Encompasses personalized advertising, emotion recognition, and customer tracking.
• Healthcare services: Used for patient authentication and health monitoring.

These use cases are explored to understand the trade-offs between benefits and risks, and to test the policy framework in various contexts.

Principles for Action

The framework is built around 11 principles designed to guide the ethical and legal use of facial recognition technology. These principles include:

1. Be lawful: Ensure compliance with national and regional laws, including GDPR in the EU.
2. Be audited by a third party: Use an independent audit framework to validate compliance.
3. Report to an oversight body: Engage with regulatory authorities to ensure transparency and accountability.
4. Run an impact assessment: Conduct assessments on the social, legal and ethical impacts of use cases.
5. Bias and discrimination: Identify, measure and mitigate biases in the system.
6. Proportional use: Ensure facial recognition systems are appropriate for their intended purpose.
7. Privacy by design: Integrate privacy considerations into the system's design and operation.
8. Accountability: Establish a culture of accountability and public governance principles.
9. Risk assessment and audit: Evaluate risks related to privacy, errors, biases, and security.
10. Performance: Follow accuracy and performance standards for both lab and field tests.
11. Right for information and consent: Ensure users are informed and provide explicit consent.

Methodology

The framework was developed through a four-step methodology:

1. Define principles for action: Establish a shared understanding of responsible use.
2. Design methodologies by use case: Create best practices for responsible system design.
3. Assess compliance: Use an assessment questionnaire tailored to the use case.
4. Validate through audit: Implement an audit framework by a trusted third party to ensure compliance.

The framework is intended to be tested on a case-by-case basis and is expected to evolve based on the pilot project outcomes.

Pilot Project on Flow Management

The flow management use case was selected as the first test scenario due to its potential for growth and clear risk profile. The pilot is expected to last 18 months, with the following phases:

• Scoping (April–September 2019): Identify relevant applications and stakeholders.
• Co-designing (October 2019–January 2020): Develop the framework, including principles, best practices, assessment tools and audit mechanisms.
• Testing (February–July 2020): Apply the framework to the flow management use case.
• Deploying: Support the implementation of the framework in various scenarios, such as company endorsement, certification processes, and legislative adoption.

Best Practices for Responsible Facial Recognition Systems

The document outlines a set of best practices for the design and deployment of facial recognition systems, emphasizing:

• Justifying the use of facial recognition based on the problem to be solved.
• Designing data plans that reflect the characteristics of end users.
• Mitigating bias through evaluation and documentation.
• Informing end users and being transparent about the system's purpose and data handling.

Assessment Questionnaire

A first version of the assessment questionnaire is provided, aligned with the principles for action. It includes questions related to:

• Definitions of bias and metrics for evaluation.
• Risk analysis and prioritization.
• Existing practices for bias mitigation.
• Action plans and effectiveness evaluation.
• Test cases and acceptance criteria.

The questionnaire is a tool for organizations to operationalize the principles in their systems and is expected to evolve as the pilot progresses.

Conclusion

The framework aims to empower citizens and representatives to make informed decisions about the use of facial recognition technology. It emphasizes the need for ethical, legal and technical considerations in the development and deployment of such systems. The WEF encourages collaboration and participation in this policy project to ensure a balanced and inclusive approach to facial recognition technology.

This white paper represents the first step in an iterative process, with the ultimate goal of creating a sustainable governance framework for facial recognition.