This piece was written for The White House Symposium on the Future of Education R&D and Digital Learning, which took place on Oct. 5. It was originally published on Bror Saxberg’s blog and is reprinted with permission.
Technology is changing the way almost all industries work beyond recognition. It’s also changing the nature of what skills humans need to succeed. This has profound implications for our education and training systems—we will need to be much more systematic and efficient to help many more learners succeed in the long haul.
To do this means more systematic application of learning science and the evidence regarding both learning at scale and on an individual level, within the always-present constraints of regulation, economics, usability, and more. This means we need to think about learning engineering thinking—practical use of science at scale—and we’ll need more learning engineers who are prepared to help us.
The types of decisions and tasks we all need to address are changing rapidly. Take, for example:
- We are shifting from using motor skills to add value to the need for complex cognitive skills for almost all work. Auto repair, farming, construction, manufacturing—all of these now have complex cognitive components that workers have to master to stay on their game.
- Technology itself is changing rapidly, and so too are the accompanying necessary skills. Our learning environments have to evolve continuously. A stonemason in the 12th century could confidently pass his skills (physical and cognitive) to his children and grandchildren, knowing the workers constructing a cathedral would be using the same techniques for multiple generations of work. Now, robotic manufacturing systems are ever-shifting relationships between tools, people, tasks, and outcomes, and skills must rapidly evolve to keep up.
- More tasks and decisions are moving to where the expertise is, and where costs are lower. This means a job may not be available near you, because it can only be done “on the ground” in your locale— so you need to be ready to pick up new skills.
- Just as work is changing faster, we are living longer. Few of us can afford to retire after our initial skill set becomes obsolete. We need to upgrade our skills, and perhaps even reinvent ourselves every decade or less in this new world of rapidly changing execution of tasks.
All of this means our education and training systems need to be different than they have been. They were never designed to maximize the yield of trained minds (from whatever starting point) making complex decisions in rapidly evolving areas they are passionate about. There is a wide range of inefficiencies that need to be addressed.
Cognitive science tells us that expert minds are not very good at verbally describing all the expertise that’s been learned over time—in fact, when training novices, experts typically articulate less than 30 percent of their decision-making processes and what they do. Thus learning environments often, at best, target a fraction of what’s needed to succeed at work, and, on average, often end up targeting irrelevant information and skills that may have been relevant decades ago (e.g., slide rules for me, when calculators were becoming available.).
In addition, learning environments have typically assumed that failure is the learner’s fault. For instance, if you are not able to absorb “what you need” from the perorations of an “eggsbert,” then you might not be cut out for the work you'd love to do. Attrition ensues.
Learners then stumble into their first job with a fraction of what they need to be good at their current tasks. A large number make mistakes and feel badly about their career choice. Many drop out, assuming their inability to learn is their fault, that they just “aren’t right” for that career. A fraction either accidentally discover good coaches and get the right feedback to master the rest of what they need, or, on their own, grit their teeth and discover (or even create) what is necessary for mastery, and happen to not to be the ones blown up by situations thrown at them outside their newly-developing control.
What are the implications for our learning and training environments?
They need to target what the best practitioners decide and do now, not decades ago, as well as where those industries are going. This means systematic and repeated identification and unpacking of what top performers (not just popular people) decide and do on tasks.
They need to be designed to match how human learning actually works, not how we wish learning worked. For millennia, medical treatments were based on how we wished or hoped human bodies worked, and they helped a bit (and sometimes not). However, we have had rapid improvements in healthcare after a multi-decade exploration of how bodies actually work—and fail—and that information has been used to relentlessly (re)design treatment, sometimes in the face of early popular apprehension and even opposition.
We need them to be designed as sequential and personalized paths on a journey to gain expertise over many years. As with health care, we need to be better at understanding the longitudinal implications of activities and treatments as we all live longer, as well as personal variations in response to activities, not just population variations as we've done historically.
Technology has to be applied to deliver better learning, not to simply be “cool.” Once we identify what is a better learning activity for a specific outcome (e.g., more personalized feedback, faster markup of performance practice, media and text working together, not at cross-purposes), we can look to technology to make that activity more available, reliable, affordable, data-rich, and personalized. (The opposite of this would be, say, creating a well-produced video of your worst-ever college professor and making it available 24/7 globally, thus damaging the interests of millions of learners, and making the professor a weapon of mass destruction... )
The evidence gathered within learning environments has to improve in both quality and density. Historically, when it didn’t matter that a low fraction of learners reached mastery of complex cognitive content, the evaluation of how learners were doing didn’t matter much either. Now that mastery matters much more economically and socially, we need accurate and frequent indications of where each learner is and where the issues are so we can intervene sooner, before learners are far off their path to mastery. This means our evidence gathering (likely less and less about assessments, and more and more about other data collected about practice, feedback, and interactions) has to become more frequent, and more accurately tied to what experts decide and do (i.e., valid and reliable).
Learning environments and the evidence from them need to be more systematically designed for long-term success. We should be connecting, with evidence, what top performers are deciding and doing both now and in the past, from down-payments in college back to high-school and before. There has to be awareness, too, that early learning should allow learners to “skate where the puck will be,” as hard as that is to predict. It should prepare students based on promising research regarding the details of expert decisions and tasks over the next decade and more, not based on where careers are now.
We also should not forget Dana Gioia’s formulation of the purpose of education: to build “productive citizens for a free society.” Much of the above relates to redesigning learning to increase the odds that a learner will be productive in a career of their choosing. We also need to think deeply about the decisions and tasks of people as citizens: how will they be able to make well-founded decisions on cryptographic privacy vs security, gene-line modifications for health versus family advantage, complex evaluations of leaders around their mastery of economics, politics, the lessons of history? We will disagree deeply on all this at the end of the day, of course, as all democracies should—but we should be disagreeing more around well-grounded arguments, not pure personality points.
We have much—although not enough—learning science available to help shape this, with more on the way. We have many caring people working with learners and investing substantial time and effort across decades, now, more than ever in our history, in building skills (albeit inefficiently). We are beginning to see technology entering the fray, providing new capabilities and data for learning.
So what are we missing?
We need learning engineers. By this phrase (first used, as far as I am aware, in the 1960's by Herbert Simon, the computer-scientist and Nobel-prize winning economist), we mean people who are deliberately trained and focused on designing and systematically improving learning environments at scale in measurable ways. They make use of the current and new science of how learning and motivation work, and they do collect careful measurements, but the focus is on improving success and impact at scale, within constraints (economic, regulatory, practical), not research per se.
If we are designing a new chemical factory, we very likely don't want chemists designing that plant: they're neither experienced nor interested in regulatory, safety, or economic issues, nor do they possess the mix of mechanical and other skills needed to do the job. That's why there's a demand for a large group of chemical engineers: approximately 30 thousand of them currently (rounding to the nearest thousand or so) work in the US.
There are approximately zero thousand true learning engineers working in the US, rounding to the nearest thousand, who are trained and following learning science, and also working at scale within real-world constraints to design, build, and measurably iterate based on outcomes. The lack of folks like this will hold back the entire enterprise of implementing more efficient, effective, higher-yield learning environments: we will miss targeting learning efforts on what experts actually decide and do (versus what they merely say they decide and do); we will continue to include inefficient methods for learning that cause many to fail unnecessarily; we will use technology in more arbitrary rather than targeted ways (which, in failing to produce intended outcomes while being “cool”, will cause investment to continue to cycle from boom to bust); we will not generate valid and reliable evidence that we could use to target interventions early and effectively (and, indeed, we may drown in bad data that we “should” be using).
We can do better than this. We can begin to train more people on learning engineering fundamentals, to improve their own decision-making as teachers, teacher-trainers (teachers have minds too, so this is its own learning engineering challenge), purchasing decision-makers, publishers, edtech developers, venture and other funders, philanthropists, policymakers and more. We can begin laying out more clearly what learning environments would look like if they had been well-designed as learning engineered environments, and hold folks increasingly accountable to reach that standard. We can become more alert to the quality and use of learning and learning interventions data, so that we are aware of what evidence is “good enough” to make real decisions about learning environments, either at scale or for individual students.
At the end of this new day, millions more learners will discover, at all ages, highly effective routes to achieve mastery of valuable skills in areas they are passionate about. It will be clear that there is real work involved in getting there, but the paths, sometimes over many years, will be highly likely to lead to the outcomes they really desire, giving them fuel for the journey ahead. We can adapt and benefit, generation by generation, from the rapid changes around us, rather than be left behind.