This chapter serves as a primer for understanding the sweeping implications of generative AI (GenAI). It begins by establishing a clear terminological foundation, distinguishing modern GenAI from its predecessors, good old-fashioned AI (GOFAI) and machine learning. While GOFAI and machine learning solved complex, specific problems, GenAI’s ability to generate novel outputs, coupled with its cheap, scalable, and versatile nature, has rapidly woven it into the fabric of society. The text emphasizes that for every potential beneficial use, there are equally foreseeable harms, such as the spread of deepfakes and the weaponization of biases. The chapter argues that GenAI is not a neutral tool and that its deployment is often driven by corporate interests that externalize risks onto society. To foster a more informed public discourse, the chapter introduces Neil Postman’s seven questions for evaluating new technology, urging readers to move from being passive subjects to active participants in the democratic oversight of GenAI’s development.

Generative AI systems offer unprecedented capabilities for addressing access to justice challenges through their ability to understand and generate human-like text. This chapter examines how large language models (“LLMs”) can bridge the language gap between laypeople’s factual descriptions and legal concepts, providing personalized legal information and assistance with legal tasks.

The analysis explores several applications, including the automated provision of legal information using retrieval-augmented generation (“RAG”), streamlined legal intake processes, intelligent form completion assistance, and AI-enhanced mediation. These systems show promise for identifying legal issues from plain language descriptions, assisting with the drafting of legal documents, and supporting dispute resolution processes.

However, significant challenges remain, particularly regarding hallucinations where models generate convincing but incorrect information, algorithmic bias that may perpetuate existing inequalities, and transparency concerns inherent in black-box AI systems. While generative AI holds promise for democratizing legal assistance, implementation with robust safeguards, continuous evaluation, and human oversight is essential to ensure these tools enhance rather than undermine access to justice.

This chapter examines technical methods for evaluating AI applications in legal contexts, and addresses the need for standards to assess AI system performance in an industry where quality standards for human work remain largely subjective. It identifies three primary evaluation approaches: categorical evaluation using precision, recall, and F1-score metrics; numerical evaluation of continuous predictions; and more complex evaluation methods of open-texture outputs.

The chapter demonstrates these methods through case studies, including technology-assisted review in e-discovery, legal research tools with varying accuracy rates and significant hallucination challenges, and access to justice applications like Beagle+. These applications show how iterative improvements can substantially increase accuracy through enhanced prompting and retrieval-augmented generation.

The analysis emphasizes that evaluation must extend beyond technical accuracy to include real-world utility, consumer protections, and security considerations. The chapter concludes that establishing robust evaluation frameworks is essential for responsible AI deployment in legal services, particularly given the legal profession’s current reliance on subjective quality assessments rather than empirical standards.

This chapter provides a foundational introduction to generative AI and large language models, addressing their capabilities and limitations. These models can perform competitively with humans on standardized exams and legal reasoning tasks and can be applied flexibly to legal challenges through natural language prompts.

However, significant limitations constrain their deployment in legal contexts. Models produce hallucinations, generating confident but incorrect information that can mislead users making critical legal decisions. The ELIZA effect compounds this risk as users tend to anthropomorphize AI and overestimate its capabilities. Additional challenges include potential biases introduced by the training data, privacy concerns associated with hosted models, and substantial computational costs. Understanding both the opportunities and limitations is crucial for safely deploying generative AI to address access-to-justice challenges.

This chapter offers a comprehensive overview of large language models (LLMs), examining their theoretical foundations, core mechanisms, and broad-ranging implications. We begin by situating LLMs within the domain of natural language processing (NLP), tracing the evolution of language modeling from early statistical approaches to modern deep learning methods.</p>The focus then shifts to the transformative impact of the Transformer architecture, introduced in the seminal paper Attention Is All You Need. By leveraging self-attention and parallel computation, Transformers have enabled unprecedented scalability and efficiency in training large models.</p>We explore the pivotal role of transfer learning in NLP, emphasizing how pretraining on large text corpora followed by task-specific fine-tuning allows LLMs to generalize across a wide range of linguistic tasks. The chapter also discusses reinforcement learning with human feedback (RLHF)—a crucial technique for refining model outputs to better align with human preferences and values.</p>Key theoretical developments are introduced, including scaling laws, which describe how model performance improves predictably with increased data, parameters, and compute resources, and emergence, the surprising appearance of complex behaviors in sufficiently large models.</p>Beyond technical aspects, the chapter engages with deeper conceptual questions: Do LLMs genuinely "understand" language? Could advanced AI systems one day exhibit a form of consciousness, however rudimentary or speculative? These discussions draw from perspectives in cognitive science, philosophy of mind, and AI safety.</p>Finally, we explore future directions in the field, including the application of Transformer architectures beyond NLP, and the development of generative methods that extend beyond Transformer-based models, signaling a dynamic and rapidly evolving landscape in artificial intelligence.

Galen system is based on three pillars: the affected body part, the type of qualities imbalanced, and the degree of imbalance. Therefore, he only distinguishes between mental illness and impaired consciousness when there is a difference between these two entities in any of these three pillars. Thus, he distinguishes phrenitis from melancholia but not from mania. The emphasis on the system, on the other hand, enables him a very tight notion of disease, where symptoms, mechanisms, affected organ and treatment are closely linked.

Like their forerunners, post-Hellenistic doctors also grappled with the unclear boundaries between healthy versus pathologic sleep, and consciousness-unconsciousness. Furthermore, they incorporated new diseases and redefined others - like lethargy - that were specifically associated with this process. Celsus considered sleep as all-or-nothing phenomenon, without recognising different depths. Regarding mental capacities, he subsumed most of them in his idea of mens/animus. Aretaeus, on the other hand did conceive different depths of sleep, and his eclectic method enabled him to find alternative pathophysiological explanations to characterise several of its main features. Similarly, although his organization of mental capacities varied according to what he was explaining, the opposition gnômê-aisthêsis was important in his idea of mind.

4 Post-Hellenistic authors present a more compartmentalised idea of diseases in general and of impaired consciousness in particular. Unlike the Hippocratics, who barely discussed mental illness, these authors did distinguish impaired consciousness from mental illness through a classificatory system of dichotomic oppositions, additionally they discussed new conditions which are not mentioned in the HC. In most theorisations, perceptions play an increasingly relevant role to understand these conditions.

Some Hippocratic doctors regarded sleep as a healthy process, and some as a pathological one; some of them struggled to distinguish between hallucinations and nightmares, and some between deep dreamless sleep and total loss of consciousness. This chapter explores how different treatises from the Hippocratic corpus navigated these ambiguities, how they explained different depth of sleep (i.e. different levels of consciousness), and how such understanding relates to their views on mental capacities (which they subsumed in concepts such as phronesis, sunesis, gnômê, and nous).

Contrary to mainstream scholarship’s opinion, the Hippocratic corpus presents many cases of impaired consciousness, but only a few of mental illness. By looking at three study cases, this chapter describes how these doctors understood conditions where patients act weirdly or were not their usual selves, and how they construed the notion of disease.

Galen conceived sleep and wakefulness as a continuum that depended on the mixture of qualities within the ruling part of the puschê (the hêgemonikon) located in the brain. Naturally, in his system whenever pathological sleep occurred the doctor needed to determine if the brain was affected directly or by sympathy (from another organ), and the precise imbalance of qualities that needed to be counteracted by their opposites. His idea of mind was very accurately and hierarchically structured: it resided in the logical part of the soul, located in the brain, and several diseases with impaired consciousness compromised its normal functioning.

Chapter 2 covers the topic of delusional disorder. Through a case vignette with topical MCQs for consolidation of learning, readers are brought through the management of a patient with delusional disorder from first presentation to subsequent complications of the conditions and its treatment. Things covered include the different forms of delusions, psychopathology, differences with schizophrenia, co-morbid conditions, the use of pharmacological treatment including depot antipsychotics, adverse effects of commonly used medications, the use of antipsychotics in pregnancy and post-partum psychosis.

Chapter 1 covers the topic of schizophrenia. Through a case vignette with topical MCQs for consolidation of learning, readers are brought through the management of a patient with schizophrenia from first presentation to subsequent complications of the conditions and its treatment. Things covered include the different forms of delusions, psychopathology, negative and positive symptoms of schizophrenia, co-morbid conditions, typical investigations carried out, the use of pharmacological treatment, adverse effects of commonly used medications, extrapyramidal side effects and treatment-resistant schizophrenia.

Results from studies on brain activity in situations of hypoxia, application of anesthetics and other psychoactive drugs, epileptic seizures, electrical stimulation of brain areas, lucid dreams, and dream-like hallucinations of several geneses have shown that the reports of people who had perceptions and experiences related to these situations showed strong accordance with NDE reports. NDE themes can be reproduced experimentally, often in a predictable way. Such contexts and situations can be used as scientifically appropriate models for NDE release. Knowledge about and control of brain activation during the occurrence of NDE-like phenomena can be essential for understanding their generation.