This chapter explores key elements of AI as relevant to intellectual property law. Understanding how artificial intelligence works is crucial for applying legal regimes to it. Legal practitioners, especially IP lawyers, need a deep understanding of AI’s technical nuances. Intellectual property doctrines aim to achieve practical ends, and their application to AI is highly fact-dependent. Patent law, for example, requires technical expertise in addition to legal knowledge. This chapter tracks the development of AI from simple programming to highly sophisticated learning algorithms. It emphasizes how AI is rapidly evolving and that many of these systems are already being widely adopted in society. AI is transforming fields like education, law, healthcare, and finance. While AI offers numerous benefits, it also raises concerns about bias and transparency, among numerous other ethical implications.

This chapter explores the single most important difference between Anglo-American and German/Continental trial procedures: bifurcation vs. unification. Should a court determine sentence at the same time as it adjudicates verdict? Or should the criminal process be divided, with sentencing taking place after conviction, in a separate ‘penalty phase’ of the criminal process? Common law (adversarial) jurisdictions take the bifurcated approach, while in civil law (inquisitorial) systems the sentencing decision is part and parcel of the decision to convict or acquit. The chapter investigates the merits of both approaches.

Comparing the two approaches to sentencing may yield important insights. Although neither system is likely to abandon its chosen methodology in favour of the alternative, there may be elements of each which can be adopted with a view to overcoming any structural deficiencies.

This chapter explains how to estimate population parameters from data. We introduce random sampling, an approach that yields accurate estimates from limited data. We then define the bias and the standard error, which quantify the average error of an estimator and how much it varies, respectively. In addition, we derive deviation bounds and use them to prove the law of large numbers, which states that averaging many independent samples from a distribution yields an accurate estimate of its mean. An important consequence is that random sampling provides a precise estimate of means and proportions. However, we caution that this is not necessarily the case, if the data contain extreme values. Next, we discuss the central limit theorem (CLT), according to which averages of independent quantities tend to be Gaussian. We again provide a cautionary tale, warning that this does not hold in the absence of independence. Then, we explain how to use the CLT to build confidence intervals which quantify the uncertainty of estimates obtained from finite data. Finally, we introduce the bootstrap, a popular computational technique to estimate standard errors and build confidence intervals.

Experimental jurisprudence draws methods and theories from an increasingly wide variety of fields, including psychology, economics, philosophy, and political science. However, researchers interested in legal thought have thus far paid relatively little attention to its origins in development. This chapter highlights an emerging approach that leverages methods and insights from developmental science to better understand the nature and development of adult intuitions about the law. By studying children’s earliest intuitions about rules, laws, and other topics, this “intuitive jurisprudence” approach can provide new methods and theoretical frameworks for experimental jurisprudence, as well as clarify places in which the law does or does not match human intuitions about justice. Already, developmental psychology and legal scholarship may converge to be mutually informative in a number of diverse areas, and this chapter reviews several, including: intent and punishment; fairness and procedural justice; ownership and property rights; trust in testimony and evidentiary issues; and social biases and equal protection under the law.

The series of cases discussed in Part III are humbling reminders of how intertwined our patients and their support systems are with healthcare practitioners. TJ, Jimmy, Mrs. Blue, and Mrs. Winthorpe all have unique experiences in different corners of the healthcare system. Each case touches on the familiar experience of a healthcare team identifying what they believe is in the best interest of patient, and there being a factor, often the patient themselves, complicating that coming to fruition. Their experiences, and different experiences of privilege and power, or disempowerment are salient elements of their stories. These “haunting” and morally distressing cases are revisited with an additional lens of diversity, equity, identity, and bias and considerations for how ethicists might more fully integrate these critical perspectives into ethics consultation.

Bias is a topic that has received intense academic study, but its importance within experimental jurisprudence has yet to be unpacked. To fill this lack, we make the following contributions in this chapter. First, we situate the topic within this newly named – but not necessarily new – academic movement: We present recent research on bias in the law and discuss whether it rightly fits within the remit of experimental jurisprudence. Second, continuing to draw on this recent research, we unpack issues that inhere to explorations of bias, ones that are important for understanding, in the experimental jurisprudence context, participants and the data they generate as well as researchers and the data they garner and interpret. Finally, we conclude by offering words of caution and guidance as bias research within experimental jurisprudence progresses.

As part of the legal test for bias, the courts have created a fictional fair-minded observer (the FMO) to act as a conduit for reasonable public perception. A number of scholars have raised concerns that the FMO bears no resemblance to an average member of the public or reasonably reflects general public opinion. This chapter presents our original empirical pilot study on expert versus lay attitudes to judicial bias. The study compares responses of legal insiders (lawyers and judges) and nonlegal experts with a basic understanding of the law (law students) to leading cases on judicial recusal. We use vignettes based on real cases from England, Australia, and Canada that dealt with different claims of judicial bias (covering issues of race, prejudgment, and more). The study may allow us to draw conclusions about the similarities and differences between legal experts and laypeople in relation to the perception of judicial bias, and we suggest ways the full study can address methodological limitations in the pilot that would allow us to draw those conclusions with greater confidence.

For a case-control study to be a suitable design, we need a good idea about the outcome of interest (or condition) described by a strong case definition. But what if we know quite a bit about the exposures we are interested in, but we are a little hazy on the potential outcomes associated with those exposures? If we consider a scientific question like the one posed in this chapter – What happens if you eat pizza and chips every day?’ – we have specifically identified the exposures of interest, but can only guess what the outcomes might be. Okay, we could probably make fairly educated guesses about some of the potential outcomes (weight gain being foremost among these), but there remains a level of uncertainty about their timing, magnitude and variety. What is really needed to answer a question like this is a ‘cohort study’, a type of observational study in which ‘cohorts’ of people (population groups who share certain characteristics, such as being in the same work environment, or who are born in the same year) are sorted into groups on the basis of whether they have or have not been exposed to specific health-related factors.

In Chapter 6 we heard about how we can identify and quantify associations between exposures and health outcomes within populations, and even between countries. We learnt how useful cross-sectional studies were for looking at a range of risk factors and outcomes as they exist in a defined population at a particular point in time. While they have a great number of advantages, it can sometimes be difficult to sort out the direction of the relationships identified using cross-sectional approaches – that is, current risk factors or exposures may not necessarily have caused current outcomes or diseases. If we want to move towards thinking about potential causal relationships, we need an approach that allows us to determine the relative strength of relationships between exposures and outcomes and provide some hints about temporality – that is, to give us a start on determining if the exposure preceded the health event. We will need this type of study to address question posed for this chapter – what might be causing all those headaches that health science students seem to complain about.

Arriving at evidence-based solutions requires strong evidence. Usually, this evidence will be derived from quality research, such as is often published in reputable scientific journals. But how do we know whether even these studies are good through and through? There is always the potential that pesky flaws, such as bias and confounding, might can beset even the most (otherwise) perfect of studies. This is why the methods taken to avoid bias and confounding are always well-described in all good published studies, as is the potential for remaining sources of error for which the design is (inevitably) unable to account, but which might still influence findings. There is always a bit of uncertainty about any evidence provided by studies and, to add to this, the very real possibility that we are not getting the full story at all times. In a phenomenon known as ‘publication bias’, even really high quality studies may not get published if they report non-significant results.

This chapter analyzes challenges to AI decision-making based on anti-discrimination in the US, the UK, and Australia. Machine learning algorithms can be trained on datasets that contain human bias, thus resulting in predictions that are tainted with unfair discrimination. Anti-discrimination claims involve challenging the inputs of decision-making, such as the data or source code, and arguing that the flawed algorithm or data inputted into the AI system leads to discriminatory outcomes.

This book is about the science and ethics of clinical research and healthcare. We provide an overview of each chapter in its three sections. The first section reviews foundational knowledge about clinical research. The second section provides background and critique on key components and issues in clinical research, ranging from how research questions are formulated, to how to find and synthesize the research that is produced. The third section comprises four case studies of widely used evaluations and treatments. These case examples are exercises in critical thinking, applying the questions and methods outlined in other sections of the book. Each chapter suggests strategies to help clinical research be more useful for clinicians and more relevant for patients.

In order to examine our three questions, we need objective research methods. Estimating whether a treatment can work, does work, and has value requires a wide range of research strategies. Evidence establishing that a treatment works under controlled conditions does not necessarily assure benefits when the intervention is applied in clinical practice. This chapter considers the development of a research protocol, and biases that might be attributable to participant recruitment, enrollment, retention, and dissemination of findings. In practice, establishing the value of a treatment should consider an examination of the existing literature, development of thorough research plans, recognition of the strengths and weaknesses of the chosen research methods, and integration of study results within a wider body of knowledge. We challenge beliefs in a hierarchy of methods that assumes some methods, such as the RCT, are free from bias.

Gastroesophageal reflux disease is a common condition that can be controlled with proton pump inhibitors such as omeprazole. We examine randomized controlled trials (RCTs) of omeprazole and find stronger evidence of efficacy among RCTs with industry support than without. The participants in these trials were unlike most people who take proton pump inhibitors, raising questions about the external validity of RCTs. Furthermore, use of these medicines is associated with short- and longer-term adverse effects. Healthy behavior change, such as weight loss, holds promise as an alternative to proton pump inhibitors.

After reviewing a wide range of topics, we conclude that good science requires greater efforts to manage biases and to promote the ethical conduct of research. An important problem is the belief that randomized controlled trials (RCTs) are exempt from systematic bias. Throughout the book, we acknowledge the importance of RCTs, but also emphasize that they are not immune from systematic bias. A second lesson concerns conflict of interest, which must always be taken seriously. Most large RCTs are sponsored by for-profit pharmaceutical companies. We identify leverage points to address these problems. These include cultivating equipoise – the position that research investigators enter a study with the understanding that either a positive, negative, or null result is of value. We return to several other themes prominent throughout this book, including the reporting of research findings and serious problems with our system of peer review. The book concludes with recommendations for reducing conflicts of interest, improving transparency, and reimagining the peer review system.

Research is about asking and answering questions. One of the most important investments of time for a research investigator should occur before the study starts. This chapter considers the importance of well-defined research questions that have clear boundaries and scope. The specifics of the research methodologies such as sample size and data analysis are essential for high-quality research. Yet less emphasis is placed on the importance of the research question, the feasibility of the study, and the social impact of the investigation. This chapter argues that clinical research should be person- and community-centered. The population, intervention, comparator, outcome, and timeframe (PICOT) framework encompasses content that may be informative for those who use health care. The feasible, interesting, novel, ethical, and relevant (FINER) framework comes closer to focusing on questions and outcomes of importance to study participants. We offer a BASES (biases, awareness, social, equilibrium, specificity) model that builds on the FINER and PICOT systems to place greater emphasis on social context.