In the first article in this series, we looked at the argument for why we need a design-oriented, systems engineering approach to help us ultimately arrive at coherent and effective education systems for our modern society. In this, the second article of the series, we’ll unpack the system we have evolved into, in order to understand what’s not working at the systems level before looking at how we might build effective learning systems going forward.
A fundamental principle of system dynamics states that the structure of the system gives rise to its behavior (Sterman, 2001). And as such, if we are not seeing the behaviors we desire, we must look to redesign the fundamental structures of the system. Leveraging systems engineering as the fundamental approach to reengineer key organizing structures creates the opportunity to truly design and implement meaningful learning at scale. The framing is notable, because the emphasis is not on how to fix current education systems necessarily, but rather to zoom out more broadly, see the system as the engineered (purposefully or not) infrastructure that it is, and seek to purposefully engineer ways in which we can more effectively create meaningful and effective learning environments at scale.
A Structural View of the System We Have
Although education systems are complex and vary considerably, it is worth zooming out to take an ‘archetypal view’ of the system to understand what is and is not working with what we have. We can view existing educational system structures through the lens of “first-order structures” — such as curriculum/standards, assessment, grade levels, etc. — which act as core pillars to the structure of the system and influence the manifestation of the “second-order structures” — such as pedagogy, application of learning technologies, etc. In other words, the shape and nature of the first order structures dictate the type and the manner in which the second order structures integrate (Adamson, Åstrand & Darling-Hammond, 2016). As a result, these structures have a significant impact on the daily practices and ultimate outcomes of the system (Darling-Hammond, 2004).
Assessment has garnered much of the attention as the foremost first-order structure of the system that must be reformed, as it and the policies that implicate its use are the ‘tail that wags the dog’, and ultimately drives much of the changes and mechanics that translate into day-to-day practices of the system (Darling-Hammond, 2004). Although the considerable focus on improving assessments is warranted, curriculum (or the “what” of education — what is to be learned, including skills, knowledge, competencies) — or more specifically, how we organize and model the learning outcomes — is often overlooked as an area in deep need of redesign.
The curriculum of many schools is generally dictated by the state or national standards under which they operate. Content standards are policy documents, developed by a jurisdiction (e.g. Massachusetts, Ontario, Finland) to set a standardized pace and benchmark of student learning outcomes in each major domain (i.e. mathematics, literacy, science, etc.). They have become the core pillar of most educational systems because they set the course for the way in which so many other elements of the system play out — dictating what is taught when, determining the very nature in which instruction and assessment are implemented, deeply influencing the type of curriculum and instructional materials are built, and implicating what/how data is modeled and collected in many new learning technologies.
In the U.S., when the new Common Core State Standards (CCSS) were released and adopted by 46 states, we essentially saw an entire reorganization of the educational ecosystem around them. Publishers scrambled to create new instructional materials that were aligned with the CCSS (and/or retrofit previous materials to now explain how they align with it), learning technologies needed to do the same in order to be purchased/adopted, and new national assessments were developed. In Brazil, a new national curriculum was launched in late 2017, where the same dynamics are playing out. It is a pattern seen over and over again in educational systems around the world.
Once policy is set around a set of standards, it becomes the bar by which everything else is structured because it becomes the non-negotiable starting point for teachers, the end-users of the system. Instructional materials, learning technologies, assessments, professional development, all are nearly designed either using the standards as their starting point or are designed in coherence with them. In other words, because of the policy surrounding state and national standards, and therefore dictate what schools teach, all the other core elements used in educational systems are influenced, adapted, and aligned to them. When the CCSS was launched, for example, we saw a cascade effect of adjustments in the system, where two new assessment consortia were established to create the assessments that should support the new standards, and publishers moved quickly to retrofit their previous materials to align to the CCSS while also developing new materials. What can be seen is that in the end, nearly everything comes back to these standards. As such, they serve as essentially as the ‘foundational layer’ to the learning data infrastructure in education.
Yet the infrastructure and methods by which we currently manage this core element of the system — largely standalone, flat PDF documents adopted by each system that only get revisited every 10–15 years — is problematic on a number of levels — including the manner in which we attempt to align instructional tools, capture learning data, and share resources and make meaning across systems or sets of standards. Moreover, key players in education, such as assessment designers and learning scientists, work with critical material related to content, but in a way that doesn’t currently map onto or coordinate with content standards. The ways in which we model learning in the learning sciences and leading edge technologies very poorly aligns with the infrastructure set by standards, fundamentally leaving us at impasse, where the evolution and impact of modern learning technologies as a key part of this ecosystem will be seriously impeded. This inadequate infrastructure is both a data and socio-technical problem that is ripe for re-engineering. Purposefully engineering a learning data architecture to work across these practices and user groups has the potential to significantly reduce these gaps.
The Problem with Curriculum Standards
Although a panel of experts is typically employed to generate a set of standards, ultimately the finalized set is adopted and endorsed by a governing body that also puts legislation in place with the standards that has implications for schools that then must teach with them. Standards are often written intentionally to be both comprehensive yet also vague, so as not to over-prescribe what and how teaching and learning plays out in each classroom and allow room for the teacher’s professional judgment as needed when designing instruction around a given standard (Konrad et al., 2014). In other words, this is largely what distinguishes curriculum from standards: standards are not curriculum, but rather benchmark statements encompassing a mix of content and skills that must be translated into curriculum.
Standards began to emerge in the early 1990s as a response to the influential 1980s education reform movement (Beatty, 2010). Since that time, they have largely served as the core infrastructure for managing content in K-12 education, which is evidenced not only by the systemic adoption in schools, but by the extent of their reference and usage in educational technologies and products. In other words, the entire architecture of new learning technologies is often aligned to one or more sets of standards. However, using standards as the core infrastructure for managing content/curriculum in education is problematic on a number of levels:
- Lack of detail / support for educators. Standards are written intentionally to be concise, often integrating numerous topics, and often display little to no further detail on key aspects for teaching that content, such as misconceptions, evidence of mastery, etc. Moreover, a large body of research shows that effective student learning in the classroom is most dependent upon teacher expertise, including pedagogical content knowledge, understanding the role of student misconceptions and how they must be addressed in instruction, the role of specific pedagogies, as well as tasks to elicit understanding and mastery (Darling-Hammond & Ball, 1998; Bransford, Brown & Cocking, 2000).
- Integration / conflation of learning constructs within a standard. An individual standard as a unit of learning can vary greatly in scope and complexity, which impacts and complicates the way in which learning technologies and resources support and capture learning data for the teacher or learner to interpret.
- Lack of coherence / alignment across jurisdictions. Due to this conflation of constructs, as well as variation in design, sets of standards from various jurisdictions do not correlate to one another — effectively each jurisdiction working on within its own system.
- Semantic web integration. Since a common language for these learning constructs has not yet been defined, semantic web tools like knowledge graph cannot yet be easily employed — limiting our ability for information, data, and knowledge-sharing across learning objects, and more broadly in the domain learning and education.
- Standards are not ‘living’, and as such, not able to accommodate more current and relevant research findings and societal demands. Typically, a jurisdiction will redesign and adopt a set of standards every 10–15 years, with long cycles of responsiveness and adaptation to modern insights and demands.
Although standards largely serve as our current infrastructure for managing learning goals, there is a much larger ecosystem around which we study, model, and support learning. Looking more broadly at how learning constructs outside content standards is managed, we see additional challenges with existing infrastructure:
- The learning sciences’ research-to-practice divide. The massive gap between the research produced around learning and its translation to and implementation in practice — as well as the gap from insights from real-world learning environments feeding back to inform learning sciences’ models and frameworks, such as learning progressions. However, getting this knowledge into the hands of practitioners, and ultimately show up in everyday classroom practice, is a persistent gap and an ongoing challenge in education (Hattie, 2009; Hille, 2011). Today, a large swath of learning constructs — from physics and mathematics to literacy, computational thinking and beyond — and the nature of how learners come to understand them, have been studied extensively. This includes identifying misconceptions, elucidation of specific learning tasks and strategies to facilitate more effective learning, common learning trajectories, and more. Yet the ‘thin’ documentation of standards does not allow for (or has not traditionally) been able to fit in this critical knowledge as well. As a result, it’s up to professional development and access to other teacher resources to support them in this critical knowledge — which is non-uniform and mediocre at best.
- The movement towards mastery-based education and personalized learning. In many parts of the US and now much more globally, there is a pronounced movement of leading districts moving to competency-based education (Singer, 2006; Sturgis, 2015a), where personalized learning and individual paths towards mastery are supported, versus standards-based education, which focuses on ‘proficiency’. There is considerable support from a variety of education stakeholders¹, because competency-based education can more effectively support learning environments in mastery-based learning, and more effectively aligns with learning progressions and a learning sciences’-based understanding of the nature of learning. However, the field has yet to common define a model for a competency, and as a result, these leading-edge schools are left to define these on their own. In many cases, districts will appoint a small subset of teachers to collectively define the competencies they will use based on the existing standards they follow²; as a result, what constitutes a ‘competency’ varies significantly within those competencies and across districts.
- The movement towards broader skills and competencies. There is a long-standing discussion in education about the critical need to move away from solely focusing on content-area skills (i.e. math, science) to more broad or ‘higher-level’ skills such as collaboration, communication, critical thinking, creativity, inquiry, persistence, etc. (OECD, 2018; Duckworth & Gross, 2014; National Research Council, 2012). The recent surge in research on what were previously dubbed non-cognitive skills and more recently just commonly being referred to as ‘modern competencies’ have helped this movement. However the central challenge remains that these are not currently “tangible” — meaning, they are not fleshed out at the construct and assessment level, making them a bit like black boxes.
- Coordination of learning metrics across modern learning technologies. Emerging learning technologies, such as intelligent tutors, game worlds, and game-based assessments represent some of the most robust and complex learning tools available today. Due to their complexity and the way they are constructed, they are still ‘black boxes’ for many educators³ — unclear with regards to exactly what learning the tool is directly supporting and/or assessing. If we look ahead to the future, this complexity and ambiguity will only increase as these tools evolve. At the same time, there is little coherence amongst the tools themselves. Many game-based learning tools and game-based assessments are built on similar learning model frameworks, yet are constructed differently, so that data about learning (and therefore the learner) cannot coordinate between them. To be clear, this is a problem that goes beyond the scope of managing content and learning constructs.
- Supporting learning beyond the current educational system. How do learners use, need, and/or benefit from maps of learning that support them as they drive their own learning in skills in and beyond what is covered in school? How do we honor and integrate learning beyond the walls of the school?
Moving from What Was, to What Could Be
Certainly for anyone who has spent time working in the field of education, many of these tensions are very familiar. However what may not always be so evident is how these ‘structures’ truly inhibit our ability to support modern, competency-based learning environments and ecosystems.
The Tension of Learning, Systems & Complexity
The fundamental tension playing out in our current reality is to how to create the supports, structures, and systems that enable modern learning to be every child’s experience — without the system being the problem itself to impeding this outcome. It is the tension framed by German philosopher, Jurgen Habermas, who explained that all of society’s enterprises — from the family to the corporation — possess both a lifeworld and a systemsworld (1987). In societal learning, the lifeworld is made up of the traditions, practices, needs, and purposes of learners and teachers; the management decisions, protocols, policies, procedures, and accountability assurances comprise the systemsworld. The quality, health, and effectiveness of a learning environment erodes when a systemsworld is the generative force determining the nature of the lifeworld. Habermas refers to this latter situation as the “colonization” of the lifeworld by the systemsworld and attributes many of society’s ills to this situation (Sergiovanni, 2000).
Or perhaps more simply, we’re now stuck with the most complex QWERTY problem.
How can we design, construct, and implement beyond this? A modern learning systems architecture is able to both overcome these existing shortcomings, and be flexible enough to accommodate the emerging dynamics and directions of learning environments and systems. A coherent learning data architecture supports the range of needs and practices across the various stakeholders, through data models that are integrated and create congruency. In other words, we should see integration, not segmentation; coherence, not disjointedness.
Each of the aforementioned challenge areas works with mapping learning in some way, often with similar (but not standardized) language or data structures. This lack of coherence and infrastructure leaves a very fragmented knowledge base, tools, policies and practices in relation to content, learning goals, and developmental pathways. These aforementioned challenges are in fact symptoms and surface-level challenges of a deeper root-level problem: there is no unified data architecture for learning constructs (knowledge, skills and competencies). Moreover, engineering a modern competency architecture drawing from current learning sciences research and modern technology requirements would create the foundation for robust, dynamic and personalized learning pathways across schools and beyond.
Engineering a Future Learning Ecosystem
More than 25 years ago, Banathy and colleagues made a compelling case for designing systems of education and moving beyond our current image of education to create one that served everyone better (1988; 1992a; 1992b; 1995; Banathy & Jenks, 1993; Kahn & Reigeluth, 1993). Their work argued for why reform was largely piecemeal and incremental, and from a systems design approach was vital, and unpacked the layers of the system that would need to be considered in that redesign. What was missing in this work was the actualized design and implementation of such a new system.
Their ideas and arguments still hold true today; yet since that time, digital technologies have transformed our daily lives, our ways of communicating, collaborating, learning, and working. At the same time, since their work, systems engineering practices have come to bear robust approaches that helped to engineer complex solutions in a wide variety of domains — building on the idea of redesign to ultimately engineer a new reality.
This significantly shifts the landscape of what an idealized design for education might look like, and perhaps most critically, it adds a layer of technological complexity that was not there at the time of Banathy’s work. These technologies have created the opportunity to dramatically redesign systems to support deeper learning at scale in a way that was perhaps imaginable nearly three decades ago, but not yet possible. In Banathy’s time, redesigning systems structures in education might include rethinking the use of grade levels, curriculum topics, class schedules, etc. These are all parts of the current system, still, worth rethinking. Yet in the last two decades, the impact of new technologies has created a new reality and an emerging infrastructure that completely changes how learners engage with ideas, materials, other learners — across space and time. All of these innovations in technology and practice must be considered in any systems redesign, but they create the opportunity for rethinking systems structures for learning in bold and dramatic ways — some of which might not be immediately evident to those in the current system.
To be sure, this is indeed an exciting opportunity. Yet the technical aspects to such an opportunity creates complexity that will require more than just design methods, it requires engineering. We can, and should, design systems — but we must engineer solutions.
In the next article in this series, we will look at emerging steps globally to do exactly that — and the implications this has for learning environments.
 Including major foundations, as well as organizations such as Digital Promise.
 Personal communication with leaders in state systems moving to competency-based learning.
 Personal communication with dozens of educators through the work of the Playful Learning project.