The research career of Professor Treur recapitulates the history of artificial intelligence (AI). His PhD work more than 40 years ago focused on the formal logical systems in vogue in the “neat” school of AI. This approach viewed intelligence as a highly specialized computational function in the brain, largely isolated from other aspects of human existence. Today, AI draws inspiration from biology and sociology, as well as mathematics, and reflects our growing awareness of the integrated nature of human cognition. This book brings together Treur’s work over the past decade to develop a common framework, inspired by neural networks, to model cognitive, emotional, and social phenomena. Having a common framework allows a unified theory in which these different effects can interact holistically.

The first two chapters motivate the approach and describe the underlying formalism. Treur calls his approach “temporal-causal networks.” The nodes of such networks are real-valued state variables. A directed edge from one node to another indicates that the source node can modify the value of the target node, with a level of influence indicated by a weight on the edge. In addition to its value, each node has a speed factor that controls how rapidly it changes and a combination function that indicates how multiple inputs combine to modify the node’s value.

The next nine chapters develop temporal-causal networks to model scenarios in two areas. Chapters 3 through 6 model emotional dynamics. Treur argues that emotions are the “glue” unifying all mental and social processes, and exhibits specific models explaining dreaming, the value of dreams for fear extinction, and how emotions undergird rational decision making. Chapters 7 through 11 model social interactions, working from the perspective that internal emotional and cognitive processes allow us first to mirror the thoughts of others, and then to assume responsibility for our own actions, to have empathy with others, and to make decisions based on our understanding of ourselves and others, including dynamically adjusting the network of external relations we maintain with other people.

The next three chapters discuss analysis methods for the temporal-causal formalism, including mechanisms for capturing emergent dynamics and for calibrating a network against real-world observations.

The last four chapters of the book do not consider specific models or their evaluation, but rather look at the philosophical and cultural implications of such models. Chapter 15 is an extensive exploration of the nature of causality and how it aligns with the formalism. Two chapters look at the implications of such models for smart applications and education, while the final chapter summarizes the overall approach.

Treur’s temporal-causal networks are almost identical to the stock-and-flow diagrams developed by Jay Forrester [1] and widely applied under the name of system dynamics (SD). Treur’s nodes are the stocks of a stock-and-flow model, while his edges correspond to the flows, and the weights are a simplified form of Forrester’s decision functions (sometimes called rate equations) on flows. The speed factors in a temporal-causal network are closely related to stock-and-flow delays. Treur’s combining functions offer a richer set of options than do stock-and-flow diagrams, in which stocks simply sum their various inputs, but in most cases Treur does not allow state variables to modulate gains, while in stock-and-flow models the feedback from stocks to decision functions is ubiquitous. Both formalisms lead naturally to representation as difference or differential equations, allowing simulation by numerical integration.

Treur mentions some of Forrester’s articles in describing the general movement toward dynamic models of causation, but does not discuss the close parallel between the two formalisms, which is perhaps not surprising given the very different applications to which they are applied. Forrester developed his approach in the mid-1950s to exploit the (then new) insights of mathematical control theory to understand the dynamics of business systems, such as employment levels and inventory changes. The method was then extended to problems such as global sustainability. For Forrester and his colleagues, human cognition is modeled entirely in the decision functions that modulate flows between stocks. By contrast, Treur’s state variables are elements of individual human cognition, and each of his models seeks to explain how an individual performs some cognitive or emotional task. The formal similarity of the two kinds of models would make it straightforward to link the micro and macro models together, providing a management or sustainment model with a much richer representation of underlying cognitive dynamics, while in turn coupling the cognitive model to the emergent dynamics imposed by flows at the macro level. One looks forward to another generation exploring this generalization.

The book is a collection of relatively independent chapters, each with its own bibliography. There is a brief index (five pages for a 500-page book), but no integrated bibliography, and no exercises, so its use in a classroom setting would require supplementary material. The volume will be of great interest to modeling practitioners and cognitive scientists, who will find great stimulation in the explanations it offers of various cognitive dynamics.