Learning Isn’t an Algorithm

Justin Reich knows it is strange to write about technology’s power to transform education during a pandemic that has demanded even the most analog-minded teachers to utilize it. “The best possible future,” he contends in the prologue to Failure to Disrupt: Why Technology Alone Can’t Transform Education, “will be one where we recognize the incredible importance of our formal education systems to the social order, and we provide these systems with adequate funding, support, and respect. Our learning technologies are only as strong as the communities of educators who guide their use.” Reich’s insight neatly encapsulates the book’s argument that technological innovations alone, despite their promise, aren’t enough to change the world. 

In this eerily timed book, Reich, a professor of comparative media studies at the Massachusetts Institute of Technology, argues that learning-at-scale educational technologies will never measure up to the hyperbolic claims of their proponents. Innovations such as massive open online courses (MOOCs) and adaptive courseware haven’t validated predictions of their ability to transform education or closed achievement gaps. What’s needed instead, Reich contends, is slow and steady incremental change, with deliberate attention given to making education more equitable.

Systems change is needed, Reich argues, because educational systems are conservative. The basic structure of education, in which a single teacher works with a room full of students, has remained unchanged for centuries. Reich explains that because people teach the way they were taught, learning-at-scale technologies like MOOCs and adaptive courseware have not radically changed education. That is, pedagogy does not keep pace with technology. Instead, the strongly traditional educational system domesticates innovative tech, bending it to fit standard teaching and learning approaches, albeit in a digital format.

For example, MOOC enthusiasts once promised to offer a low-cost educational opportunity to anyone in the world, but in their current form, Reich explains, MOOCs “are now primarily supplements to existing infrastructure for professional master’s degrees and executive education programs.” The flash-card app Quizlet is another example that Reich provides to demonstrate learning technologies’ limited potential to reform existing education systems. Quizlet, he asserts, is popular and effective, but not transformative; it simply digitizes an already-proven method.

Reich identifies three existing categories of large-scale learning technologies, or online learning environments, that can reach many students with only a few experts—instructor guided, algorithm guided, and peer guided—to frame his analysis. He describes these in relation to the two leading philosophies of education: the pail-filling instructionist approach, in which students are empty vessels into which teachers pour knowledge and information; and the fire-lighting constructivist approach, in which teachers ignite students’ interest and curiosity by creating conditions in which students learn by doing.

There’s a limit to tech’s ability to facilitate learning, especially as it pertains to critical thinking, complex communication, and unstructured problem-solving.

MOOCs are a good example of instructor-guided, pail-filling learning technology. A course is developed based on what the faculty member or subject matter expert determines is the best sequence of material to facilitate learning. The path is linear, preset, and universal for all learners.

Another pail-filling technology, algorithm-guided learning environments respond to input from each individual learner. Every student experiences a custom path through the material based on their accumulated interactions with the system, which reveals their strengths and weaknesses in the subject matter. Students who need additional support are directed to review materials and exercises for remediation, while those who demonstrate mastery accelerate to new or more challenging concepts. In this way, everyone’s unique learning needs are met.

Peer-guided learning environments, aligned with the fire-lighting philosophy, are best suited for informal learning. Reich describes the Rainbow Loom craze of 2013, in which a global network of online users posted YouTube videos to show others how to create increasingly complex bracelets and other designs using brightly colored rubber bands. In this example, people teach and learn from each other with no formal system in place. Although these environments function well in social interactions, they are too unstructured to be effective in traditional school settings, which are designed to measure individual learning, not collaboration. Reich argues that apart from applications such as extracurricular coding clubs, peer-guided environments seem to have no place in schools.

Reich also notes the recent development of a fourth category, game-based learning or gamification, which spans both the pail-filling and fire-lighting philosophies. Gamification can be either instructor led, such as games that are simply math practice problems disguised as fun adventures, or peer led, as in Minecraft, for which you can learn tips and tricks for progressing in the game from people posting online. As with MOOCs and adaptive courseware, gamification has failed to radically transform education. While learners are taught to play the game, Reich questions their ability to “flexibly deploy” gaming skills to other paradigms, including everyday life.

In the second half of the book, Reich outlines the complex problems that limit the ability of any large-scale learning environment to reduce inequities. He discusses the “curse of the familiar,” in which innovative technologies are often diluted to fit existing educational systems. He then examines how the sociologist Robert Merton’s Matthew effect (“Success breeds more success”) has played out in education technology; the “edtech Matthew Effect,” Reich explains, occurs when those who have access to technology and privilege benefit more from learning-at-scale technologies than people from less advantaged backgrounds. The “trap of routine assessment” describes how computerized assessments are only good at evaluating the things computers do well, while the jobs of tomorrow will require people who are able to do the messy things we can’t automate. The final dilemma, the “toxic power of data and experiments,” addresses the complicated ethical concerns around collecting students’ data without explicit permission.

Based on this critical assessment, Reich recommends a long-term commitment to designing and deploying these technologies via collaborations among a diverse mix of students, parents, teachers, and researchers. Studying the implementation of these technologies, and making steady, small improvements over time—what Reich calls “tinkering”—will yield better results than the typical hyped-up and short-lived claims about the ability of at-scale technologies to transform education. Finally, commitment to systems change is required, investing in professional development-focused communities of educators, improvements to local infrastructure, and educational programs to train parents to help their students use tech more effectively. In short, deploying at-scale learning technology in schools and colleges is not a silver bullet. Significant wraparound support of both people and systems is required for these solutions to make a difference.

While Reich’s analysis and recommendations are grounded in extensive experience, robust research, and factual historical information, there’s something missing—not just in this analysis, but in learning-at-scale technologies as a whole.

That missing piece is teachers. Large-scale technologies fail because they try to replace teachers with computer programming, which is incapable of the pedagogical expertise needed to facilitate learning. Teachers bring both subject matter proficiency and pedagogical skills, and they adapt this knowledge and expertise to best meet the needs of the learners in front of them—because the individual learners also shape the cohort experience, as well as how that teacher teaches to that specific classroom. The Technological, Pedagogical, and Content Knowledge (TPACK) framework, an approach that addresses problems in implementing tech in the classroom, explains why large-scale technological learning environments will never be effective. Teachers need all three kinds of knowledge, all of which factor differently depending on the educational context. Only people can assess what is needed in the moment. It’s not programmable. Students aren’t robots. Technology can’t teach the way that people—with their pedagogical training, wisdom gained through experience, and professional expertise to adjust their tactics as needed—can.

Reich’s analysis neglects the value of teachers to help students learn. Students from historically marginalized groups or low-income backgrounds don’t need impersonal technology to help them learn and overcome the challenges they face. They need human connection, mentorship, and inspiration.

To understand why, it’s important to revisit Reich’s argument about the trap of routine assessment—that is, what computers can’t do. He notes that the reach of automated assessments is limited. Computers can’t solve unstructured problems or communicate above a certain level of complexity. Reich includes as an example the continued need of airline employees to help check in travelers at airline counters. Computer kiosks can handle the routine check-in experience, but once the process requires more than basic communication or service, a staff member must resolve the issue.

The same is true for education. While certain aspects of content delivery and assessment can be programmed, there’s a limit to tech’s ability to facilitate learning, especially as it pertains to critical thinking, complex communication, and unstructured problem-solving.

Reich does not infer that these nonautomated skills are precisely those that teachers are capable of providing in the human interaction, engagement, and encouragement that students need to learn how to navigate, negotiate, and think critically. No technology can provide this level of complex guidance.

To be fair, Reich acknowledges that “most learners require support and human contact,” and calls for better preparation of educators to meaningfully implement large-scale learning technologies. He argues for more robust professional development, especially for schools in low-income neighborhoods, as well as sustained professional learning communities. But he fails to call out the inability of computers to teach—to communicate with complexity and solve the messy, human problems.

Reich proposes systemic changes that are big enough to drive a major culture shift in education. But he does not provide practical steps forward for individuals committed to the educational enterprise. The lack of practical steps, combined with the oversight regarding the centrality of teachers, leaves a gap between theory and action. Despite asserting this point in his prologue, he doesn’t reinforce it in the book. Computers can’t teach. Only people can. Consequently, readers are left wondering what they can do to address the insufficiencies of remote learning technologies—all too crucial during the seemingly interminable COVID-19 pandemic.