Highlights | Open internet and inclusive technology | Artificial intelligence (AI) | Inclusive technologies | Online tracking industry / AdTech | Profiling practices

UN Special Rapporteur analyses AI’s impact on human rights

In October 2018, the United Nations (UN) Special Rapporteur for the promotion and protection of the right to freedom of opinion and expression, David Kaye, released his report on the implications of artificial intelligence (AI) technologies for human rights. The report was submitted to the UN General Assembly on 29 August 2018 but has only been published recently.

By EDRi · November 7, 2018

The text focuses in particular on freedom of expression and opinion, privacy and non-discrimination. In the report, the UN Special Rapporteur David Kaye first clarifies what he understands by artificial intelligence and what using AI entails for the current digital environment, debunking several myths. He then provides an overview of all potential human rights affected by relevant technological developments, before laying down a framework for a human rights-based approach to these new technologies.

----------------------------------------------------------------- Support our work - make a recurrent donation! https://edri.org/supporters/ -----------------------------------------------------------------

1. Artificial intelligence is not a neutral technology

David Kaye defines artificial intelligence as a “constellation of processes and technologies enabling computers to complement or replace specific tasks otherwise performed by humans” through “computer code […] carrying instructions to translate data into conclusions, information or outputs.” He states that AI is still highly dependent on human intervention, as humans need to design the systems, define their objectives and organise the datasets for the algorithms to function properly. The report points out that AI is therefore not a neutral technology, as the use of its outputs remains in the hands of humans.

Current forms of AI systems are far from flawless, as they demand human scrutiny and sometimes even correction. The report considers that AI systems’ automated character, the quality of data analysis as well as systems’ adaptability are sources of bias. Automated decisions may produce discriminatory effects as they rely exclusively on specific criteria, without necessarily balancing them, and they undermine scrutiny and transparency over the outcomes. AI systems also rely on huge amounts of data that has questionable origins and accuracy. Furthermore, AI can identify correlations that can be mistaken for causations. David Kaye points at the main problem of adaptability when losing human supervision: it poses challenges to ensuring transparency and accountability.

2. Current uses of artificial intelligence interfere with human rights

David Kaye describes three main applications of AI technology that pose important threats to several human rights.

The first problem raised is AI’s effect on freedom of expression and opinion. On one hand, “artificial intelligence shapes the world of information in a way that is opaque to the user” and conceals its role in determining what the user sees and consumes. On the other, the personalisation of information display has been shown to reinforce biases and “incentivize the promotion and recommendation of inflammatory content or disinformation in order to sustain users’ online engagement”. These practices impact individuals’ self-determination and autonomy to form and develop personal opinions based on factual and varied information, therefore threatening freedom of expression and opinion.

Secondly, similar concerns can be raised in relation to our right to privacy, in particular with regard to AI-enabled micro-targeting for advertisement purposes. As David Kaye states, profiling and targeting users foster mass collection of personal data, and lead to inferring “sensitive information about people that they have not provided or confirmed”. The few possibilities to control personal data collected and generated by AI systems put into question the respect of privacy.

Third, the Special Rapporteur highlights AI as an important threat to our rights to freedom of expression and non-discrimination due to AI’s increasingly-allocated role in the moderation and filtering of content online. Despite some companies’ claims that artificial intelligence can support exceeded human capacities, the report sees the recourse to automate moderation as impeding the exercise of human rights. In fact, artificial intelligence is unable to resist discriminatory assumptions or to grasp sarcasm and the cultural context for each piece of content published. As a result, freedom of expression and our right not to be discriminated against can be severely hampered by delegating complex censorship exercises to AI and private actors.

3. A set of recommendations for both companies and States

Recalling that “ethics” is not a cover for companies and public authorities to neglect binding and enforceable human rights-based regulation, the UN Special Rapporteur recommends that “any efforts to develop State policy or regulation in the field of artificial intelligence should ensure consideration of human rights concerns”.

David Kaye suggests human rights should guide development of business practices, AI design and deployment and calls for enhanced transparency, disclosure obligations and robust data protection legislation – including effective means for remedy. Online service providers should make clear which decisions are made with human review and which by artificial intelligence systems alone. This information should be accompanied by explanations of the decision-making logic used by algorithms. Further, the “existence, purpose, constitution and impact” of AI systems should be disclosed in an effort to improve the level of individual users’ education around this topic. The report also recommends to make available and publicise data on the “frequency at which AI systems are subject to complaints and requests for remedies, as well as the types and effectiveness of remedies available”.

States are identified as key actors responsible for creating a legislative framework hospitable to a pluralistic information landscape, preventing technology monopolies and supportive of network and device neutrality.

Lastly, the Special Rapporteur provides useful tools to oversee AI development:

  1. human rights impact assessments performed prior, during and after the use of AI systems;
  2. external audits and consultations with human rights organisations;
  3. enabled individual choice thanks to notice and consent;
  4. effective remedy processes to end human rights violations.

UN Special Rapporteur on Freedom of Expression and Opinion Report on AI and Freedom of Expression (29.08.2018)

Civil society calls for evidence-based solutions to disinformation

(Contribution by Chloé Berthélémy, EDRi intern)