According to the authors, Dan Curtis proposed artificial psychology as a theoretical discipline in 1963 (p. 164). During the 56 years since, it can be expected to have broadly expanded in both complexity and overarching clarity; this book demonstrates both to some degree. However, it always keeps its focus on the future need for a true social science of artificial psychology to support forms of artificial intelligence (AI) not yet realized.

The work consists of extensive discursive and experimental contributions, but the numerous networking threads are perhaps more significant. As AI systems increase in complexity, more and more traits resemble issues addressed in psychology. Besides the technical challenges of integrating AI systems into society, “we must also understand how such a system would be received and perceived by people and how we expect any type of artificially intelligent system to react to and perceive people” (p. 163). There is also some fun here, for example, the discussions of machine analogs of self-soothing, deep breathing, acupressure, and mindfulness in chapter 8.

Having established that the book serves a great purpose in bringing together varieties of systems thinking and methodology, let me turn to some of the difficulties. This is not a primer on the subject, or even a survey that would appeal to a broad audience. I am not a stranger to AI--I was educated under Marvin Minsky and John McCarthy, and I occasionally taught it over two decades--but I am unable to follow a number of discussions. When the writing is clear but content-free, the reading becomes less rewarding. This makes me a fan of the last chapter (14), “Conclusions and Next Steps.” The summaries, arguments, and reservations are finely put, clear, and mostly encouraging.

The book is plagued by Springer typographical errors, acronyms before their expansions, an incomplete index, and diagrams that have drawing errors or lack immediate interpretability. The obscurity of the diagrams seems to arise from their being extracted or adapted from a number of disciplines and presented without explanation. But there are also such things as the double knowledge transition strategy box in Figure 1.4, besides the issue of how this can run in a cycle as drawn.

Here is the lead sentence of a paragraph on page 19: “The Internal Family Systems Theory (IFST) is an individual theory that relates system within.” The remainder of the paragraph gives a better introduction to IFST, a fine therapy, and its applicability. But it ends with this awkward sentence: “Now IFST is a newer theory than the feedback loops but very related.” Note that the earlier feedback discussion claimed that negative feedback leads to being off course, out of balance, whereas it is the essence of control.

Here is one last example:

Learning without context can lead to serious decision and inference problems within the system. Instead, the systematic approach presented here, combining the RNA contextual approach, with a cognitive framework, in the artificial intelligence system, allows the framework that can handle cognitive processing of information and context, turning them into actionable intelligence. (p. 119)