Previous 51 / 54 Next

10. Our approach to human rights and responsible product development Five Pillars of Responsible AI Responsible AI at Meta is built on five pillars: • Privacy and security • Fairness and inclusion • Robustness and safety • Transparency and control • Accountability and governance We’ve made progress in addressing many of these issues, and there remains much work to be done. In 2022, we worked to: • Expand language support • Our open-source No Language Left Behind effort harnesses AI to deliver high-quality translations directly between 200 languages. • Our Massively Multilingual Speech (MMS) recognition model supports speech-to-text and text-to-speech for 1,107 languages and supports language identification for over 4,000 languages. Some of the supported languages, such as Tatuyo, have only a few hundred speakers, and for most of these languages, no prior speech technology exists. We publicly share the models and code. • Wikipedia teamed up with Meta AI to fact-check citations using Meta’s open-sourced AI program, Sphere. • Address bias • In our drive toward ensuring that AI systems treat everyone fairly, we introduced and open-sourced several datasets and models responsible for addressing this gap in privacy-protected ways. • We trained an AI model for reducing demographic biases in text — a demographic text perturber — which can help break stereotypical associations present in natural language programing (NLP) data. • We introduced and open-sourced two new datasets to help measure fairness and mitigate potential bias in NLP models. These datasets include a more comprehensive representation of different demographic dimensions to measure fairness in these models, including terms for gender identity, age, race, and people with disabilities. • Better understand problematic content associations • We assembled a cross-disciplinary team, including people from our Civil Rights, Human Rights, Engineering, RAI, AI Research, Policy, and Product teams, to better understand problematic content associations in several of our end-to-end systems. We worked to Human rights report 51

Select target paragraph3

● ● ●

Human rights report - Insights and actions 2022

Report

Preview

Language

English

Author(s)

META

Date Published: Sep 20, 2023

Abstract: This is Meta’s second annual Human Rights Report, covering the calendar year 2022. It demonstrates how we are living up to the commitments made in our Corporate Human Rights Policy, which is based on the United Nations Guiding Principles on Business and Human Rights (UNGPs). The report shows how we are building on our work to respect human rights, how we assess potential impacts that could be linked to our services, and what we are doing to mitigate these risks.

Over the years, Meta’s services and apps have helped grassroots movements grow rapidly, challenge orthodoxies, advocate for rights, and in doing so, change the world. 2022 was no exception. As new challenges emerge, our human rights work evolves to address such issues as the use of our services in countries in conflict, threats to an open internet, and the rise of artificial intelligence.

We have strengthened our governance systems to advance our work toward respecting human rights across all our products and services. This includes continuing to empower the Oversight Board, which published 12 content moderation decisions and made 91 recommendations in 2022. Our efforts to strengthen governance also include training our employees. During the year, we launched human rights training titled Bigger than Meta: Human Rights company-wide.

In 2022, Meta passed its Global Network Initiative assessment, which included an examination of our policies and processes as well as eight specific illustrative case studies involving government takedown requests and data requests.

Human rights due diligence is central to our Corporate Human Rights Policy. In this report, we disclose key findings from our Comprehensive Human Rights Salient Risk Assessment (CSRA) and provide an update on our implementation of the recommendations borne of the due diligence work.

In 2022, we conducted the CSRA in partnership with Business for Social Responsibility (BSR), an independent organization with expertise in human rights. The analysis identified eight potential salient risks as highest priority, based on UNGP criteria. These were 1.) Freedom of Opinion and Expression; 2.) Privacy; 3.) Equality and Non-Discrimination; 4.) Life, Liberty, and Security of Person; 5.) Best Interests of the Child; 6.) Public Participation, to Vote, and to Be Elected; 7.) Freedom of Association and Assembly; and 8.) Right to Health. The assessment found that the potential risks relating to how Meta delivers our services and profit models are varied and nuanced.

Link: Source

Primary Documents

2022-Meta-Human-Rights-Report.pdf
English
Download View

Myanmar Internet Project

Select target paragraph3

Human rights report - Insights and actions 2022

Primary Documents

Notifications