The Importance of Redundancy in Education

Because schools cannot provide all the different environments that different children might need, effective education requires a coordinated range of redundant learning opportunities in and out of school. We can learn how to provide the needed redundancy efficiently. Click “The Importance of Redundancy in Education” above to read more.

Schools in the U.S. evolved over the same time line as factories.  Like factories, schools were to be “scientifically managed” so that they provided education as efficiently as possible.  A single teacher would teach groups of children of the same age, with a different teacher for each grade.  A curriculum would be set, and ideally the managers of school systems would know just what is being taught on a given day to all the children at a given grade level. School, on a different time scale, was to look like an assembly line, with the same “treatment” for each student, just as a car being built in a factory would get a particular weld made at the same station as the previous car coming down the line. Just as engineers keep looking for ways to improve the car assembly process, education researchers keep searching for improved teaching methods that work for every child more efficiently.

Of course, not every car is assembled perfectly.  Even today, the average car has things that dealers need to fix.  But, the process has become more reliable.  For example, Consumer Reports says that from 2011 to 2018, the industry went from 4% of cars needing warranty repairs on brakes to less than 1%.  School systems do not offer “warranty repairs” (actually, a few do), but a look at any newspaper report of standardized test scores for schools would suggest that in some areas, if schools had to remediate inadequate learning by students, the repair rate would be more like 60-80% than like 1-4%.

Making cars is easier to do reliably.  Once an assembly line is in decent shape, the robot screwing in a particular screw can count on finding the screw hole in exactly the same place on every car coming down the line. In the case of students, each student is a bit different, so a single process does not work for everyone.  New knowledge is built on top of old knowledge, so what a student already knows determines what they will learn from a particular experience. Teachers need to be able to quickly determine what relevant knowledge a child already has and then to adapt their teaching to assure that it affords the chance to build on that prior knowledge.

A uniform factory approach to educating each child will not work. We use the term “pedagogical content knowledge” to refer to what a teacher knows about how different children come to a particular subject with different prior knowledge and about how to build on each child’s prior understanding. A good teacher will have broad pedagogical content knowledge that allows them to serve a wide range of students. In addition, teachers need to detect and respond to other aspects of a child’s situation that affect willingness to persist in learning and willingness to engage in various activities that might promote learning.

But, even when a teacher is well prepared and hardworking, school does not work well for every child. The process of making an automobile assembly line reliable and efficient is much simpler than the process of making a school reliable in assuring that every child learns. In fact, rather than looking at the assembly line as the model for schooling, student variability is the core issue an education system must address.

When a process needs to be able to adapt to differing and not entirely predictable circumstances, we generally engineer redundancy into the process to prevent failures.  The recent discussions of failures of the Boeing 737 MAX 8 aircraft are an example of this.  The approach to assuring that a bad situation does not arise will require a number of steps.  These include increasing the redundancy of pitch sensors, making it easier for pilots to determine that the automated systems are pushing the plane’s nose down too much, training pilots to deal with automated system failures, possibly improving the reliability of manual control of aircraft trim surfaces, and improving displays that might guide pilot activity with respect to aircraft pitch.

We might benefit from looking at the problem of educating every child in the same way we look at aircraft reliability.  Just as engineers and pilots are calling for triple redundancy to prevent plane crashes due to errors in pitch control, we might consider having more redundancy of learning opportunities in education. What would it mean to take this path?

Charters are not enough. Currently, the people calling for more charter schools and for school vouchers are responding to the need for redundancy.  The idea is that if there are many different kinds of schools, each child will likely find a school that works for them. However, the range of flight situations is much more constrained than the range of variation in children’s knowledge and dispositions, and the minimalist approach to redundancy has not worked well even in aviation, as we have learned from the MAX 8. Major repairs are now underway to deal with that problem.

So, what are the problems with the current charter-school approach to educational redundancy?  First, most charters are not that different from public schools.  In fact, the charter process in many states makes it hard to be very different.  The state cares how many days the school is in session, how many hours per day, what the ratio of teachers to students is, and what material is covered at each age level. In the end, attending one charter school offers no more redundancy of learning opportunities than attending one public school. In addition, even with the variability that does exist, it is not clear that either parents or children are particularly adept at deciding what school to attend.  Factors like geographical proximity to the home, alignment with after-school childcare, alignment with family religion, and even the appearance of the school building are likely to be bigger determinants of school choice than which experiences are provided for learning algebra or what kind of mentoring is provided to improve student writing.

So far, I have focused mainly on the subject-matter curriculum of schools.  However, there is more education that every child needs.  In my book, Learning for the Age of Artificial Intelligence, I list eight competences that every child should acquire to have a decent shot at a good life in the age of smart machines (see this blog). These pose a bigger challenge for schools, whether public or private or charter. These competences often were not even taught when current teachers were in school, but they now are critical.  That is, they are critical for almost everyone except public school teachers and university professors who have tenure.  The rest of us need to be prepared to have our jobs automated away.  This means we need the human skills less likely to be automated and also must be competent enough to survive a period of retraining during which there may be no income.

To be blunt, all our kids need these competences now, few people acquired all of them in the past, and teachers are better insulated from needing them than their students will be.  This suggests that children likely will need learning opportunities not likely to be provided reliably by traditional schools.  Schools might do the job for some students, but others will fall through the cracks.  This might be because the school day is not organized in ways that allow them time to develop the competences fully.  It might be because their teachers cannot teach what they themselves have not mastered. It might be because different children need different learning opportunities and it just is not possible for one teacher to provide the variety of experiences needed by 20-30 children. No school is 100% reliable in providing the competences needed to thrive in the AI age!

In other aspects of life, when no system can perform with adequate reliability, we move to a redundant set of systems that cover for each other. I would never fly a plane that did not have redundant systems to assure the pilot knows whether the plane is climbing or dropping (it turns out that it often is hard to tell by just looking out the window). For critical medical matters, such as major surgery, we have redundant systems to monitor the patient and even for details like assuring that sponges are not left in the body.  We need this approach for education, too.

It is time to see our education system as including schools, after-school programs, outside organizations, and any other possibilities that seem likely to be productive.  Not all of these opportunities need to be publicly funded.  However, we will need ways to assure that the goal of achieving competence for the AI age is achievable for every student.

We can learn from experiences with health care systems. We have a similar redundant approach in the health care world.  Some of that involves medical specialization – we see cardiologists for heart problems and dermatologists for skin problems.  However, there is additional redundancy, with some people seeing chiropractors, some seeing exercise coaches, some taking yoga classes, some getting nutritional help, etc.  Different people see different combinations of care providers, driven partly by their dispositions and partly by their medical needs.

With health care, we generally operate by having a primary care provider for each person.  Historically, these have been family medicine doctors, internists, and pediatricians, mostly.  In the future, nurse practitioners and physician assistants may play a bigger role, especially if we get serious about providing health care for all.  In addition, health care has started to be integrated for each patient using electronic medical records.  So, for example, when I had a digestive problem, the gastroenterologist was able to see a lot of information about me that was collected by an internist, as well as information collected by other specialists.  This allowed diagnosis with less redundancy of medical tests and helped assure that I would not be subject to treatment options for my gut that might be damaging to other body systems. While I happen to use an HMO that has a stake in integrating care, we can expect over time that health insurers will begin to support electronic medical records for their clients who are free to use a range of independent providers.

An integrated learning record. The key to assuring redundant learning opportunities that work for every child in providing the needed competences for life in the AI age will be such a shared record, an electronic learning record. A child’s classroom teacher may provide some of the integration, just as a primary physician provides some integration of my health care.  However, real coordination will not be possible unless learning outcomes data and data on learning approaches that already have been tried are collected and easily available to whoever helps a student consider new learning opportunities.

This will not be easy, but it will be possible.  We will need to change privacy laws to keep protecting students but not be barriers to coordinated efforts for each student’s formation. We will need to find ways to enable learning opportunity providers to speak some common language when describing what they provided each student and what that student became able to do. We will need to support and data mine a database in which learning opportunities and learning outcomes are stored, so we can begin to understand what kinds of activities can best supplement school for each child.

In a future blog, I will provide a science-fictiony sketch of what this all might look like. As with medicine, achieving a redundant but coordinated education system that serves every child will require overcoming pitfalls.  There also likely will be the occasional pathology.  However, our intolerance for risk should be tempered by what we all know: our schools currently are not preparing many children adequately for life in the AI age, and the odds of any child being adequately educated are determined substantially by wealth and racial/ethnic status. I believe it is worth risking some smaller problems in order to solve the big problem that we fail too many of our children and that the cost of educational failure is rising in the age of smart machines.

Leave a Reply