Logo Assembly Logo Hemicycle

Preventing discrimination caused by the use of artificial intelligence

Resolution 2343 (2020)

Parliamentary Assembly
Text adopted by the Standing Committee, acting on behalf of the Assembly, on 22 October 2020 (see Doc. 15151, report of the Committee on Equality and Non-Discrimination, rapporteur: Mr Christophe Lacroix).See also Recommendation 2183 (2020).
1 Artificial intelligence (AI) is transforming the way we live. It enables massive upscaling of processes and is already used by a wide range of private and public entities, across fields as diverse as selection procedures that determine access to employment and education, the evaluation of individual entitlements to welfare payments or to credit, or the targeting of advertising and news.
2 Many uses of AI can have a direct impact on equality of access to fundamental rights, including the right to private life and the protection of personal data; access to justice and the right to a fair trial, in particular as regards the presumption of innocence and the burden of proof; access to employment, education, housing and health; and access to public services and welfare. The use of AI has been found to cause or exacerbate discrimination in these fields, leading to denials of access to rights that disproportionately affect certain groups – often women, minorities and those who are already the most vulnerable and marginalised. Its use in information flows has also been linked to the spread of online hate that spills over into all other social interactions.
3 Machine learning used to build AI-based systems relies on vast datasets (big data), much of which is personal data. Effective guarantees of the protection of personal data remain essential in this context. At the same time, data are by nature biased, as they reflect discrimination already present in society as well as the bias of those who collect and analyse them. Choices about which data to use and which to ignore in AI-based systems, as well as a lack of data on key issues, the use of proxies and the difficulties inherent in quantifying abstract concepts, can also lead to discriminatory results. Biased datasets are at the heart of many cases of discrimination caused by the use of AI and remain a major issue to be resolved in this field.
4 The design and purpose of AI-based systems are also crucial. Algorithms optimised for efficiency, profitability or other objectives, without taking due account of the need to guarantee equality and non-discrimination, may cause direct or indirect discrimination, including discrimination by association, on a wide variety of grounds, including sex, gender, age, national or ethnic origin, colour, language, religious convictions, sexual orientation, gender identity, sex characteristics, social origin, civil status, disability or health status. This makes it especially important, wherever the use of AI-based systems may have an impact on access to fundamental rights, that, from the outset, full respect for equality and non-discrimination is taken into account in their design and that they are rigorously tested before being deployed, as well as regularly after their deployment, in order to ensure that these rights are guaranteed.
5 The complexity of AI systems, and the fact that they are frequently developed by private companies and treated as their intellectual property, can lead to serious issues of transparency and accountability regarding decisions made using these systems. This can make discrimination extremely difficult to prove and can hinder access to justice, in particular where the burden of proof is placed on the victim and/or where the machine is assumed by default to have made the correct decision, violating the presumption of innocence.
6 The lack of diversity in many technology companies and professions heightens the risk that AI systems will be developed without due regard to their potentially discriminatory impacts on some individuals and groups in society. Women and minorities’ access to science, technology, engineering and mathematics (STEM) professions needs to be improved, and a true culture of respect for diversity urgently needs to be developed within these professional circles. Using interdisciplinary and intercultural approaches throughout all stages of the design of AI systems would also contribute to strengthening them from the standpoint of equality and non-discrimination.
7 Finally, strong, clear and universally accepted and applicable ethical principles must underpin the development and deployment of all AI-based systems. The Parliamentary Assembly considers that these principles can be grouped under the following broad headings: transparency, including accessibility and explicability; justice and fairness, including non-discrimination; human responsibility for decisions, including liability and the availability of remedies, safety and security; and privacy and the protection of personal data.
8 The Assembly welcomes the fact that both public and private actors have begun to examine and develop ethical and human rights standards applicable to the use of AI. It welcomes in particular the Committee of Ministers’ Recommendation Rec/CM(2020)1 on the human rights impacts of algorithmic systems, along with its accompanying Guidelines, and the recommendation of the Council of Europe Commissioner for Human Rights entitled “Unboxing Artificial Intelligence: 10 steps to protect Human Rights”. The Assembly endorses the general proposals made in these texts for application also in the areas of equality and non-discrimination.
9 The Assembly stresses that legislators must not hide behind the complexities of AI to prevent the introduction of regulations designed to protect and promote equality and non-discrimination in this field: the human rights issues at stake are clear and require action. In addition to ethical principles, procedures, tools and methods for regulating and auditing AI-based systems are needed in order to ensure their compliance with international human rights standards, and in particular with the rights to equality and non-discrimination. Given the strong transnational and international dimensions of AI-based technologies, international standards also appear to be needed in this field.
10 In the light of these considerations, the Assembly calls on member States to:
10.1 review their anti-discrimination legislation and amend it as necessary, so as to ensure that it covers all cases where direct or indirect discrimination, including discrimination by association, may be caused by the use of AI, and that complainants have full access to justice; in the latter respect, pay particular attention to guaranteeing the presumption of innocence and ensuring that victims of discrimination do not face a disproportionate burden of proof;
10.2 draw up clear national legislation, standards and procedures to ensure that AI-based systems comply with the rights to equality and non-discrimination wherever the enjoyment of these rights may be affected by the use of such systems;
10.3 ensure that equality bodies are fully empowered to address issues of equality and non-discrimination that arise due to the use of AI, and to support individuals bringing cases in this field, and that equality bodies have all the necessary resources to carry out these tasks.
11 In order to ensure that the use of AI-based technologies by public authorities is subject to adequate parliamentary oversight and public scrutiny, the Assembly calls on national parliaments to:
11.1 make the use of such technologies a part of regular parliamentary debates, and ensure that an adequate structure for such debates exists;
11.2 require the government to notify the parliament before such technology is deployed;
11.3 require the use of such technologies by the authorities to be systematically recorded in a public register.
12 In order to address underlying issues of diversity and inclusion in the field of AI, the Assembly further calls on member States to:
12.1 promote the inclusion of women, girls and minorities in STEM education paths, from the earliest ages and to the highest levels, and work together with industry to ensure that diversity and inclusion are fostered throughout career paths;
12.2 support research into data bias and the means by which its impact can be effectively countered in AI-based systems;
12.3 promote digital literacy and access to digital tools by all members of society.
13 The Assembly invites all entities, both public and private, working on and with AI-based systems, to ensure that respect for equality and non-discrimination is integrated from the outset in the design of such systems, and adequately tested before their deployment, wherever these systems may have an impact on the exercise of or access to fundamental rights. To this end, it invites these entities to consider building capacity for a human rights impact assessment framework for the development and deployment of AI-based systems by both private and public entities. In addition, it encourages the use of interdisciplinary and diverse teams at all stages of the development and deployment of AI-based systems.
14 Finally, the Assembly calls on national parliaments to support work being carried out at international level, in particular through the Council of Europe’s Ad hoc Committee on Artificial Intelligence (CAHAI), to ensure that international human rights standards are effectively applied in the field of AI, and that respect for the principles of equality and non-discrimination is guaranteed in this field.