On March 18, 2018, at around 10 p.m., Elaine Herzberg was wheeling her bicycle across a street in Tempe, Arizona, when she was struck and killed by a self-driving car. Although there was a human operator behind the wheel, an autonomous system--artificial intelligence--was in full control. This incident, like others involving interactions between people and AI technologies, raises a host of ethical and proto-legal questions. What moral obligations did the system's programmers have to prevent their creation from taking a human life? And who was responsible for Herzberg's death? "Artificial intelligence" refers to systems that can be designed to take cues from their environment and, based on those inputs, proceed to solve problems, assess risks, make predictions, and take actions. In the era predating powerful computers and big data, such systems were programmed by humans and followed rules of human invention, but advances in technology have led to the development of new approaches.

Today, artificial intelligence and machine learning tools are ubiquitous across sectors--used for everything from determining an individual's credit worthiness to enabling law enforcement surveillance--and rapidly evolving. Despite this, few nations have rules in place to oversee these systems or mitigate the harms they could cause. On April 21, the European Commission released a draft of its proposed AI regulation, the world's first legal framework addressing the risks posed by artificial intelligence. The draft regulation makes some notable strides, prohibiting the use of certain harmful AI systems and reining in harmful uses of some high-risk algorithmic systems. However, the Commission's proposed regulation displays gaps which, if not addressed, could limit its effectiveness in holding some of the biggest developers and deployers of algorithmic systems accountable.

Over the past three years, celebrities have been appearing across social media in improbable scenarios. You may have recently caught a grinning Tom Cruise doing magic tricks with a coin or Nicolas Cage appearing as Lois Lane in Man of Steel. Most of us now recognize these clips as deepfakes--startlingly realistic videos created using artificial intelligence. In 2017, they began circulating on message boards like Reddit as altered videos from anonymous users; the term is a portmanteau of "deep learning"--the process used to train an algorithm to doctor a scene--and "fake." Deepfakes once required working knowledge of AI-enabled technology, but today, anyone can make their own using free software like FakeApp or Faceswap. All it takes is some sample footage and a large data set of photos (one reason celebrities are targeted is the easy availability of high-quality facial images) and the app can convincingly swap out one person's face for another's.

In recent years, the problem of addressing fairness in Machine Learning (ML) and automatic decision-making has attracted a lot of attention in the scientific communities dealing with Artificial Intelligence. A plethora of different definitions of fairness in ML have been proposed, that consider different notions of what is a "fair decision" in situations impacting individuals in the population. The precise differences, implications and "orthogonality" between these notions have not yet been fully analyzed in the literature. In this work, we try to make some order out of this zoo of definitions.

Artificial Intelligence (AI) has been compared to electricity: it is a general-purpose technology with applications in all domains of human activity. Electricity has found uses that no one envisaged when the first electrical systems were designed and, in practice, life would be completely different without this technology. Ideally, the Act would have developed the two central ideas addressed by the White Paper: creating legislation that stimulates innovation, while at the same time guaranteeing trust. However, in its current form, the document has a few drawbacks and needs to mature to meet the expectations of the AI community, in particular, and of society, in general. The main sections of the Act are concerned with prohibited practices, high-risk systems, transparency requirements, and governance.

In a study published today, OpenAI, the lab best known for its research on large language models, claims it's discovered a way to improve the "behavior" of language models with respect to ethical, moral, and societal values. The approach, OpenAI says, can give developers the tools to dictate the tone and personality of a model depending on the prompt that the model's given. Despite the potential of natural language models like GPT-3, many blockers exist. The models can't always answer math problems correctly or respond to questions without paraphrasing training data, and it's well-established that they amplify the biases in data on which they were trained. That's problematic in the language domain, because a portion of the data is often sourced from communities with pervasive gender, race, and religious prejudices.

For attorneys, "discovery" is the critical, information gathering phase of a case. A time that's spent collecting and reviewing evidence that will eventually become the building blocks for future arguments. In times of yore, this process was primarily physical--it involved gathering actual documents, pouring over paper and ink photos, interviewing real live witnesses, and combing through objects archived in evidence storage lockers. These days, however, building a case isn't so much about reviewing what you can see, taste, feel, and smell, as it is about what you can't. It's about combing through huge amounts of unstructured material that only exist as 0s and 1s inside a database somewhere.

Some would contend that there is the law and then there is everything else. You've likely heard that well-worn line before. It is certainly eye-catching and memorable. What makes the catchphrase especially interesting is that somewhere in that morass is so-called soft law, sitting somewhat in a no man's zone. Yes, there is a nebulous grey area that is not quite a law and yet oftentimes provides a law-like shaping and tonal directive toward what we can do, including whether our actions are seemingly lawful or ostensibly could veer into becoming unlawful.

Identifying academic plagiarism is a pressing problem, among others, for research institutions, publishers, and funding organizations. Detection approaches proposed so far analyze lexical, syntactical, and semantic text similarity. These approaches find copied, moderately reworded, and literally translated text. However, reliably detecting disguised plagiarism, such as strong paraphrases, sense-for-sense translations, and the reuse of non-textual content and ideas, is an open research problem. The thesis addresses this problem by proposing plagiarism detection approaches that implement a different concept: analyzing non-textual content in academic documents, specifically citations, images, and mathematical content. To validate the effectiveness of the proposed detection approaches, the thesis presents five evaluations that use real cases of academic plagiarism and exploratory searches for unknown cases. The evaluation results show that non-textual content elements contain a high degree of semantic information, are language-independent, and largely immutable to the alterations that authors typically perform to conceal plagiarism. Analyzing non-textual content complements text-based detection approaches and increases the detection effectiveness, particularly for disguised forms of academic plagiarism. To demonstrate the benefit of combining non-textual and text-based detection methods, the thesis describes the first plagiarism detection system that integrates the analysis of citation-based, image-based, math-based, and text-based document similarity. The system's user interface employs visualizations that significantly reduce the effort and time users must invest in examining content similarity.

As AI systems are integrated into high stakes social domains, researchers now examine how to design and operate them in a safe and ethical manner. However, the criteria for identifying and diagnosing safety risks in complex social contexts remain unclear and contested. In this paper, we examine the vagueness in debates about the safety and ethical behavior of AI systems. We show how this vagueness cannot be resolved through mathematical formalism alone, instead requiring deliberation about the politics of development as well as the context of deployment. Drawing from a new sociotechnical lexicon, we redefine vagueness in terms of distinct design challenges at key stages in AI system development. The resulting framework of Hard Choices in Artificial Intelligence (HCAI) empowers developers by 1) identifying points of overlap between design decisions and major sociotechnical challenges; 2) motivating the creation of stakeholder feedback channels so that safety issues can be exhaustively addressed. As such, HCAI contributes to a timely debate about the status of AI development in democratic societies, arguing that deliberation should be the goal of AI Safety, not just the procedure by which it is ensured.