The Future of Instruction Modeling

Instruction modeling is still in its early stages. This chapter discusses promising directions in which instruction modeling could develop in coming years. This includes increasing the richness of interfaces used in instruction modeling programs (e.g., by allowing students to enter responses in free form and have them graded via natural language processing); applying instruction modeling to subjects beyond mathematics, including English, foreign language, and science; using educational data mining to create automated “coaches” to help teachers better implement instruction modeling programs in their classrooms; creating approaches to instruction modeling that allow for rapid authorship of content; redesigning schools (in schedules as well as architecture) to optimize the use of instruction modeling; and putting in place government policies to encourage the use of comprehensive blended learning programs (such as those developed through instruction modeling).

How to Conduct Instruction Modeling

This chapter provides a step-by-step guide to conducting instruction modeling. Once an instructional tradition to be modeled has been chosen, the first step is to create a “generic model,” which is a detailed description of the general pedagogical goals that teachers in this tradition pursue and the methods they use to achieve them. The second step is to design a learning platform suitable to modeling the selected tradition; the platform includes both software components and the needed offline (implementation) behaviors. The third step is to create instructional content for the platform; each lesson should be carefully designed to reproduce the experiences of the corresponding offline lesson. We discuss trade-offs between more sequential (“waterfall”) and more iterative (“agile”) approaches.

A Comparison with Other Approaches

What are the relative merits of instruction modeling and other approaches to the design of blended learning programs? This chapter discusses several prevailing approaches, including applied learning science, personalization, and the use of big data in education. Many programs are designed around a single claimed feature of good instruction; terming such thinking “featurism,” this chapter argues that it is reductionist and less likely to be successful than more comprehensive approaches (such as instruction modeling). However, instruction modeling is not simply an alternative to other approaches: as the example of cognitive psychology illustrates, instruction modeling can often be fruitfully combined with other methods. Just as good software developers blend different approaches (e.g., using usability testing and the psychology of attention in designing interfaces), good instructional designers should draw on a wide range of techniques. This chapter discusses how instruction modeling can work in concert with big data, natural language processing, and other important approaches.

Identifying Good Instruction . . .

Instruction modeling uses blended learning to reproduce good offline instruction. Thus, any discussion of instruction modeling must first answer the question of how to recognize good instruction in the first place. This question is more difficult than it first appears. This chapter offers a perspective on the answer, providing a foundation for everything that follows in the book. While test scores give the illusion of perfectly measuring student knowledge, they are in fact imperfect indicators. Conclusions about the effectiveness of instructional programs can only be drawn from carefully designed experiments. Even then, quantitative methods paint only an incomplete picture, and therefore should be supplemented by humanistic methods: philosophical discussions, reviews by experts, ethnographies, and histories. We can only hope to tell apart good instruction from bad by combining insights from all of these sources into a single, comprehensive understanding.

Conclusion

What are some of the broader lessons of this book? One of the central arguments is for the importance of instructional content in blended learning. Good instruction without strong content is a contradiction in terms, and creating good content is exceptionally difficult. And yet, the most widely known methods for designing blended learning do not offer much guidance in this matter. This book argues that, whether through instruction modeling or by some other means, designers must recognize the importance of good content and adopt some systematic approach for producing it. A second major argument made in the book is for the importance of combining multiple design approaches. There are many pieces that must be in place for a blended learning program to succeed (including the content, the interface, and the implementation supports), and it is unlikely that any one approach—including instruction modeling—will see to all of them. Designers must think comprehensively, and draw from a wide toolkit of methods in order to attend to all of the essential components of good instruction.

Instruction Modeling

This chapter describes the core ideas behind instruction modeling. A promising way to improve mathematics instruction is to import successful approaches from other countries; however, it is exceptionally difficult to do this, since instructional traditions are cultural and the volume of teaching expertise that needs to be transferred is vast. Computers offer a possible way to ease the barriers. Expert systems (invented c. 1970) are a type of artificial intelligence system that uses rules to mimic human decision-making. Following the pattern suggested by expert systems, an instruction modeler studies high-quality offline instruction and then designs computer programs that aim to recreate this instruction. Many important activities cannot be automated, and therefore instruction modeling is necessarily blended learning: some instruction takes place online, while other activities are led by classroom teachers. To illustrate these ideas, this chapter describes several instruction modeling programs created by Reasoning Mind. It also discusses Russian mathematics education, explaining why it is a successful instructional tradition and a suitable choice for instruction modeling.

Good Instruction in Mathematics

This chapter applies the lessons of Chapter 1 (which discusses how to identify good instruction) to the case of mathematics. There has been much controversy about what makes for good instruction in mathematics. Nevertheless, scientific and humanistic sources do allow us to paint a picture. Some instructional methods are less guided (such as pure discovery learning) and others more guided (like teacher-led instruction); scientific and humanistic evidence are in agreement that general guidance is needed, but should not come at the expense of student cognitive engagement. The evidence also consistently shows that instruction should emphasize genuine understanding of the underlying reasons for mathematical principles. Skills (such as fluency in computations) are not in opposition to concepts, but rather in mutual support. Solving varied and unexpected problems is essential in good mathematics instruction. Mathematical “rigor” (meaning precision in expression) plays an important role in mathematical thought, but should be carefully balanced with accessibility for children. While such principles give general guidance, knowing them is not enough to create excellent instructional programs: they need to be applied consistently in each moment of each lesson. Getting these details right is challenging, and can only be done through years of trial and error. This helps explain why good instructional traditions in mathematics are so rare.

The Case for Instruction Modeling

Relying on the examples and lessons of the preceding chapters, this chapter offers a case for the use of instruction modeling. The central value of instruction modeling is that it offers a consistent way to develop high-quality instructional content for blended learning programs. While this may appear insignificant, in reality it is of great importance. Good instructional traditions in mathematics are exceedingly rare, and have developed their content—including textbooks and the content of individual lessons—through decades of trial and error. Instruction modeling offers a way for designers to avoid the great risk of attempting to do the same from scratch. This is the theoretical case for instruction modeling; the practical case comes from the success of the specific blended learning programs developed through its use. We survey the research literature on the efficacy of these programs. The evidence suggests that instruction modeling is not merely a theoretically appealing approach but also one that can be used to develop programs that succeed in practical use.

Implementation

Implementation is of paramount importance in blended learning. Reasoning Mind learned this lesson when one of the organization’s first programs was offered to teachers with very little training and support, leading to a year of seriously flawed program use. Subsequently, Reasoning Mind added extensive teacher training and implementation support, which addressed the problem. Implementation is important in any blended learning program, but is especially challenging for a program developed using instruction modeling: such programs by design involve a reform in the curriculum and instructional methods, and therefore can only succeed when teachers are given the support needed to make the change. One lesson learned is that the more instructionally comprehensive a blended learning program is, the more difficult it is for schools to adopt and implement it. This is a central problem for blended learning, since programs can only make a difference in proportion to their use. Thus, unless conditions are created (for example, through government policies) to make it easier for schools to use comprehensive programs, it is unrealistic to expect that blended learning will lead to meaningful educational improvements.

The Role of Learning Science in Instruction Modeling

This chapter discusses the interplay between instruction modeling and applied learning science in developing blended learning programs. Far from being two alternative methods, these two approaches complement one another. We discuss in depth the Edifice program; this program systematically applies Mayer’s cognitive theory of multimedia learning (CTML), which is perhaps the most developed and empirically supported theory guiding the design of blended learning interfaces.