Beyond a Human Rights-based approach to AI Governance: Promise, Pitfalls, Plea


Paper by Nathalie A. Smuha: “This paper discusses the establishment of a governance framework to secure the development and deployment of “good AI”, and describes the quest for a morally objective compass to steer it. Asserting that human rights can provide such compass, this paper first examines what a human rights-based approach to AI governance entails, and sets out the promise it propagates. Subsequently, it examines the pitfalls associated with human rights, particularly focusing on the criticism that these rights may be too Western, too individualistic, too narrow in scope and too abstract to form the basis of sound AI governance. After rebutting these reproaches, a plea is made to move beyond the calls for a human rights-based approach, and start taking the necessary steps to attain its realisation. It is argued that, without elucidating the applicability and enforceability of human rights in the context of AI; adopting legal rules that concretise those rights where appropriate; enhancing existing enforcement mechanisms; and securing an underlying societal infrastructure that enables human rights in the first place, any human rights-based governance framework for AI risks falling short of its purpose….(More)”.

The human rights impacts of migration control technologies


Petra Molnar at EDRI: “At the start of this new decade, over 70 million people have been forced to move due to conflict, instability, environmental factors, and economic reasons. As a response to the increased migration into the European Union, many states are looking into various technological experiments to strengthen border enforcement and manage migration. These experiments range from Big Data predictions about population movements in the Mediterranean to automated decision-making in immigration applications and Artificial Intelligence (AI) lie detectors at European borders. However, often these technological experiments do not consider the profound human rights ramifications and real impacts on human lives

A human laboratory of high risk experiments

Technologies of migration management operate in a global context. They reinforce institutions, cultures, policies and laws, and exacerbate the gap between the public and the private sector, where the power to design and deploy innovation comes at the expense of oversight and accountability. Technologies have the power to shape democracy and influence elections, through which they can reinforce the politics of exclusion. The development of technology also reinforces power asymmetries between countries and influence our thinking around which countries can push for innovation, while other spaces like conflict zones and refugee camps become sites of experimentation. The development of technology is not inherently democratic and issues of informed consent and right of refusal are particularly important to think about in humanitarian and forced migration contexts. For example, under the justification of efficiency, refugees in Jordan have their irises scanned in order to receive their weekly rations. Some refugees in the Azraq camp have reported feeling like they did not have the option to refuse to have their irises scanned, because if they did not participate, they would not get food. This is not free and informed consent….(More)”.

Algorithms and Contract Law


Paper by Lauren Henry Scholz: “Generalist confusion about the technology behind complex algorithms has led to inconsistent case law for algorithmic contracts. Case law explicitly grounded in the principle that algorithms are constructive agents for the companies they serve would provide a clear basis for enforceability of algorithmic contracts that is both principled from a technological perspective and is readily intelligible and able to be applied by generalists….(More)”.

Facial Recognition Software requires Checks and Balances


David Eaves,  and Naeha Rashid in Policy Options: “A few weeks ago, members of the Nexus traveller identification program were notified that Canadian Border Services is upgrading its automated system, from iris scanners to facial recognition technology. This is meant to simplify identification and increase efficiency without compromising security. But it also raises profound questions concerning how we discuss and develop public policies around such technology – questions that may not be receiving sufficiently open debate in the rush toward promised greater security.

Analogous to the U.S. Customs and Border Protection (CBP) program Global Entry, Nexus is a joint Canada-US border control system designed for low-risk, pre-approved travellers. Nexus does provide a public good, and there are valid reasons to improve surveillance at airports. Even before 9/11, border surveillance was an accepted annoyance and since then, checkpoint operations have become more vigilant and complex in response to the public demand for safety.

Nexus is one of the first North America government-sponsored services to adopt facial recognition, and as such it could be a pilot program that other services will follow. Left unchecked, the technology will likely become ubiquitous at North American border crossings within the next decade, and it will probably be adopted by governments to solve domestic policy challenges.

Facial recognition software is imperfect and has documented bias, but it will continue to improve and become superior to humans in identifying individuals. Given this, questions arise such as, what policies guide the use of this technology? What policies should inform future government use? In our headlong rush toward enhanced security, we risk replicating the justification the used by the private sector in an attempt to balance effectiveness, efficiency and privacy.

One key question involves citizens’ capacity to consent. Previously, Nexus members submitted to fingerprint and retinal scans – biometric markers that are relatively unique and enable government to verify identity at the border. Facial recognition technology uses visual data and seeks, analyzes, and stores identifying facial information in a database, which is then used to compare with new images and video….(More)”.

Accelerating AI with synthetic data


Essay by Khaled El Emam: “The application of artificial intelligence and machine learning to solve today’s problems requires access to large amounts of data. One of the key obstacles faced by analysts is access to this data (for example, these issues were reflected in reports from the General Accountability Office and the McKinsey Institute).

Synthetic data can help solve this data problem in a privacy preserving manner.

What is synthetic data ?

Data synthesis is an emerging privacy-enhancing technology that can enable access to realistic data, which is information that may be synthetic, but has the properties of an original dataset. It also simultaneously ensures that such information can be used and disclosed with reduced obligations under contemporary privacy statutes. Synthetic data retains the statistical properties of the original data. Therefore, there are an increasing number of use cases where it would serve as a proxy for real data.

Synthetic data is created by taking an original (real) dataset and then building a model to characterize the distributions and relationships in that data — this is called the “synthesizer.” The synthesizer is typically an artificial neural network or other machine learning technique that learns these (original) data characteristics. Once that model is created, it can be used to generate synthetic data. The data is generated from the model and does not have a 1:1 mapping to real data, meaning that the likelihood of mapping the synthetic records to real individuals would be very small — it is not considered personal information.

Many different types of data can be synthesized, including images, video, audio, text and structured data. The main focus in this article is on the synthesis of structured data.

Even though data can be generated in this manner, that does not mean it cannot be personal information. If the synthesizer is overfit to real data, then the generated data will replicate the original real data. Therefore, the synthesizer has to be constructed in a manner to avoid such overfitting. A formal privacy assurance should also be performed on the synthesized data to validate that there is a weak mapping between synthetic records to individuals….(More)”.

Government by Algorithm: Artificial Intelligence in Federal Administrative Agencies


The Administrative Conference of the United States: “Artificial intelligence (AI) promises to transform how government agencies do their work. Rapid developments in AI have the potential to reduce the cost of core governance functions, improve the quality of decisions, and unleash the power of administrative data, thereby making government performance more efficient and effective. Agencies that use AI to realize these gains will also confront important questions about the proper design of algorithms and user interfaces, the respective scope of human and machine decision-making, the boundaries between public actions and private contracting, their own capacity to learn over time using AI, and whether the use of AI is even permitted.

These are important issues for public debate and academic inquiry. Yet little is known about how agencies are currently using AI systems beyond a few headlinegrabbing examples or surface-level descriptions. Moreover, even amidst growing public and  scholarly discussion about how society might regulate government use of AI, little attention has been devoted to how agencies acquire such tools in the first place or oversee their use. In an effort to fill these gaps, the Administrative Conference of the United States (ACUS) commissioned this report from researchers at Stanford University and New York University. The research team included a diverse set of lawyers, law students, computer scientists, and social scientists with the capacity to analyze these cutting-edge issues from technical, legal, and policy angles. The resulting report offers three cuts at federal agency use of AI:

  • a rigorous canvass of AI use at the 142 most significant federal departments, agencies, and sub-agencies (Part I)
  • a series of in-depth but accessible case studies of specific AI applications at seven leading agencies covering a range of governance tasks (Part II); and
  • a set of cross-cutting analyses of the institutional, legal, and policy challenges raised by agency use of AI (Part III)….(More)”

This emoji could mean your suicide risk is high, according to AI


Rebecca Ruiz at Mashable: “Since its founding in 2013, the free mental health support service Crisis Text Line has focused on using data and technology to better aid those who reach out for help. 

Unlike helplines that offer assistance based on the order in which users dialed, texted, or messaged, Crisis Text Line has an algorithm that determines who is in most urgent need of counseling. The nonprofit is particularly interested in learning which emoji and words texters use when their suicide risk is high, so as to quickly connect them with a counselor. Crisis Text Line just released new insights about those patterns. 

Based on its analysis of 129 million messages processed between 2013 and the end of 2019, the nonprofit found that the pill emoji, or 💊, was 4.4 times more likely to end in a life-threatening situation than the word suicide. 

Other words that indicate imminent danger include 800mg, acetaminophen, excedrin, and antifreeze; those are two to three times more likely than the word suicide to involve an active rescue of the texter. The loudly crying emoji face, or 😭, is similarly high-risk. In general, the words that trigger the greatest alarm suggest the texter has a method or plan to attempt suicide or may be in the process of taking their own life. …(More)”.

Our personal health history is too valuable to be harvested by the tech giants


Eerke Boiten at The Guardian: “…It is clear that the black box society does not only feed on internet surveillance information. Databases collected by public bodies are becoming more and more part of the dark data economy. Last month, it emerged that a data broker in receipt of the UK’s national pupil database had shared its access with gambling companies. This is likely to be the tip of the iceberg; even where initial recipients of shared data might be checked and vetted, it is much harder to oversee who the data is passed on to from there.

Health data, the rich population-wide information held within the NHS, is another such example. Pharmaceutical companies and internet giants have been eyeing the NHS’s extensive databases for commercial exploitation for many years. Google infamously claimed it could save 100,000 lives if only it had free rein with all our health data. If there really is such value hidden in NHS data, do we really want Google to extract it to sell it to us? Google still holds health data that its subsidiary DeepMind Health obtained illegally from the NHS in 2016.

Although many health data-sharing schemes, such as in the NHS’s register of approved data releases], are said to be “anonymised”, this offers a limited guarantee against abuse.

There is just too much information included in health data that points to other aspects of patients’ lives and existence. If recipients of anonymised health data want to use it to re-identify individuals, they will often be able to do so by combining it, for example, with publicly available information. That this would be illegal under UK data protection law is a small consolation as it would be extremely hard to detect.

It is clear that providing access to public organisations’ data for research purposes can serve the greater good and it is unrealistic to expect bodies such as the NHS to keep this all in-house.

However, there are other methods by which to do this, beyond the sharing of anonymised databases. CeLSIUS, for example, a physical facility where researchers can interrogate data under tightly controlled conditions for specific registered purposes, holds UK census information over many years.

These arrangements prevent abuse, such as through deanonymisation, do not have the problem of shared data being passed on to third parties and ensure complete transparency of the use of the data. Online analogues of such set-ups do not yet exist, but that is where the future of safe and transparent access to sensitive data lies….(More)”.

How to Put the Data Subject's Sovereignty into Practice. Ethical Considerations and Governance Perspectives



Paper by Peter Dabrock: “Ethical considerations and governance approaches of AI are at a crossroads. Either one tries to convey the impression that one can bring back a status quo ante of our given “onlife”-era, or one accepts to get responsibly involved in a digital world in which informational self-determination can no longer be safeguarded and fostered through the old fashioned data protection principles of informed consent, purpose limitation and data economy. The main focus of the talk is on how under the given conditions of AI and machine learning, data sovereignty (interpreted as controllability [not control (!)] of the data subject over the use of her data throughout the entire data processing cycle) can be strengthened without hindering innovation dynamics of digital economy and social cohesion of fully digitized societies. In order to put this approach into practice the talk combines a presentation of the concept of data sovereignty put forward by the German Ethics Council with recent research trends in effectively applying the AI ethics principles of explainability and enforceability…(More)”.

Realizing the Potential of AI Localism


Stefaan G. Verhulst and Mona Sloane at Project Syndicate: “Every new technology rides a wave from hype to dismay. But even by the usual standards, artificial intelligence has had a turbulent run. Is AI a society-renewing hero or a jobs-destroying villain? As always, the truth is not so categorical.

As a general-purpose technology, AI will be what we make of it, with its ultimate impact determined by the governance frameworks we build. As calls for new AI policies grow louder, there is an opportunity to shape the legal and regulatory infrastructure in ways that maximize AI’s benefits and limit its potential harms.

Until recently, AI governance has been discussed primarily at the national level. But most national AI strategies – particularly China’s – are focused on gaining or maintaining a competitive advantage globally. They are essentially business plans designed to attract investment and boost corporate competitiveness, usually with an added emphasis on enhancing national security.

This singular focus on competition has meant that framing rules and regulations for AI has been ignored. But cities are increasingly stepping into the void, with New York, Toronto, Dubai, Yokohama, and others serving as “laboratories” for governance innovation. Cities are experimenting with a range of policies, from bans on facial-recognition technology and certain other AI applications to the creation of data collaboratives. They are also making major investments in responsible AI research, localized high-potential tech ecosystems, and citizen-led initiatives.

This “AI localism” is in keeping with the broader trend in “New Localism,” as described by public-policy scholars Bruce Katz and the late Jeremy Nowak. Municipal and other local jurisdictions are increasingly taking it upon themselves to address a broad range of environmental, economic, and social challenges, and the domain of technology is no exception.

For example, New York, Seattle, and other cities have embraced what Ira Rubinstein of New York University calls “privacy localism,” by filling significant gaps in federal and state legislation, particularly when it comes to surveillance. Similarly, in the absence of a national or global broadband strategy, many cities have pursued “broadband localism,” by taking steps to bridge the service gap left by private-sector operators.

As a general approach to problem solving, localism offers both immediacy and proximity. Because it is managed within tightly defined geographic regions, it affords policymakers a better understanding of the tradeoffs involved. By calibrating algorithms and AI policies for local conditions, policymakers have a better chance of creating positive feedback loops that will result in greater effectiveness and accountability….(More)”.