Chapter 4 examines communicative settings where the use of machine translation is particularly likely to involve high levels of risk. The chapter looks at guidelines about machine translation use and at the issue of consent. Two types of consent are examined, namely using machine translation to seek some type of consent, and consent that concerns whether the use of machine translation itself is consensual. The chapter then explores some of the direct reasons why machine translation is used in high-risk scenarios. These reasons include urgency, service user preferences and unreliable human language services. The project’s participating professionals were not short of stories to tell about human interpreters who had not turned up for appointments or telephone interpreting connections that frequently crashed. Incidents of this nature are considered within a broader context where limited resources and outsourced human language services normalise the reliance on machine translation in ways that increase risks and affect standards of care.

This book concludes with a summary of key types of machine translation use discussed in the project. The conclusion outlines ethical principles of multilingual AI use including the potential or intended legal value of a message, the stability of the information and its potential to be reused, and the need for any uses of machine translation to be transparent and as far as possible consensual and cybersecure. The conclusion also examines some of the challenges posed by the broader project. It offers a reflection on the question of accountability and on the difficulties of living well with technology in environments that elevate cost efficiency above other values.

Compared to shallow infusion, semi-deep infusion concerns opening the blackbox of AI systems using knowledge sources. This chapter will provide a detailed grounding of model interpretability, highlighting the state-of-theart methods that promise interpretable AI, the limitations of these methods, and how knowledge infusion in AI can help make models interpretable without sacrificing uncertainty while handling context sensitivity, and userlevel explainability.

To create genuinely effective artificial intelligence (AI), we need to build systems that think, explain, and reason like humans. This perspective from Gary Marcus aligns with Andrew Ng’s view that the hype around big data is overblown and that AI must advance in intelligence. Early on in the heyday of machine learning (ML), Pedro Domingos (2012) observed that “Data alone is not enough.” These experts agree that merely scaling up AI models to billions of parameters has led to fundamental challenges such as “hallucinations” (Thoppilan et al., 2022a).

Explainability and safety engender trust. These require a model to exhibit consistency and reliability. To achieve these, it is necessary to use and analyze data and knowledge with statistical and symbolic AI methods relevant to the AI application – neither alone will do. Consequently, we argue and seek to demonstrate that the neurosymbolic AI approach is better suited for making AI a trusted AI system. We present the CREST framework that shows how consistency, reliability, user-level explainability, and safety are built on neurosymbolic methods that use data and knowledge to support requirements for critical applications such as health and well-being. This chapter focuses on large language models (LLMs) as the chosen AI system within the CREST framework. LLMs have garnered substantial attention from researchers due to their versatility in handling a broad array of natural language processing (NLP) scenarios. For example, ChatGPT and Google’s MedPaLM have emerged as highly promising platforms for providing information in general and healthrelated queries, respectively. Nevertheless, these models remain black boxes despite incorporating human feedback and instruction-guided tuning. For instance, ChatGPT can generate unsafe responses despite instituting safety guardrails. CREST presents a plausible approach harnessing procedural and graph-based knowledge within a neurosymbolic framework to shed light on the challenges with LLMs.

A knowledge graph (KG) is a machine-readable structured representation of knowledge consisting of entities (entity and entity type) and relationships in various forms (e.g., labeled property graphs and resource description frameworks (RDFs)) (Sheth et al., 2019b). KiL based on Machine Learning/Deep Learning seamlessly integrates external knowledge to address challenging problems in low-resource and open-domain natural language processing tasks and domain-specific problems. Domain-specific problems require the application of task-specific knowledge (implicit or explicit) to generic AI models. For example, to detect emerging events in a stream of crisis-related posts (e.g., Hurricane, COVID-19 Pandemic), a generic language model (e.g., Word2Vec Mikolov et al., 2013, BERT) can be fine tuned using the concepts and relationships found in disaster ontology (e.g., empathi from Gaur et al., 2019a). Lowresource problems are characterized by having few labeled samples, making further labeling difficult in terms of effort, quality, and time. For instance, annotating millions of posts from users in various mental health communities on Reddit would require (a) establishing guidelines for annotation, (b) training annotators, (c) resolving annotation conflicts, and (d) enriching quality over multiple iterations to achieve high annotator agreement. A study by Gaur et al. (2021b) proposed a KiL pipeline to annotate such extensive social data at scale, shifting the human role from annotators to evaluators.

Traditional dialog agents in conversational information seeking have repeatedly focused on entities in the user query (Rao and Daumé III, 2018; Zamani et al., 2020). Consequently, the generated questions are redundant and lack diversity, losing user engagement.

Process knowledge is an ordered set of information that maps to evidencebased guidelines or categories of conceptual understanding to experts in a domain. For instance, The American Academy of Family Physicians (AAFP) develops clinical practice guidelines (CPGs) that serve as a framework for clinical decisions and supporting best practices. CPGs allow systematic assessment to optimize patient care. In addition, the US Departments of Agriculture (USDA) and Health and Human Services (HHS) develop dietary guidelines for Americans that serve as a recommendation for meeting nutrient needs, promoting health, and preventing disease. An AI system adapted to process knowledge can handle uncertainty in prediction, and the predicted outcomes are safe and user-level explainable. Further, an AI system can consider process knowledge as meta-information to capture the sequential context necessary for carrying out a structured conversation. Also, it allows the developer of the AI system to probe its internal decision making using application-specific guidelines or specifications that inform the synchrony between the end-user’s thought process and the model’s functioning.

A major focus in this chapter will be to learn various ways in which datasets can be transformed using external knowledge. Shallow infusion concerns semantic data transformation and provides the following benefits:

We define the third category of knowledge infusion, that is, deep infusion of knowledge, as a paradigm that couples the latent representation learned by deep neural networks with KGs exploiting the semantic relationships between entities. This chapter will provide a theoretical background to achieve deep infusion, as illustrated in Figures 6.1 and 6.2. We aim to:

Humans interact with the environment using a combination of perception – transforming sensory inputs from their environment into symbols – and cognition – mapping symbols to knowledge about the environment for supporting abstraction, reasoning by analogy, and long-term planning. Human perceptioninspired machine perception, in the context of AI, refers to large-scale pattern recognition from raw data using neural networks trained using self-supervised learning objectives such as next-word prediction or object recognition. In contrast, machine cognition encompasses more complex computations, such as using knowledge of the environment to guide reasoning, analogy, and long-term planning. Humans can also control and explain their cognitive functions. This seems to require the retention of symbolic mappings from perception outputs to knowledge about their environment. For example, humans can follow and explain the guidelines and safety constraints driving their decision making in safety-critical applications such as healthcare, criminal justice, and autonomous driving. While data-driven neural network-based AI algorithms effectively model machine perception, symbolic knowledgebased AI is better suited for modeling machine cognition. This is because symbolic knowledge structures support explicit representations of mappings from perception outputs to the knowledge, enabling traceability and auditing of the AI system’s decisions. Such audit trails are useful for enforcing application aspects of safety, such as regulatory compliance and explainability, through tracking the AI system’s inputs, outputs, and intermediate steps. This chapter introduces neurosymbolic AI, combining neural networks and knowledgeguided symbolic approaches to create more capable and flexible AI systems. These systems have immense potential to advance both algorithm-level (e.g., abstraction, analogy, reasoning) and application-level (e.g., explainable and safety-constrained decision-making) capabilities of AI systems.