RightsCon Tunis 2019

Algorithmic decision-making has become synonymous with inexplicable decision-making, often described as a black box. The panel will unpack and discuss AI explainability from both a technical and a legal angle. Speakers will touch on what makes algorithms so difficult to explain, look into what technical and legal solutions are being proposed to mitigate this problem, and the Intuitive Appeal of Explainable Machines.

Artificial Intelligence (AI) refers to a category of computer systems which perform tasks more efficiently and make decisions with a degree of autonomy that normally require human intelligence. There is a lot of discussion around “ethical AI” at the moment, which has lead to the publication of several papers, principles and studies recently on the ethical use of AI. The purpose of this panel is to take a pragmatic perspective on how AI technology is being deployed in the telecommunications sector. We will explore the current safeguarding mechanisms on the ethical and rights respecting use of AI in telecommunications, and aim to draw best practise and evaluate whether the current mechanisms are seen as sufficient from stakeholders' perspective. Questions to be addressed: How is AI used inside the telecoms companies and in the customer interface? How can the UNGPs be applied to the design and use of AI? How should corporate transparency be defined in the use of AI? Are there overarching themes in the current corporate principles on the use of AI which could be combined into industry principles?

This session will explore how human rights frameworks can be integrated into the governance of artificial intelligence and machine learning technologies. Building on international human rights law and the UN Guiding Principles on Business and Human Rights, it will discuss the value the existing human rights framework brings to the conversation about “ethical AI.” With a diverse panel drawn from international organisations, civil society, academics, and private sector, the discussion will explore the varying impact of these technologies in different countries and highlight both recent accomplishments and where further research is needed to inform policy development. The rapid development and deployment of artificial intelligence has many stakeholders concerned, and in the past several years, we have witnessed a proliferation of principles documents that attempt to guide the ethical and rights-respecting development and deployment. These declarations may each have individual value, but do they represent an emerging consensus about the proper use of AI/ML tools? If a consensus is emerging, does it reflect the reality of how AI tools are being deployed globally and cross-culturally, in non-democratic states as well as democratic ones? Do current principles reflect international human rights norms, and how do they anticipate and address governance challenges?

Online platforms are increasingly facing calls from lawmakers and regulators to address harmful content of various types, and also frequently wish to moderate their own content for their own terms of use and policies. The scale of the content on many platforms makes some form of algorithmic function a necessary part of content moderation, and recently interest has grown in using AI technologies such as machine learning to play a part too. The goal of this session is to exchange views between expert panelists, and the public, on the potential for AI to be used in content moderation, while respecting fundamental rights.

In July 2018, Microsoft shared its call for government regulation and responsible industry governance of facial recognition technology. This technology clearly brings important societal benefits, but we recognize the risks which need broader study and discussion. We’ve been pursuing these issues with technologists, civil society, academics and policymakers to expand our understanding of the risks, including in the contexts of bias, privacy and surveillance, and to develop our 6 principles to manage our development and use of facial recognition. We believe there are three problems that governments need to address: 1. Certain uses of facial recognition technology increase the risk of decisions and outcomes that are biased. 2. The widespread use of this technology can lead to new intrusions into privacy. 3. The use of facial recognition technology by a government or law enforcement for mass surveillance can harm democratic freedoms. All three of these problems should be addressed through legislation by requiring transparency, enabling third part testing, ensuring meaningful human review, avoiding the use for unlawful discrimination, ensuring notice, clarifying consent, and limiting on-going government surveillance of specified individuals.

Informed consent for survey and broader population data is critical to ethical applications of humanitarian intervention, from the use of household surveys to national population data. However, how much of the data we collect is collected by consent, and what does "informed" in the age in AI? While Artificial Intelligence may allow humanitarians to achieve truly meaningful achievements -- to predict and prescribe interventions before the famine or before the outbreak -- is the data behind those future successes ethical? Is data collected for one study but repurposed to train algorithms what those original subjects agreed to? If the platforms we store this data on today are upgraded to incorporate deep learning functions, is that original data fair game? This session aims to explore how data on individuals represents virtual bodies, subject to an interpretation of international law that forbids experimentation, and that even the examination of this data in aggregate does not fulfill our fundamental "do no harm" policies. We ask: what is missing to bring truly informed consent to AI?

Artificial intelligence systems are already playing a major role in how we live our lives on and off the internet. Since data collection has fueled the internet economy, the demand to apply logical systems to make sense of and profit off of this information has surged. This has made algorithmic systems a powerful tool in targeted advertising, content curation, automation, and scientific research. The use of AI however, has been fraught with problems, many times arising from the inherent bias and discrimination that are too easily embedded into the algorithms. There is a growing demand for fairness and accountability in these systems. However, much of these efforts would be constrained by emerging global rules on AI in trade agreements. Universally, they restrict regulators from protecting citizens from algorithmic bias and misuse. This panel will discuss why and how trade agreements contain AI governance rules that will define and constrain policy making at a domestic level. We will exchange ideas on the way forward to advocate for algorithmic transparency in future trade agreements and what CSOs can do.