Galitsky summarizes research and practice from two decades aimed at describing, defining, implementing, and applying computer systems that model, simulate, elucidate, and rehabilitate autism. While “rehabilitation” is a frequently stated goal, it feels like the term of a neurotypical. “Guided education” sounds better to this neurotypical. As a parallel goal, the book outlines the effort to trace and train interactions among multiple artificial intelligence (AI) or robotic agents. The quote in the front matter summarizes this connection: “To teach a child with autism, one might first consider how to program a robot.” Part of the value of the book is learning that this statement involves no condescension.
The first two chapters introduce diverse accounts and theories of autism, including a defective theory of mind (ToM)--whereby one person knows that another person thinks, reasons about his own and others’ mental states, and more. The next three chapters recount extant views, computational aspects, and a machine embodiment of ToM. The sixth chapter adds the physical world to the mental world. Chapter 7 is devoted to handling autistic cognition with a hybrid system that utilizes active learning and formal reasoning via logic programming based on antiunification, with the goal of training action adjustment in subjects.
Chapter 8 shifts to real applications for real people, providing examples of training children with autism using computer-based tools and other tools constructed with the insight of the author and his collaborators. These include emotional remediation, expression of feelings, and language improvement. The ninth chapter closes the book by connecting reasoning with behavior in the real world, including discussions of causes, diagnosis, cases, and applied behavior analysis.
I hope the reader is impressed by the author’s dedication and determination to produce tools that can provide (and have provided) benefit to many children and adults with autism. There is a lot to learn about how that is done and about the author’s own exploration in theoretical and face-to-face settings by reading this book ... except this book is unreadable. Let me first describe the conceptual difficulty of reading it. It combines very technical themes from psychology, education, rehabilitation/remediation, computer science, AI, and logic programming. I am not a novice in the last three, but the entire text is a real challenge nonetheless. Yet there is much to be gained through this book.
Let’s turn to the more difficult and pervasive aspect of the unreadability: the innumerable typographical errors, grammatical errors, and infelicitous wordings surviving because of an almost total lack of editing. I do not know how a prestigious house like Springer could so ill serve an author by releasing a book that has an error rate between one and five per hundred words. As long as I look fuzzily or cross-eyed at the text, it goes well, but if I have to focus on it, then I begin to wonder whether I am in fact neurotypical.
I kept a log of more egregious errors in order to recount them here, but gave up. Some are attributable to Russian phrasing, but multiple issues with number, case, gender, article, referent, and other syntax should have been rectified by professional copy editing. This is not just a written word problem; there are also instances in the technical expressions of logic, diagrams, and programming. Here are four diverse examples: (1) page 23, Figure 2.3 “UTERAL LANGUAGE” (literal?); (2) page 108 (last example) ADJ is actually an ADV (the table is a garbled misappropriation from the cited Colorado website); (3) page 120, bottom: confusion of “who” and “whom” and “know” and “now” (and later, on page 163, “week” and “weak”); and (4) page 138, line 5b: “Nick presents his communicative actions of herself.”
More reviews about this item: Amazon
