The Learning We Design For

The concept of “self-efficacy” keeps popping into my brain lately. I understand it to be your belief in your ability to execute a course of action to accomplish a goal. A more popularized and related term of late is “agency.” Learner agency is commonly desired among teachers. We want students to be motivated to learn, to have clarity of what they know and don’t know, and to have the ability to pursue more clarity around the things they don’t. Agency is also often used with “voice” and “choice”—the idea that students need to be represented in their learning and be able to make choices about what and how they learn.

These are all admirable aims, and then we lay down the hammer: We want to develop learner agency, voice, and choice within the constraints of a standards-based accountability system. Are we developing self-efficacy to support their ability to learn and be successful adults, or are we developing self-efficacy to support them in performing well on an end-of-year assessment?

This tension is heavy. This tension wears down teachers and students. This tension is only whispered about in schools.

In ”America’s Children Are Unwell. Are Schools Part of the Problem?” Jia Lynn Yang suggests that the intense spike in children learning disability and mental health diagnoses should be attributed to traditional schooling models as much as, if not more than, genetics, tech devices, medications and processed foods. She points to standards, testing, accountability, and packed schedules and lack of free time and play as major issues with how we “do school” these days. She also suggests that an increase in home learning is the result of dissatisfaction with schools. The pandemic contributed to this. Parents were able to get a front row seat to their children’s learning when they were learning at home, and many did not like what they saw. Emily Hanford’s Sold a Story podcast opens with these stories. She also suggests that once children were learning at home, many parents finally noticed that their children couldn’t read. Parents were appalled when they realized not only were their children not being taught how to read, but also that they were being taught in ways that harmed their reading progress.

Schools tend to not hold up to much scrutiny when we look at what is being learned, which, as Yang suggests, has eroded trust in the system. And what’s even more baffling is that since No Child Left Behind in the early 2000s, schools have been held more accountable than ever before even with basically stagnant or declining results (e.g., “A New Nation's Report Card Shows Drops in Science, Math and Reading Scores” and “NAEP Scores for Class of 2024 Show Major Declines, with Fewer Students College Ready”). So how have we built an educational system that is so heavily accountable to learning and yet anyone who looks closely believes they see less learning happening now than before?

I believe there is a fundamental mismatch between our goals and our design.

Our society has changed and continues to change and yet our schools are bulwarks of tradition. No one seems to agree how to change them. It also seems too risky. It’s about children’s lives. And so we don’t move forward.

While there are innovative school models focused on different interests or workforce preparation and even schools that use AI run apps in place of teachers, those are novelties. In large part, they still have to meet standards, and students still have to take standardized assessments. It’s the water we swim in even as we build a super fancy pool float with all the bells and whistles.

The changes we need are less about approach or model and more about the construct of learning. We need to be swimming in different water.

Learning is a change in thinking or behavior. Learning requires time, flexibility, and uncertainty. How we learn math is different from how we learn to analyze a text. Some knowledge and skills need to be taught directly and some learning happens indirectly through repeated and/or varied experiences. Some knowledge and skills are evolutionarily “switched on” through exposure. Think language development: Babies babble to try out different sounds but largely, over time, stop using sounds not heard in their environment. That then eventually results in production of words, which then get attached to meaning in our brains. This is a combination of exposure, observation, and experiences. Parents might point to the family dog and say “puppy,” and then do this several times at random when the opportunity presents itself or refer to the puppy while their young child is listening and observing. The child might eventually say some version of “puppy” or come up with their own word for the dog. These experiences over time reflect a teaching and learning process that isn’t controlled. We intuitively and almost unconsciously assess whether the child seems to be getting it, although, we are typically very patient to let the learning develop on the child’s timeframe.

When it comes to school subjects, though, there is an expectation of standardized learning on a specific timeframe with formal assessments. There’s an assumption that all content must be broken down into discrete knowledge and skills and directly taught and assessed. If kids learn something, the idea is that it must have been taught well by a teacher. But students are learning a ton of things in school, many of which aren’t about academics and not being taught directly at all. What do we want our children to learn? What kind of experiences do we want them to have? What kind of adult life do we want them to live? This is where no one seems to agree, and so we all just settle back into convention and how we’ve always done it. At least there’s agreement in that approach, even as it doesn’t seem to be working for us nor producing the learning we want.

This leads me to consider the proliferation of learning apps and the incorporation of them in classrooms. Educational technology has promised more “personalized” or individualized instruction and with it, the promise of more learner agency, voice, and choice. However, two somewhat recent Substack posts about learning applications (“Why Education Can Never be Fun” by Daisy Christodoulou and “Why Most Education Apps Fail” by Carl Hendrick) suggest that what hangs up learning apps is the learning. What makes an app or game fun or “engaging” is fundamentally antithetical to learning. If app designers are able to rid themselves of the albatross of learning hanging around their necks, then they can design an app that engages kids and keeps them coming back for more. But, the moment that learning is expected, those engagement design patterns fail or at least fall flat compared to those without a learning expectation. Since our society is now conditioned to seek engagement, or at least as long as there are apps that focus solely on engagement, children tend to prefer those.

What’s interesting and nuanced in the Hendrick post is the idea that he suggests that learning happens in very predictable ways, able to be broken down and codified or marked. While true for some learning, this is not the case for complex content and ideas as Hendrick ironically himself describes in another recent post, “The Lethal Mutation of Retrieval Practice.”

To me, this begs the question: “What learning are we seeking, and how are we designing our systems to support that learning?”

Looking at how learning apps are designed can help expose this mismatch between learning goals and design. We need to consider what learning can be taught via apps and what still needs to be taught by humans and in a social context. We also can look at how AI technologies affect the answers to this question, if at all.

At Learning Tapestry, we designed Wonderwood, an app that makes use of AI technologies, to try to honor the ideas of learner agency, voice, and choice. The app gives children access to a safe space to pursue their curiosities about topics. They are exposed to new and connected knowledge and ideas that build over time. But we hit a wall. We weren’t going to sacrifice learning for pure engagement, which meant that our design didn’t focus on keeping kids on app and increasing time used. We were relying on parents to see the value and carve out time in the day for their children to use Wonderwood. But getting attention and finding time are insanely difficult these days, and so we consistently lost out. Not to mention the backlash against screentime, which also meant time spent on a device was even more limited.

We also envisioned Wonderwood being used as a supplement to a classroom library (or, if no classroom library existed, it could be the classroom library), a place for students to extend their learning about the topics in the curriculum. But there is an expectation for learning apps to follow the model of school: discrete knowledge and skills that are identified, explicitly taught, and then tested. There isn’t space for nor priority given to learning that is complex and nebulously defined. That kind of learning feels inefficient and hard to measure, so it is downplayed and even avoided. What’s hopeful and promising with AI technologies, though, is the idea that learning apps might be able to handle complex learning, especially as it is used alongside and to enhance human instruction (e.g., “AI Tutors with a Little Human Help Offer Reliable Instruction Study Finds” by Greg Toppo in The74). Whereas edtech products have traditionally been limited to explicitly teaching discrete knowledge and skills, the abilities of AI technologies to reason, engage in conversation, and offer critique or different perspectives and approaches, opens up opportunities to teach and learn complex ideas and concepts in support of what a human teacher provides.

So we have schools that focus on teaching what can be measured, while also claiming to value learner agency, voice, and choice. We want students to develop self-efficacy but then deny them ownership and control over their learning. From a design perspective, this is the same tension that shows up in learning apps: engagement and learning are often at odds and in tension with each other. Even so, teachers absorb a multitude of incongruencies daily and attempt to create a cohesive learning experience out of them. It’s no wonder you hear often of teachers who “close their door and just teach” and ignore what’s happening around them.

The question isn’t whether schools are broken. They aren’t. They work. They produce a particular kind of learning that is measurable and comparable. The question is whether that’s the learning we want, and then how do we design schools that honor the time, flexibility, and uncertainty that learning requires?

Until we design systems that actually support the full range of learning we value, self-efficacy, learner agency, voice, and choice are just lovely, unrealized ideas. We keep asking teachers and students to overcome the system’s design flaws with individual effort. They try, but the system keeps producing what it was designed to produce.