AI Ethics: Seven Traps


Blog Post by Annette Zimmermann and Bendert Zevenbergen: “… In what follows, we outline seven ‘AI ethics traps’. In doing so, we hope to provide a resource for readers who want to understand and navigate the public debate on the ethics of AI better, who want to contribute to ongoing discussions in an informed and nuanced way, and who want to think critically and constructively about ethical considerations in science and technology more broadly. Of course, not everybody who contributes to the current debate on AI Ethics is guilty of endorsing any or all of these traps: the traps articulate extreme versions of a range of possible misconceptions, formulated in a deliberately strong way to highlight the ways in which one might prematurely dismiss ethical reasoning about AI as futile.

1. The reductionism trap:

“Doing the morally right thing is essentially the same as acting in a fair way. (or: transparent, or egalitarian, or <substitute any other value>). So ethics is the same as fairness (or transparency, or equality, etc.). If we’re being fair, then we’re being ethical.”

            Even though the problem of algorithmic bias and its unfair impact on decision outcomes is an urgent problem, it does not exhaust the ethical problem space. As important as algorithmic fairness is, it is crucial to avoid reducing ethics to a fairness problem alone. Instead, it is important to pay attention to how the ethically valuable goal of optimizing for a specific value like fairness interacts with other important ethical goals. Such goals could include—amongst many others—the goal of creating transparent and explainable systems which are open to democratic oversight and contestation, the goal of improving the predictive accuracy of machine learning systems, the goal of avoiding paternalistic infringements of autonomy rights, or the goal of protecting the privacy interests of data subjects. Sometimes, these different values may conflict: we cannot always optimize for everything at once. This makes it all the more important to adopt a sufficiently rich, pluralistic view of the full range of relevant ethical values at stake—only then can one reflect critically on what kinds of ethical trade-offs one may have to confront.

2. The simplicity trap:

“In order to make ethics practical and action-guiding, we need to distill our moral framework into a user-friendly compliance checklist. After we’ve decided on a particular path of action, we’ll go through that checklist to make sure that we’re being ethical.”

            Given the high visibility and urgency of ethical dilemmas arising in the context of AI, it is not surprising that there are more and more calls to develop actionable AI ethics checklists. For instance, a 2018 draft report by the European Commission’s High-Level Expert Group on Artificial Intelligence specifies a preliminary ‘assessment list’ for ‘trustworthy AI’. While the report plausibly acknowledges that such an assessment list must be context-sensitive and that it is not exhaustive, it nevertheless identifies a list of ten fixed ethical goals, including privacy and transparency. But can and should ethical values be articulated in a checklist in the first place? It is worth examining this underlying assumption critically. After all, a checklist implies a one-off review process: on that view, developers or policy-makers could determine whether a particular system is ethically defensible at a specific moment in time, and then move on without confronting any further ethical concerns once the checklist criteria have been satisfied once. But ethical reasoning cannot be a static one-off assessment: it required an ongoing process of reflection, deliberation, and contestation. Simplicity is good—but the willingness to reconsider simple frameworks, when required, is better. Setting a fixed ethical agenda ahead of time risks obscuring new ethical problems that may arise at a later point in time, or ongoing ethical problems that become apparent to human decision-makers only later.

3. The relativism trap:

“We all disagree about what is morally valuable, so it’s pointless to imagine that there is a universalbaseline against which we can use in order to evaluate moral choices. Nothing is objectively morally good: things can only be morally good relative to each person’s individual value framework.”

            Public discourse on the ethics of AI frequently produces little more than an exchange of personal opinions or institutional positions. In light of pervasive moral disagreement, it is easy to conclude that ethical reasoning can never stand on firm ground: it always seems to be relative to a person’s views and context. But this does not mean that ethical reasoning about AI and its social and political implications is futile: some ethical arguments about AI may ultimately be more persuasive than others. While it may not always be possible to determine ‘the one right answer’, it is often possible to identify at least  some paths of action are clearly wrong, and some paths of action that are comparatively better (if not optimal all things considered). If that is the case, comparing the respective merits of ethical arguments can be action-guiding for developers and policy-makers, despite the presence of moral disagreement. Thus, it is possible and indeed constructive for AI ethics to welcome value pluralism, without collapsing into extreme value relativism.

4. The value alignment trap:

“If relativism is wrong (see #3), there must be one morally right answer. We need to find that right answer, and ensure that everyone in our organisation acts in alignment with that answer. If our ethical reasoning leads to moral disagreement, that means that we have failed.”…(More)”.

Seeing, Naming, Knowing


Essay by Nora N. Khan for Brooklyn Rail: “…. Throughout this essay, I use “machine eye” as a metaphor for the unmoored orb, a kind of truly omnidirectional camera (meaning, a camera that can look in every direction and vector that defines the dimensions of a sphere), and as a symbolic shorthand for the sum of four distinct realms in which automated vision is deployed as a service. (Vision as a Service, reads the selling tag for a new AI surveillance camera company).10 Those four general realms are: 

1. Massive AI systems fueled by the public’s flexible datasets of their personal images, creating a visual culture entirely out of digitized images. 

2. Facial recognition technologies and neural networks improving atop their databases. 

3. The advancement of predictive policing to sort people by types. 

4. The combination of location-based tracking, license plate-reading, and heat sensors to render skein-like, live, evolving maps of people moving, marked as likely to do X.

Though we live the results of its seeing, and its interpretation of its seeing, for now I would hold on blaming ourselves for this situation. We are, after all, the living instantiations of a few thousand years of such violent seeing globally, enacted through imperialism, colonialism, caste stratification, nationalist purges, internal class struggle, and all the evolving theory to support and galvanize the above. Technology simply recasts, concentrates, and amplifies these “tendencies.” They can be hard to see at first because the eye’s seeing seems innocuous, and is designed to seem so. It is a direct expression of the ideology of software, which reflects its makers’ desires. These makers are lauded as American pioneers, innovators, genius-heroes living in the Bay Area in the late 1970s, vibrating at a highly specific frequency, the generative nexus of failed communalism and an emerging Californian Ideology. That seductive ideology has been exported all over the world, and we are only now contending with its impact.

Because the workings of machine visual culture are so remote from our sense perception, and because it so acutely determines our material (economic, social), and affective futures, I invite you to see underneath the eye’s outer glass shell, its holder, beyond it, to the grid that organizes its “mind.” That mind simulates a strain of ideology about who exactly gets to gather data about those on that grid below, and how that data should be mobilized to predict the movements and desires of the grid dwellers. This mind, a vast computational regime we are embedded in, drives the machine eye. And this computational regime has specific values that determine what is seen, how it is seen, and what that seeing means….(More)”.

The Bad Pupil


CCCBLab: “In recent years we have been witnessing a constant trickle of news on artificial intelligence, machine learning and computer vision. We are told that machines learn, see, create… and all this builds up a discourse based on novelty, on a possible future and on a series of worries and hopes. It is difficult, sometimes, to figure out in this landscape which are real developments, and which are fantasies or warnings. And, undoubtedly, this fog that surrounds it forms part of the power that we grant, both in the present and on credit, to these tools, and of the negative and positive concerns that they arouse in us. Many of these discourses may fall into the field of false debates or, at least, of the return of old debates. Thus, in the classical artistic field, associated with the discourse on creation and authorship, there is discussion regarding the entity to be awarded to the images created with these tools. (Yet wasn’t the argument against photography in art that it was an image created automatically and without human participation? And wasn’t that also an argument in favour of taking it and using it to put an end to a certain idea of art?)

Metaphors are essential in the discourse on all digital tools and the power that they have. Are expressions such as “intelligence”, “vision”, “learning”, “neural” and the entire range of similar words the most adequate for defining these types of tools? Probably not, above all if their metaphorical nature is sidestepped. We would not understand them in the same way if we called them tools of probabilistic classification or if instead of saying that an artificial intelligence “has painted” a Rembrandt, we said that it has produced a statistical reproduction of his style (something which is still surprising, and to be celebrated, of course). These names construct an entity for these tools that endows them with a supposed autonomy and independence upon which their future authority is based.

Because that is what it’s about in many discourses: constructing a characterisation that legitimises an objective or non-human capacity in data analysis….

We now find ourselves in what is, probably, the point of the first cultural reception of these tools. Of their development in fields of research and applications that have already been derived, we are moving on to their presence in the public discourse. It is in this situation and context, where we do not fully know the breadth and characteristics of these technologies (meaning fears are more abstract and diffuse and, thus, more present and powerful), when it is especially important to understand what we are talking about, to appropriate the tools and to intervene in the discourses. Before their possibilities are restricted and solidified until they seem indisputable, it is necessary to experiment with them and reflect on them; taking advantage of the fact that we can still easily perceive them as in creation, malleable and open.

In our projects The Bad Pupil. Critical pedagogy for artificial intelligences and Latent Spaces. Machinic Imaginations we have tried to approach to these tools and their imaginary. In the statement of intentions of the former, we expressed our desire, in the face of the regulatory context and the metaphor of machine learning, to defend the bad pupil as one who escapes the norm. And also how, faced with an artificial intelligence that seeks to replicate the human on inhuman scales, it is necessary to defend and construct a non-mimetic one that produces unexpected relations and images.

Fragment of De zeven werken van barmhartigheid, Meester van Alkmaar, 1504 (Rijksmuseum, Amsterdam) analysed with YOLO9000 | The Bad Pupil - Estampa

Fragment of De zeven werken van barmhartigheid, Meester van Alkmaar, 1504 (Rijksmuseum, Amsterdam) analysed with YOLO9000 | The Bad Pupil – Estampa

Both projects are also attempts to appropriate these tools, which means, first of all, escaping industrial barriers and their standards. In this field in which mass data are an asset within reach of big companies, employing quantitively poor datasets and non-industrial calculation potentials is not just a need but a demand….(More)”.

How AI Can Cure the Big Idea Famine


Saahil Jayraj Dama at JoDS: “Today too many people are still deprived of basic amenities such as medicine, while current patent laws continue to convolute and impede innovation. But if allowed, AI can provide an opportunity to redefine this paradigm and be the catalyst for change—if….

Which brings us to the most befitting answer: No one owns the intellectual property rights to AI-generated creations, and these creations fall into the public domain. This may seem unpalatable at first, especially since intellectual property laws have played such a fundamental role in our society so far. We have been conditioned to a point where it seems almost unimaginable that some creations should directly enter the public domain upon their birth.

But, doctrinally, this is the only proposition that stays consistent to extant intellectual property laws. Works created by AI have no rightful owner because the application of mind to generate the creation, along with the actual generation of the creation, would entirely be done by the AI system. Human involvement is ancillary and is limited to creating an environment within which such a creation can take form.

This can be better understood through a hypothetical example: If an AI system were to invent a groundbreaking pharmaceutical ingredient which completely treats balding, then the system would likely begin by understanding the problem and state of prior art. It would undertake research on causes of balding, existing cures, problems with existing cures, and whether its proposed cure would have any harmful side effects. It would also possibly combine research and knowledge across various domains, which could range from Ayurveda to modern-day biochemistry, before developing its invention.

The developer can lay as much stake to this invention as the team behind AlphaGo for beating Lee Sedol at Go. The user is even further detached from the exercise of ingenuity: She would be the person who first thought, “We should build a Go playing AI system,” and direct the AI system to learn Go by watching certain videos and playing against itself. Despite the intervention of all these entities, the fact remains that the victory only belongs to AlphaGo itself.

Doctrinal issues aside, this solution ties in with what people need from intellectual property laws: more openness and accessibility. The demands for improved access to medicines and knowledge, fights against cultural monopolies, and brazen violations of unjust intellectual property laws are all symptomatic of the growing public discontent against strong intellectual property laws. Through AI, we can design legal systems which address these concerns and reform the heavy handed approach that has been adopted toward intellectual property rights so far.

Tying the Threads Together

For the above to materialize, governments and legislators need to accept that our present intellectual property system is broken and inconsistent with what people want. Too many people are being deprived of basic amenities such as medicines, patent trolls and patent thickets are slowing innovation, educational material is still outside the reach of most people, and culture is not spreading as widely as it should. AI can provide an opportunity for us to redefine this paradigm—it can lead to a society that draws and benefits from an enriched public domain.

However, this approach does come with built-in cynicism because it contemplates an almost complete overhaul of the system. One could argue that if open access for AI-generated creations does become the norm, then innovation and creativity would suffer as people would no longer have the incentive to create. People may even refuse to use their AI systems, and instead stick to producing inventions and creative works by themselves. This would be detrimental to scientific and cultural progress and would also slow adoption of AI systems in society.

Yet, judging by the pace at which these systems have progressed so far and what they can currently do, it is easy to imagine a reality where humans developing inventions and producing creative works almost becomes an afterthought. If a machine can access all the world’s publicly available knowledge and information to develop an invention, or study a user’s likes and dislikes while producing a new musical composition, it is easy to see how humans would, eventually, be pushed out of the loop. AI-generated creations are, thus, inevitable.

The incentive theory will have to be reimagined, too. Constant innovation coupled with market forces will change the system from “incentive-to-create” to “incentive-to-create-well.” While every book, movie, song, and invention is treated at par under the law, only the best inventions and creative works will thrive under the new model. If a particular developer’s AI system can write incredible dialogue for a comedy film or invent the most efficient car engines, the market would want more of these AI systems. Thus incentive will not be eliminated, it will just take a different form.

It is true that writing about such grand schemes is significantly tougher than practically implementing them. But, for any idea to succeed, it must start with a discussion such as this one. Admittedly, we are still a moonshot away from any country granting formal recognition to open access as the basis of its intellectual property laws. And even if a country were to do this, it faces a plethora of hoops to jump through, such as conducting feasibility-testing and dealing with international and internal pressure. Despite these issues, facilitating better access through AI systems remains an objective worth achieving for any society that takes pride in being democratic and equal….(More)”.

What Would More Democratic A.I. Look Like?


Blog post by Andrew Burgess: “Something curious is happening in Finland. Though much of the global debate around artificial intelligence (A.I.) has become concerned with unaccountable, proprietary systems that could control our lives, the Finnish government has instead decided to embrace the opportunity by rolling out a nationwide educational campaign.

Conceived in 2017, shortly after Finland’s A.I. strategy was announced, the government wants to rebuild the country’s economy around the high-end opportunities of artificial intelligence, and has launched a national programto train 1 percent of the population — that’s 55,000 people — in the basics of A.I. “We’ll never have so much money that we will be the leader of artificial intelligence,” said economic minister Mika Lintilä at the launch. “But how we use it — that’s something different.”

Artificial intelligence can have many positive applications, from being trained to identify cancerous cells in biopsy screenings, predict weather patterns that can help farmers increase their crop yields, and improve traffic efficiency.

But some believe that A.I. expertise is currently too concentrated in the hands of just a few companies with opaque business models, meaning resources are being diverted away from projects that could be more socially, rather than commercially, beneficial. Finland’s approach of making A.I. accessible and understandable to its citizens is part of a broader movement of people who want to democratize the technology, putting utility and opportunity ahead of profit.

This shift toward “democratic A.I.” has three main principles: that all society will be impacted by A.I. and therefore its creators have a responsibility to build open, fair, and explainable A.I. services; that A.I. should be used for social benefit and not just for private profit; and that because A.I. learns from vast quantities of data, the citizens who create that data — about their shopping habits, health records, or transport needs — have a right to say and understand how it is used.

A growing movement across industry and academia believes that A.I. needs to be treated like any other “public awareness” program — just like the scheme rolled out in Finland….(More)”.

Data Trusts as an AI Governance Mechanism


Paper by Chris Reed and Irene YH Ng: “This paper is a response to the Singapore Personal Data Protection Commission consultation on a draft AI Governance Framework. It analyses the five data trust models proposed by the UK Open Data Institute and identifies that only the contractual and corporate models are likely to be legally suitable for achieving the aims of a data trust.

The paper further explains how data trusts might be used as in the governance of AI, and investigates the barriers which Singapore’s data protection law presents to the use of data trusts and how those barriers might be overcome. Its conclusion is that a mixed contractual/corporate model, with an element of regulatory oversight and audit to ensure consumer confidence that data is being used appropriately, could produce a useful AI governance tool…(More)”.

Deep Medicine: How Artificial Intelligence Can Make Healthcare Human Again


Book by Eric Topol: “Medicine has become inhuman, to disastrous effect. The doctor-patient relationship–the heart of medicine–is broken: doctors are too distracted and overwhelmed to truly connect with their patients, and medical errors and misdiagnoses abound. In Deep Medicine, leading physician Eric Topol reveals how artificial intelligence can help. AI has the potential to transform everything doctors do, from notetaking and medical scans to diagnosis and treatment, greatly cutting down the cost of medicine and reducing human mortality. By freeing physicians from the tasks that interfere with human connection, AI will create space for the real healing that takes place between a doctor who can listen and a patient who needs to be heard. Innovative, provocative, and hopeful, Deep Medicine shows us how the awesome power of AI can make medicine better, for all the humans involved….(More)”.

Machine Ethics: The Design and Governance of Ethical AI and Autonomous Systems


Introduction by A.F. Winfield, K. Michael, J. Pitt, V. Evers of Special Issue of Proceedings of the IEEE: “…The primary focus of this special issue is machine ethics, that is the question of how autonomous systems can be imbued with ethical values. Ethical autonomous systems are needed because, inevitably, near future systems are moral agents; consider driverless cars, or medical diagnosis AIs, both of which will need to make choices with ethical consequences. This special issue includes papers that describe both implicit ethical agents, that is machines designed to avoid unethical outcomes, and explicit ethical agents: machines which either encode or learn ethics and determine actions based on those ethics. Of course ethical machines are socio-technical systems thus, as a secondary focus, this issue includes papers that explore the societal and regulatory implications of machine ethics, including the question of ethical governance. Ethical governance is needed in order to develop standards and processes that allow us to transparently and robustly assure the safety of ethical autonomous systems and hence build public trust and confidence….(More)?

Regulating disinformation with artificial intelligence


Paper for the European Parliamentary Research Service: “This study examines the consequences of the increasingly prevalent use of artificial intelligence (AI) disinformation initiatives upon freedom of expression, pluralism and the functioning of a democratic polity. The study examines the trade-offs in using automated technology to limit the spread of disinformation online. It presents options (from self-regulatory to legislative) to regulate automated content recognition (ACR) technologies in this context. Special attention is paid to the opportunities for the European Union as a whole to take the lead in setting the framework for designing these technologies in a way that enhances accountability and transparency and respects free speech. The present project reviews some of the key academic and policy ideas on technology and disinformation and highlights their relevance to European policy.

Chapter 1 introduces the background to the study and presents the definitions used. Chapter 2 scopes the policy boundaries of disinformation from economic, societal and technological perspectives, focusing on the media context, behavioural economics and technological regulation. Chapter 3 maps and evaluates existing regulatory and technological responses to disinformation. In Chapter 4, policy options are presented, paying particular attention to interactions between technological solutions, freedom of expression and media pluralism….(More)”.

Is Ethical A.I. Even Possible?


Cade Metz at The New York Times: ” When a news article revealed that Clarifaiwas working with the Pentagon and some employees questioned the ethics of building artificial intelligence that analyzed video captured by drones, the company said the project would save the lives of civilians and soldiers.

“Clarifai’s mission is to accelerate the progress of humanity with continually improving A.I.,” read a blog post from Matt Zeiler, the company’s founder and chief executive, and a prominent A.I. researcher. Later, in a news media interview, Mr. Zeiler announced a new management position that would ensure all company projects were ethically sound.

As activists, researchers, and journalists voice concerns over the rise of artificial intelligence, warning against biased, deceptive and malicious applications, the companies building this technology are responding. From tech giants like Google and Microsoft to scrappy A.I. start-ups, many are creating corporate principles meant to ensure their systems are designed and deployed in an ethical way. Some set up ethics officers or review boards to oversee these principles.

But tensions continue to rise as some question whether these promises will ultimately be kept. Companies can change course. Idealism can bow to financial pressure. Some activists — and even some companies — are beginning to argue that the only way to ensure ethical practices is through government regulation....

As companies and governments deploy these A.I. technologies, researchers are also realizing that some systems are woefully biased. Facial recognition services, for instance, can be significantly less accurate when trying to identify women or someone with darker skin. Other systems may include security holes unlike any seen in the past. Researchers have shown that driverless cars can be fooled into seeing things that are not really there.

All this means that building ethical artificial intelligence is an enormously complex task. It gets even harder when stakeholders realize that ethics are in the eye of the beholder.

As some Microsoft employees protest the company’s military contracts, Mr. Smith said that American tech companies had long supported the military and that they must continue to do so. “The U.S. military is charged with protecting the freedoms of this country,” he told the conference. “We have to stand by the people who are risking their lives.”

Though some Clarifai employees draw an ethical line at autonomous weapons, others do not. Mr. Zeiler argued that autonomous weapons will ultimately save lives because they would be more accurate than weapons controlled by human operators. “A.I. is an essential tool in helping weapons become more accurate, reducing collateral damage, minimizing civilian casualties and friendly fire incidents,” he said in a statement.

Google worked on the same Pentagon project as Clarifai, and after a protest from company employees, the tech giant ultimately ended its involvement. But like Clarifai, as many as 20 other companies have worked on the project without bowing to ethical concerns.

After the controversy over its Pentagon work, Google laid down a set of “A.I. principles” meant as a guide for future projects. But even with the corporate rules in place, some employees left the company in protest. The new principles are open to interpretation. And they are overseen by executives who must also protect the company’s financial interests….

In their open letter, the Clarifai employees said they were unsure whether regulation was the answer to the many ethical questions swirling around A.I. technology, arguing that the immediate responsibility rested with the company itself….(More)”.