IGF 2018 WS #302 The Toronto Declaration on machine learning: the way forward

Organizer 1: Civil Society, Western European and Others Group (WEOG)

Artificial intelligence, and specifically machine learning, is increasing in use in numerous sectors - yet these relatively new systems are already generating or contributing to concrete discriminatory harms. , From scoring systems used in the criminal justice which discriminate against people of color, to facial recognition systems that are more accurate when trying to identify white men rather than women of color, to the use of algorithms for recruitment or education processes which have disadvantaged candidates with a potential health issue or from lower income from jobs or university placement, the examples of discrimination are piling up. We must acknowledge and address the risks of these technologies reinforcing or even institutionalising inequalities and discrimination and put safeguards in place.

Faced with the reality of these early impacts on human rights, Access Now and Amnesty International felt the urgent need to develop a document anchored in human rights law that would address the issue and present recommendations to states and the private sector on how to design, develop and implement machine learning in a way that prevents and mitigate the risk of discrimination and promotes equality. The Toronto Declaration was the product of collaborative work between industry, government, civil society and academia, and was launched at launched at RightsCon Toronto in May 2018.

Format: 

Birds of a Feather - 60 Min

Interventions: 

Representatives from Access Now and Amnesty International will provide an overview of the work around the Declaration since its launch, including where it has had influence,, the current signatories, and concrete uses of the document (10 min). Then participants to the session will be invited to discuss priorities for the implementation of the recommendations in the Declaration from a user, developer and government perspective (15 min for each aspect).

Based on the “bird of a feather” model, Access Now and Amnesty International will invite participants and experts on artificial intelligence from academia, the private sector, government and civil society to participate to the discussion.

Diversity: 

The session will ensure participation from civil society, academia, government and private sector to reflect a diversity of viewpoints. Gender balance and diverse geographical representation will also be ensured.

Following the presentation of the work done with the Declaration since its launch at RightsCon Toronto in May 2019, speakers from industry, governments, academia and civil society will be invited to this informal session brainstorm ways to practically implement the recommendations highlighted in the Declaration and reflect on challenges and opportunities to address the impact of discrimination in machine learning.

Discussion Facilitation: 

Access Now and Amnesty International will facilitate the discussion in the room. One moderator will be assign a specific segment of the discussion on recommendations for users, developers and governments while the fourth person will be tasked with online moderation.

Participants to the session will be shared in advance with the Declaration and a broad agenda of the session open for feedback to prepare and foster discussion ahead of the session.

Online Participation: 

The organiser will set up a public discussion pad for remote participants to contribute to the brainstorming session.

Contact Information

United Nations
Secretariat of the Internet Governance Forum (IGF)

Villa Le Bocage
Palais des Nations,
CH-1211 Geneva 10
Switzerland

igf [at] un [dot] org
+41 (0) 229 173 678