In a later chapter in their book Algorithms of Education, Kalervo N. Gulson, Sam Sellar, and P. Taylor Webb describe the Class Care System in Chinese schools, where software analyzes classroom footage to categorize student states into behavioral categories like “listening,” “writing,” and “sleeping.” Algorithms translate this facial data into statistical data by generating scores between 0 and 100 that are then used by teachers and administrators to modify instruction or intervene. While this artificial intelligence (AI) system includes human actors by allowing them to use new kinds of data to make decisions in the classroom, it also excludes them by blackboxing the (exclusively machine-readable) algorithms and pattern recognition codes at work here. Does a system like this provide more control to humans in matters of decision-making in education, or does it actually take away control? This paradox of the blurring of the assumed distinctions between human and machine agency lies at the heart of the authors’ key argument in this provocative book about AI-based governance in education.
Discussions about AI in education today could be considered an extension of long-standing conversations about datafication. However, many critical commentaries often border on popular cautionary science fiction tropes of robots replacing humans in the classroom and in policy making. Gulson, Sellar, and Webb contest this human-machine dichotomy and emphasize that AI in education “does not involve direct replacement of human minds and bodies, but rather it produces new ways of thinking through the conjunction of human and nonhuman cognition” (2). While analyses of education policy and governance until now have focused on the agencies, identities, and rationalities of human actors and systems, what might they look like in this new era where humans and machines interact to coconstruct decisions about education? In addressing this question, the authors of Algorithms of Education do not take a clear normative stance; they steer clear of both ominous commentaries about the harms AI will foster as well as a blinded excitement for a “revolutionary future.” Instead, they take a balanced, nuanced approach to explore how new rationalities are emerging in education governance, the possibilities they present, and the things we need to be cautious about.
While education policy has been historically influenced by human actors and rationalities, the current Al-based education governance is creating new policy tools and technologies that both build on and transform the earlier ones. AI has expanded the idea of education governance from just being state controlled to now involving a constellation of actors from edtech firms and major technological companies to data scientists, supercomputers, and cloud technologies. This has produced a “synthetic governance” in education, the core idea in this book, which the authors define as a combination of “human classifications, rationalities, values, and calculative practices”; new forms of computation and nonhuman rationalities; and “the new directions made possible for education governance by algorithms and AI” (4). Gulson, Sellar, and Webb call on readers to understand this new synthetic governance in education as “not human or machine governance, but human and machine governance” (132). This book is an attempt to outline this new regime of governance, illustrate how it is playing out, pose questions about the sufficiency of existing conceptual tools and methods to analyze education governance in examining it, and offer possibilities for the same.
The idea of synthetic governance is conceptualized and strengthened through several recurring themes. One is the paradox of control involved in AI-based governance, where humans are increasing algorithmization, datafication, and digitization in education to gain more control but at the risk of losing control at the same time as more and more of these processes take place behind the black box of machine learning and neural networks. Another theme challenges the commonly considered human-machine dichotomy, as the authors consider their work to be “an experiment in creating new lines of development for thinking about education governance by conceptualizing the cooperation of human and machine cognition in ways that blur both their assumed and actual distinctions” (37). These themes sharpen the overall argument of the book and encourage the reader to reorient their approach to thinking about AI in education.
In service of this argument, the book is organized into three conceptual chapters, three empirical chapters, and a concluding chapter that provokes future work. Chapter 1 elaborates on how the existing “Anglo-governance model” has been based on data-driven rationalities and is beginning to transform toward a synthetic governance model. Chapter 2 conceptualizes this synthetic governance model in depth, proposing alternative theoretical perspectives to understand AI-based governance in education and the new possibililies and aspirations it presents. Following an outline of the authors’ methodological approach in chapter 3, chapters 4–6 focus on different examples of synthetic governance playing out in education systems around the globe. These include the economic and political consequences of the National Schools Interoperability Program’s infrastructure in Australia, the applications of machine learning and pattern making in the facial recognition system in Chinese classrooms, and the workings of a data analytics center in an Australian state department of education. The concluding chapter offers some tentative answers to the critical questions the book raises about possible actions with or about synthetic governance.
The book’s content and structure do not necessarily make teachers or educators—whose daily work might be most impacted by these changing rationalities in education policy—its intended audience. However, it provides much for scholars, policy makers, administrators, and technology professionals to gain from and think about.
The main strength of Algorithms of Education lies in its lack of a normative stance about Al-based governance, which adds significant nuance to the authors’ key argument. By embracing both the possibilities and shortcomings of synthetic governance, the book generates insights that can be used by critics of AI in education to supplement their cautionary stances and by proponents of AI to rethink and improve on their work. Additionally, this book promises to be a foundational text for future analytical studies of AI-centered developments in education policy and governance, since, according to the authors, the emerging synthetic modes of governance will create new values, desires, and expectations and “new conditions for contesting how education policy should be made and for what purposes” (128). Long after reading the book, readers might find themselves still deep in thought, realigning their understanding of education governance, grappling with questions of power and ethics, or even being inspired to find alternative approaches to analyze or conceptualize policies.
Some readers might find the first few chapters to be difficult reading due to their language or content. It might take some time for generalist readers without relevant knowledge to get acclimatized to dense discussions about AI, datafication, governance, and technology in chapters 1–3, especially due to a lack of sufficient examples. However, persisting with these sections and getting to the empirical chapters is worth the effort, as these chapters provide the required illustrative narratives that such readers may look for. The empirical chapters nicely capture aspects of AI-based governance in practice and lead to the conclusion in the last chapter.
In chapter 3, Gulson, Sellar, and Webb provide a brief but interesting example to help readers understand why new synthetic modes of governance require new ways of analyzing and understanding education policy. Building on that example, imagine a situation where the education policy in a state in India is shaped by AI-generated insights based on inputs from a data analytics center in the United Kingdom as well as local stakeholders and compiled using code written in Silicon Valley. In such a case, where is the policy being made and who is making it? What might one study or analyze to understand or critique this policy? These questions don’t have easy answers, and engaging with them requires thinking in new ways and reimagining human-machine interactions in our world. Algorithms of Education provides the right push in that direction.