Tinkering with tech: How the pandemic exposed the flaws of digital education

Schools in the United States have placed a lot of faith into the possibilities of education technology for improving both access and learning outcomes. But the pandemic has revealed that most technology is used to replicate traditional school routines. Justin Reich (MIT) explains why incremental improvement may be preferable to sweeping change and why technology alone cannot fix the U.S. education system.

This article is part of our dossier "Digital classrooms - Transatlantic perspectives on lessons from the pandemic".

Over the past two decades, education technology evangelists have made remarkable claims about how new technologies will transform educational systems. In 2009, the late Harvard Business School professor Clay Christensen and colleagues predicted that by 2019, half of all US secondary school classes would be taught in online and blended formats, that these new courses would cost one third as much to provide and would lead to better learning outcomes for students.[1]

Founders of online education organizations and companies echoed this. In a 2011 TED talk, Sal Khan, founder of Khan Academy, proposed using short videos to reinvent education. Students could solve personalized math problems at their own pace on computers while teachers led small groups for enrichment or remediation. Udacity founder Sebastian Thrun predicted in 2012 that only ten mega-institutions of higher education may exist in 50 years, and his company could be among them. And renowned education researcher Sugata Mitra claimed in 2016 that students didn’t need schools or teachers at all: groups of children with access to the internet could teach themselves anything.[2]

In some respects, these claims suggested that education was ripe for disruption and transformation. New technologies would provide powerful new learning experiences, usher in new organizational forms of classrooms and schools, and lead to better outcomes for students at lower costs.

Then, in 2020, the world was blighted by a pandemic. Schools serving over a billion learners shut down, according to the United Nations (UNESCO).  Primary and secondary schools rapidly switched online, while facing issues like continuing to provide meals for families and shifting their technology stock from schools to homes. Teachers with no experience in remote education began developing new curricula and lessons and finding new approaches in a shifting and uncertain policy landscape, with changing and conflicting guidance from schools, districts, and states.[3]

Contrary to the promise that new technologies would enhance the existing educational offerings, these new forms of online learning had to compete against a hobbled, overburdened system. For most learners and families, remote online learning has been something between a disappointment and a disaster.

As one of my hometown elementary school principals phrased it, “We used to have one physical space that served 285 students; now we are trying to support students in 285 different spaces.”

The two defining learning technologies of the pandemic

To be sure, education technologies have played an important role in the bridge to remote learning during the pandemic, but the most widely used are not new entrants. Neither MOOCs, intelligent tutors, virtual reality, artificial intelligence, nor other new-fangled tools are powering the transition, but rather two of the oldest technologies available: learning management systems and video conferencing.

Learning management systems were theorized in the 1960s and 1970s, commercialized in the 1990s, and made open source in the 2000s. They have become ubiquitous at all levels of education, with systems like SeeSaw and Google Classroom used among younger grades, and Canvas, Schoology and others used in secondary and post-secondary education. They let students and teachers pass documents like worksheets back and forth. They can do other things—host discussion forums, portfolios, peer review, and autograded quizzes—but they primarily help manage the documents of school.

Video telephony, as it was called when introduced in the 1930s, has been the other major learning technology of the pandemic. In some respects, the importance of video conferencing is obvious. In typical classrooms, much of the activity involves teachers talking to groups of students. Video telephony, in theory, allows this interaction to happen at a distance.

But for many instructional designers and other experts in online learning, the surge in this type of synchronous learning during extended school closures has been something of a surprise. Synchronous learning has a number of serious problems. It excludes many students with insufficient bandwidth or inadequate devices and poses a challenge for families with several young children or for students attending classes across time zones. Many simple, essential classroom teaching moves—like looking directly at or being near a student who is losing attention—do not work over video conference. A basic request like “Turn and talk to your neighbor for two minutes” happens instantly in a classroom, but is much more difficult to orchestrate effectively by video conference.

These challenges are substantially magnified for the youngest students. To participate in a video conference meeting, a person must be able to turn on a computer, access a browser, log into a video conference classroom with a URL that includes a random 10-digit code, attend to and follow instructions from the disembodied head of a speaker, toggle video and mute buttons appropriately, and manage this entire interface while paying attention to the content being taught. Most six-year-olds lack the motivation, executive function, working memory, and fine motor skills to accomplish this suite of tasks without the full-time supervision of an adult. Coordinating attention and activity on video conferences is even difficult for adults.

As many students and families are learning, this particular combination of technologies—learning management systems and video conferencing—is not a particularly robust replacement for in-person learning, especially for the youngest students. So why has this combination been so prominent? The answer lies in the conservativism of schools.

Schools as “small-c” conservative systems

Over the last forty years, education researchers have found one reliable pattern in the history of classroom technology: when teachers get access to new technologies, they primarily use them to extend existing practices. Technologies don’t disrupt existing systems, existing systems domesticate new technology. Developing and perfecting new pedagogical practices is time-consuming, and new technologies add complexity to these patterns. When teachers and school systems evaluate new technologies, they usually want to replicate common, traditional routines. For all the shortcomings of learning management systems and video conferencing tools, they let teachers conduct remote learning in ways similar to typical school routines and classroom practices.

To understand the shift to remote learning during the pandemic is to recognize schools’ responses as a profoundly conservative reaction to a global crisis. In secondary schools, many teachers walked from their lecterns to their home-office web-cams, and kept teaching largely as they had before. The adaptations have been harder in primary schools. Still, teachers generally have expected students to sit at their laptops instead of their desks, and to participate in schooling organized around daily periods of math, literacy, the arts and similar subjects.

Whether this readily observable pattern is also desirable is an open question. To education technology enthusiasts, the conservatism of schools is typically perceived as a serious shortcoming: if only schools were willing to reorganize themselves to benefit from powerful applications of new tools! In her anthropological study of the One Laptop per Child program, The Charisma Machine (2020), Morgan Ames calls this the “charismatic” approach towards technology, which puts tremendous faith in new technologies to reorganize social systems.

Another view is that schools are enormously complex institutions with a variety of stakeholders pursuing a diverse set of objectives. Schools teach students to tie shoes, factor polynomials, understand the causes of the Civil War, speak a foreign language, love their country, critique their country, square dance, recognize the pluperfect, get along on the playground, and much more. Schools have found a variety of organizational forms and routines—age-graded classrooms, class schedules, textbooks, marking periods, desk arrangements—that work to balance these competing demands. Under this view, when reformers try to use technologies to sweep away existing schools and replace them with new models, they find that big changes in one part of the system disrupt delicate balances in others.

The tinkerer’s approach to education technology

From this perspective, sweeping change is neither possible nor desirable. The work of improving schools should been seen not as making one big transformative move (like reorganizing schools around new technologies), but as a deliberate steady process of incremental improvement, getting a thousand things a little bit more right each day, each semester, each year. Morgan Ames calls this the “tinkering” approach to school improvement, drawing on the work of David Tyack and Larry Cuban in their 1995 book Tinkering Toward Utopia.

For those hoping that schools might adopt radical new approaches to teaching and learning after their experience in the pandemic, this incremental stance towards continuous improvement will likely mean disappointment. But those who see human development as a long journey, with two steps back for every three steps forward, will understand the value of such small, shoulder-to-the-wheel improvements.

While charismatic technologists see learning technologies as multifaceted and transformative, tinkerers tend to think of learning technologies as a rather uneven set of specific tools, an odd collection of ratchet heads rather than a Swiss Army knife. For instance, over the past 60 years, technology developers have programmed a set of adaptive tutors that attempt to personalize instruction for students. They pose problems to students, evaluate their answers automatically, and then point students to easier or more difficult subsequent material.

These learning technologies work well for some subjects, but not others. The judicious use of technology in schools involves matching effective technologies with the right pedagogical goals. They work for teaching some mathematics and computer programming. For instance, there are decent, gamified, adaptive tutors for subjects in elementary grade mathematics—programs like Dreambox and STMath. At the same time, remote teaching mathematics to younger students through video conferencing is quite challenging; so in this domain there is a nice overlap between a well-developed set of learning technologies and a need in schools.

Evidence suggests these tools are not very effective in teaching reading skills. There are some reasonably effective adaptive tutors for the early phases of language acquisition—where students are learning to say hello and conjugate basic verbs. But we don’t have particularly good learning technologies for teaching phonics. And these tools are much less effective at later stages of language learning, where students are learning about new cultures through reading primary sources, for example. For teaching most topics in social studies and science they hardly exist at all.

Why are there better adaptive tutors for mathematics than for reading, science, and social studies? The answers are both technical and political. In the United States public school system, math is a standardized testing subject in grades 3-8 and in high school, while science and social studies are tested much less, if at all. Publishers develop products for “markets” within schools, and these markets are often formed by testing requirements. Tutors work better in math than in reading because there are better autograding tools for mathematics. Computers are good at identifying correct answers when the complete set of possible right answers can be known in advance, as with computation problems. Some aspects of reading are tractable in the same way (such as identifying silent letters in a word pronunciation) but many are not (such as enumerating the full list of plausible motivations for a character in a fictional story).

So during the school closures, many elementary school teachers have used a combination of synchronous teaching and common printed or online classroom materials. Improving remote instruction in these topics is not about buying the right software, but about working with communities of educators to identify effective curriculum materials and develop new kinds of online teaching routines.

First and foremost it is also about addressing the digital fault lines, which map closely to other structural inequalities in society. The pandemic has both unearthed and exacerbated opportunity gaps that are already a deep moral stain on our society and must be a central focus of future policy changes. Millions of families in the United States do not have adequate access to broadband internet or lack computing devices for each of their children. Many lack the resources to provide safe, quiet spaces for learning and tutors for additional support. Students who have been poorly served by traditional education settings often experience a greater “online penalty” than others.[4]

Schools Cannot Fix Society

Perhaps the most important lesson of the pandemic is that school systems cannot be the sole infrastructure provider and safety net for children and their families. Teachers and educators have done an extraordinary job of managing a challenging pivot to remote learning, while providing meals and technology to children and families, retrofitting buildings for improved ventilation, offering frontline health screening, addressing special education and mental health needs, and filling gaps in the social safety net around shelter, clothing, and other necessities.

Overburdened schools cannot do all of this. Welfare systems should provide adequate funding for food and shelter for children and their families. Public health systems should provide screening, testing, tracing, and isolation facilities for children and their families.  More robust paid time off and family leave policies should ensure that families whose loved ones fall ill don’t have to choose between work, school, and risking infection. A federal government initiative to provide broadband access, like the early electrification movement in the 20th century, should connect every home and residence.

Dramatic improvements in learning will not come from the invention of new technologies, they will come from a commitment to strengthen the social fabric around children. Once these fundamental issues are adressed there is an important role for EdTech in the future of schooling, especially in view of how future calamities such as extreme weather events or pandemics will affect how and where students learn. But our learning technologies are only as powerful and effective as the communities that guide their use.

This article is part of our dossier "Digital classrooms - Transatlantic perspectives on lessons from the pandemic".



Ames, M (2019). The charisma machine: The life, death and Legacy of One Laptop Per Child. Cambridge: MIT Press.

Christensen, C., Johnson, C., Horn, M. (2008). Disrupting class. McGraw-Hill.

Dynarski, S. (2018). Online courses are harming the students who need the most help. New York Times, Jan. 18, 2018. https://www.nytimes.com/2018/01/19/business/online-courses-are-harming-the-students-who-need-the-most-help.html

Kaupp, R. (2012). Online penalty: The impact of online instruction on the Latino-White achievement gap. Journal of Applied Research in the Community College19(2), 3-11.

Khan, S. (2011). Let’s use video to reinvent education. Retrieved from https://www.youtube.com/watch?v=nTFEUsudhfs

Mitra, S. (2016). The future of learning. Proceedings of the 2016 ACM Learning at Scale Conference. https://dl.acm.org/doi/proceedings/10.1145/2876034

Reich, J., Buttimer, C., Coleman, D., Colwell, R., Faruqi, F., Larke, L. (2020). What's lost, What's left, What's next: Lessons learned from the lived experiences of teachers during the 2020 novel coronavirus pandemic. https://edarxiv.org/8exp9 

Tyack, D. & Cuban, L. (1995). Tinkering toward utopia. Cambridge, MA: Harvard University Press.

UNESCO. (2020). Global monitoring of school closures caused by COVID-19.  https://en.unesco.org/covid19/educationresponse


[1] Christensen, C., Johnson, C., Horn, M. (2008). Disrupting class. McGraw-Hill, 101.

[2] Many of the arguments in this report are abridged from Reich, J. (2020). Failure to disrupt: Why technology alone can’t transform education. Cambridge: Harvard UP.

[3] See Reich, J., Buttimer, C., Coleman, D., Colwell, R., Faruqi, F., Larke, L. (2020). What's lost, What's left, What's next: Lessons learned from the lived experiences of teachers during the 2020 novel coronavirus pandemic.

[4]Kaupp, R. (2012). Online penalty: The impact of online instruction on the Latino-White achievement gap. Journal of Applied Research in the Community College19(2), 3-11.