How do you represent knowledge about a world to make it easy to acquire, debug, maintain, communicate, share, and reason with that knowledge? This chapter explores how to specify the meaning of symbols in intelligent agents, how to use the meaning for knowledge-based debugging and explanation, and, finally, how an agent can represent its own reasoning and how this may be used to build knowledge-based systems. As Smith points out in the quote above, the problems of ontology are central for building intelligent computational agents.

Knowledge Sharing

Having an appropriate representation is only part of the story of building a knowledge-based agent. We also should ensure that the knowledge can be acquired from people and from data. The knowledge for any non-trivial domain comes from diverse sources and at multiple points in time. Multiple sources need to interoperate, to work together, at both a syntactic level and a semantic level.

Recall (page 67) that an ontology is a specification of the meanings of the symbols in an information system. Here an information system can be a knowledge base, a sensor such as a thermometer, or some other source of information. The meaning of the symbols is sometimes just in the mind of the knowledge base designer, in a user manual, or in comments with the knowledge base. Increasingly, the specification of the meaning is in machine-interpretable form. This formal specification of the meaning is important for semantic interoperability – the ability of different knowledge bases to work together at a semantic level so that the meanings of symbols are respected.

Agents in real environments are inevitably forced to make decisions based on incomplete information. Even when an agent senses the world to find out more information, it rarely finds out the exact state of the world. For example, a doctor does not know exactly what is going on inside a patient, a teacher does not know exactly what a student understands, and a robot does not know what is in a room it left a few minutes ago. When an intelligent agent must act, it has to use whatever information it has. This chapter considers reasoning with uncertainty that arises whenever an agent is not omniscient. This is used in Chapter 9 as a basis for acting with uncertainty. This chapter starts with probability, shows how to represent the world by making appropriate independence assumptions, and shows how to reason with such representations.

Probability

To make a good decision, an agent cannot simply assume what the world is like and act according to that assumption. It must consider multiple hypotheses when making a decision. Consider the following example.

This chapter considers a simple form of a knowledge base that is told what facts and rules hold in the world. An agent can use such a knowledge base, together with its observations, to determine what else must be true in the world. When queried about what must be true given a knowledge base, it answers the query without enumerating the possible worlds, or even generating any possible worlds. This chapter presents a number of reasoning formalisms that use propositions. They differ in what is being proved, what background knowledge must be provided, and how observations are handled.

Propositions

Statements about the world provide constraints about what could be true. Constraints could be specified extensionally (page 130) as tables of legal assignments to variables, or intensionally (page 130) in terms of formulas. For many domains propositions provide a suitable language for providing intensional constraints.

There are a number of reasons for using propositions for specifying constraints and queries:

This book is about artificial intelligence, a field built on centuries of thought, which has been a recognized discipline for over 60 years. As Buchanan points out in the quote above, we now have the tools to test hypotheses about the nature of thought itself, as well as to solve practical tasks. Deep scientific and engineering problems have already been solved and many more are waiting to be solved. Many practical applications are currently deployed and the potential exists for an almost unlimited number of future applications. In this book, we present the principles that underlie intelligent computational agents. These principles can help you understand current and future work in AI and equip you to contribute to the discipline yourself.

What is Artificial Intelligence?

Artificial intelligence, or AI, is the field that studies the synthesis and analysis of computational agents that act intelligently. Let us examine each part of this definition.

An agent is something that acts in an environment; it does something. Agents include worms, dogs, thermostats, airplanes, robots, humans, companies, and countries.

We are interested in what an agent does; that is, how it acts. We judge an agent by its actions.

An agent acts intelligently when

The representation dimension (page 23) has, as its top level, reasoning in terms of individuals and relations. Reasoning in terms of relations allows for compact representations that can be built before the agent encounters particular individuals. When an agent finds out about an individual, it can make inferences about that individual. This chapter outlines how, in planning, learning and probabilistic reasoning, feature-based representations can be expanded to deal also with individuals and relations. In each of these areas, the relational representation benefits from being able to be built before the individuals are known and, therefore, before the features are known. As Russell points out in the quote above, relational reasoning brings great advantages over propositional and feature-based representations.

Planning with Individuals and Relations

A robot that can deliver parcels to people needs a model of the world before it knows which parcels exist and which people may need deliveries. It might need to be programmed before it knows the environment which it will inhabit. A tutoring system needs to work for multiple students and multiple problems, and needs to be programmed before it knows about the students and all of the problems. A purchasing agent, when it is being designed and built, will not know about all of the hotels and rooms it can book, and will not know about the people and their goals or preferences. In all of these cases, the agent's goals and its environment are described in terms of individuals and relations.

Artificial Intelligence: Foundations of Computational Agents is a book about the science of artificial intelligence (AI). AI is the study of the design of intelligent computational agents. The book is structured as a textbook but it is designed to be accessible to a wide audience.

We wrote this book because we are excited about the emergence of AI as an integrated science. As with any science being developed, AI has a coherent, formal theory and a rambunctious experimental wing. Here we balance theory and experiment and show how to link them together intimately. We develop the science of AI together with its engineering applications. We believe the adage, “There is nothing so practical as a good theory.” The spirit of our approach is captured by the dictum, “Everything should be made as simple as possible, but not simpler.” We must build the science on solid foundations; we present the foundations, but only sketch, and give some examples of, the complexity required to build useful intelligent systems. Although the resulting systems will be complex, the foundations and the building blocks should be simple.

This second edition results from extensive revision throughout the text. We have restructured the material based on feedback from instructors who have used the book in classes. We have brought it up to date to reflect the current state of the art, made parts that were difficult for students more straightforward, added more intuitive explanations, and coordinated the pseudocode algorithms with new open-source implementations of the algorithms in Python and Prolog. We have resisted the temptation to just keep adding more material. AI research is expanding so rapidly now that the volume of potential new text material is vast. However, research teaches us not only what works but also what does not work so well, allowing us to be highly selective. We have included more material on machine learning techniques that have proven successful. However, research also has trends and fashions. We have removed techniques that have been shown to be less promising, but we distinguish them from the techniques that are merely out of fashion. We include some currently unfashionable material if the problems attacked still remain and the techniques have the potential to form the basis for future research and development.