The ultimate goal of education is to teach students how to learn. The importance of the educational reform brought about by the research on “Deep Learning” in the field of learning sciences is self-evident. Since the concept of “Deep Learning” emerged in 1976, the United States has gradually constructed a theoretical and practical system of “Deep Learning” through clarifying its connotation, formulating frameworks, and exploring implementation steps. After more than forty years of exploration and development, the research on “Deep Learning” in the U.S. has formed many unique experiences. Referring to the experiences and lessons from the U.S. research on “Deep Learning,” China’s research should first clarify its conceptual connotation theoretically, construct a corresponding goal system in top-level design, and gradually deepen related empirical research in educational practice.
United States; Deep Learning; Academic Review; Reflection
“Deep Learning” is a high-investment learning process where learners pursue effective learning transfer and the resolution of real problems through understanding the essence of knowledge and critically applying learning content, with higher-order thinking as the main cognitive activity. In the current era of information explosion, there are higher demands on people’s ability to acquire and process information. Accordingly, the ability for “Deep Learning” has become an urgent learning capability that people need to master. The rise of research on deep learning is a conscious response of modern society to the demands of the times, and cultivating adolescents’ deep learning abilities has increasingly become an important topic in education across countries. The U.S. is the birthplace of research on “Deep Learning,” with studies on deep learning in basic education starting in the late 1970s. Over the past forty years, through the research process of “initial phase – development phase – deepening phase,” the U.S. has consistently focused on theoretical research and practical exploration of deep learning, viewing it as a key factor in students’ core competencies.
In recent years, with the use of new teaching methods such as Massive Open Online Courses (MOOC), Small Private Online Courses (SPOC), and flipped classrooms, research on “Deep Learning” across various subjects has emerged continuously in the educational field and has gradually gained significant attention from many frontline teachers. Currently, research in China on deep learning is still in its infancy, with related academic studies showing characteristics of fragmentation and superficiality, lacking necessary theoretical and practical support or remaining stuck at the technical level of discussion. The issues regarding the local concept of deep learning, the construction of a theoretical framework, and the selection of implementation paths remain unresolved, and consensus has yet to be formed. The stone from another hill can be used to polish jade. By sorting through the relevant literature on U.S. “Deep Learning” research over the past forty years, analyzing its historical evolution and characteristics should provide some reflection for promoting the localization of related theories and practices in China.
Since the late 20th century, with the continuous development and deepening of research in brain science, cognitive science, and artificial intelligence, studies in the field of learning and teaching have gradually shifted from initial attention to forms and technical aspects to the profound meanings and processes of learning. Going beyond superficial learning of symbolic knowledge and focusing on deep learning, pursuing personal meaning of public knowledge has become a fundamental direction of international educational reform. In the 1960s, U.S. primary and secondary schools began large-scale reforms of curricula and teaching based on Bruner’s cognitive structure theory. However, this massive reform of curricula and teaching did not achieve the expected goals, as the reformed curriculum content severely deviated from the actual education in U.S. primary and secondary schools, and teaching overly emphasized mechanical memorization. Frontline teachers believed that “for materials that must be memorized and practiced, the exercise method should be adopted, and one should not think that memorization or practice is dull.” This resulted in a large amount of “superficial learning” and led to a decline in educational quality. In response, the U.S. launched the “Back to Basics” movement in the 1970s to address the decline in educational quality. During this movement, some scholars focused on students’ learning processes and methods, proposing solutions to educational issues in social transformation, leading to the emergence of deep learning research.
The literature for this study is sourced from the WOS (Web of Science) database’s sub-database – Web of Science TM Core Collection. As a globally influential database, Web of Science TM Core Collection has accumulated over 10,000 authoritative academic journals and more than 100,000 academic conference proceedings. The search terms input into this database included: search term “Deeper Learning”; field “Learning Sciences & Education & America”. There were no time restrictions on the search, resulting in 156 documents retrieved, of which 120 effective documents were obtained after excluding irrelevant and duplicate information.
The concept of “Deep Learning” was first proposed by Ference Marton and Roger Säljö in 1976 in the article “The Distinction between Different Kinds of Learning: Outcomes and Processes.” After 1976, the overall number of research studies on deep learning in the U.S. showed an upward trend, but the total was relatively small, and research mainly focused on theoretical studies of related concepts; after 2000, scholars expanded their research fields by exploring and refining implementation paths for deep learning. This period, termed the development phase, saw the top five keywords in this field’s research being “Learning approaches,” “Problem-based learning,” “Educational technology,” “Critical thinking,” and “Teaching.” In 2012, with the advancement of the William and Flora Hewlett Foundation’s deep learning practice project and the “Deeper Learning Network” (DLN), the growth rate of research somewhat slowed, indicating that deep learning research in the U.S. had entered the deepening phase.
In the field of education, many scholars have explored and elaborated on the learning process, delving into the depth of learning as well. In the 1950s, educational cognitive psychologists, represented by Bloom, conducted ongoing explorations in the categorization of learning objectives and educational goals, laying the theoretical foundation for deep learning research.
Bloom and his team created a distinctive educational objective classification system after eight years of research. In “Taxonomy of Educational Objectives (Book 1) – Cognitive Domain,” Bloom categorized educational objectives related to cognitive abilities and skills based on the principle from lower-order to higher-order: understanding, application, analysis, synthesis, and evaluation. Since the complexity of behaviors varies across different categories, the educational objective classification exhibits significant hierarchical characteristics, with “application” being higher than “understanding,” and “analysis” being higher than “application,” and so on. Following Bloom, American educational psychologist Gagné first proposed cumulative learning theory in his 1971 article “Teaching Based on Research on Learning.” The core essence of this theory includes four levels of intellectual skills: discrimination, concepts, rules, and higher-order rules, with objectives gradually progressing from shallow to deep. Gagné’s theory facilitated the in-depth development of learning objective classification. Following Bloom and Gagné, Merrill was another scholar who systematically elaborated on educational objective classification theory. In 1971, he proposed a classification system of educational objectives composed of four levels and ten types of acquired behaviors in the article “Determining the Psychological Conditions Necessary for Teaching Outcomes.” Among these, cognitive objectives consist of memory behaviors and complex cognitive behaviors, which further include more detailed cognitive hierarchies. In summary, the early scholars’ in-depth exploration of educational and learning objectives from the perspective of learning psychology laid the foundation for the future design and development of learning forms, especially deep learning.
Academically recognized deep learning research began with a series of groundbreaking experiments conducted by Marton et al. at the University of Gothenburg in 1976, aimed at exploring the different levels of information processing adopted by students when reading large amounts of prose. The study found that students typically had two tendencies when understanding the content of articles: some students attempted to grasp the article as a whole, summarizing its main idea through the transfer of old and new knowledge and deep thinking; while others focused on paragraphs that might be questioned, attempting to memorize and recite the article’s content. The researchers argued that students are at different levels when learning knowledge, and the corresponding differences in levels are described based on whether learners engage in “surface-level” or “deep-level” processing. Based on this, they divided the ways in which students acquire and process information into “surface learning” and “deep learning.” This was the first time the concept of “Deep Learning” was proposed in the field of education, gradually gaining academic attention as a more in-depth learning method compared to the prevalent “surface learning” in classroom teaching. U.S. scholars Entwistle and Ramsden subsequently deepened Marton et al.’s related research, extending the connotation of deep learning to the formation mechanisms of learning and analyzing the characteristics of deep learners in the classroom. They found that, unlike surface learners, deep learners acquire knowledge by exploring beyond the main points, engaging in active participation and critical thinking about information. In the learning process, they show great interest in the information learned and maintain composure; deep learning occurs when learners engage in focused, long-term reflection and transcendent thinking about a problem or domain.
Subsequently, another important shift in deep learning research was to regard deep learning as a high-investment learning method primarily characterized by “higher-order thinking” as the main cognitive activity. In 1995, researchers such as Harel interpreted the state of deep learners during deep learning. The research found that during deep learning, learners focus on broader contextual information and the internal connections between materials. They recall and associate by reconstructing the qualitative relationships between information, reshaping the causal relationships between phenomena, achieving a deep understanding of the learning content, which fully demonstrates human higher-order thinking abilities. Beattie, Collins, and McInnes believed that the goal of deep learning is to enable students to learn to understand, primarily reflected in a critical understanding of the learned content, emphasizing logical relationships and drawing evidence-based conclusions. Bransford and other scholars focused on deep learning as a learning ability directed at problem-solving, considering deep learning ability as the capacity of students to extract prior experiences to solve problems in different new contexts.
This stage marked the process from the proposal to the gradual enrichment of the concept of deep learning. Scholars’ research began with distinguishing between deep learning and surface learning, arguing that deep learning differs from mechanical memorization and simple understanding of knowledge, but rather seeks the essence of learning. The connotation of deep learning ability mainly includes the ability to transfer knowledge, creativity, and the ability to solve problems in new contexts.
In the face of challenges such as globalization, knowledge economy, and immigration issues, the U.S. seeks to improve talent quality and global competitiveness by reforming public primary and secondary education. Despite decades of efforts in educational reform, there are still concerns in the U.S. that primary and secondary students are not acquiring the knowledge and skills necessary for the future and are unable to become attractive and innovative citizens in a changing world. Since 2000, the Organization for Economic Cooperation and Development (OECD) has conducted the “Programme for International Student Assessment” (PISA) tests annually on 15-year-old students from 65 countries, with U.S. students typically scoring in the middle across all categories. Surveys indicate that, although educational reforms in the U.S. have not ceased, American adolescents’ scores in science and reading still fall below the average compared to other countries and regions. During this period, the OECD studied the teaching characteristics of high-achieving countries, revealing that students in these countries often have opportunities to engage in deep learning. In order to revive the leading position of American education in the world, and to help more students understand and learn deep learning, researchers shifted focus from the concept of deep learning to the process of how to promote students’ deep learning and gradually constructed a theoretical framework for deep learning.
Regarding how to promote students’ deep learning, scholars have made fruitful explorations from the aspects of classroom teaching, school environment, and educational technology. Marton and Booth conducted experimental research on how teachers can promote students’ deep learning through deep teaching. The research indicated that teachers have a responsibility to hold “awareness meetings” between their knowledge and that of their students, determining corresponding paths and rules through instructional improvements to ensure that the content taught can be operationalized. In terms of specific strategies, they suggested that teachers could guide the development of students’ deep learning abilities through intuitive teaching methods, such as teaching games and situational re-creation. Smith and Colby also noted that teachers’ efforts to cultivate deep learning outcomes do have certain effects. In deep teaching, they suggested the following strategies: (1) support teachers in engaging in dialogues and exchanges about surface and deep learning; (2) examine teaching practices and the resulting student learning conditions; (3) rethink classroom assessments and in-depth learning methods. Jensen and Nickelsen discovered through years of teaching practice that using mind maps in classroom teaching could build a foundational awareness for students’ deep learning and promote the sustained development of students’ higher-order thinking. To enable teachers and students to become successful learners at a deep level, they also developed a new teaching model called the “Deep Learning Circle.” The “Deep Learning Circle” points out the paths for students at different positions in the learning journey and provides references for teachers on how to guide students in deep learning.
With the deepening of research, scholars found that in addition to basic classroom teaching, the invisible curriculum of campus culture also subtly impacts students’ deep learning processes. A representative figure in this area of research is Phan Huy P, who conducted empirical research in the field of deep learning pedagogy and critical thinking development courses, and based on a questionnaire survey of one hundred students, drew a potential growth curve model. The research results showed that a relaxed and harmonious learning atmosphere helps stimulate students’ critical thinking and promote the development of their deep learning abilities. Vos and others conducted empirical research on the deep learning strategies of 235 students from four primary schools, concluding that an unconstrained game environment and atmosphere would enhance students’ motivation to participate in deep learning behaviors. Warburton explored the factors influencing deep learning in environmental education from the opposite perspective, arguing that if students have weak attention to existing interests or backgrounds, it would inhibit deep learning.
In 2012, Abbas Sadeghi and others used surveys to explore multiple factors influencing students’ deep learning, revealing that factors affecting students’ deep learning include teacher-student characteristics, learning goals, learning strategies, and academic activities. In 2013, Loyens and Sofie M conducted empirical research on students’ academic achievements and learning attitudes from the perspective of learning organizational forms, revealing that students in problem-based learning groups have stronger learning motivation and opportunities to adopt more effective learning methods. Overall, the problem-based learning model facilitates students’ deep learning.
Additionally, the latest research achievements in disciplines such as computer science and brain science have provided valuable resources for the development of deep learning. Remote education, mediated and carried out by computers, has made online deep learning possible for students. In 1992, Henri developed an analytical model that educators can use to better understand the learning process. This model emphasizes five dimensions of the learning process: participation, interaction, social, cognitive, and meta-agency. Henri’s model provides information on participants as learners and how they process a specific topic, facilitating online deep learning. In 1996, Oliver and McLoughlin made some changes to Henri’s analytical model based on relevant research findings from brain science, recognizing five types of interactions: social, procedural, instructional, explanatory, and cognitive. This improved model has been used to analyze various strategies in remote teaching and traditional teaching.
Increasingly, research indicates that deep learning is influenced by multiple factors, and the generative mechanisms of deep learning are complex and diverse. As scholars delved deeper into research on deep learning implementation strategies, a framework aimed at cultivating students’ deep learning competencies gradually took shape. In 2012, the core institution advocating deep learning in the U.S., the National Research Council (NRC), identified deep learning capabilities as consisting of three dimensions: cognitive, interpersonal, and personal. The William and Flora Hewlett Foundation, through systematic analysis and synthesis of existing research, proposed a basic framework for “Deep Learning” competencies, interpreting it as the development of six core abilities, including mastering core academic content, critically thinking, and solving complex problems.
In this framework, the goals of deep learning are not limited to deepening students’ understanding of knowledge but are committed to creating new knowledge and applying it to solving real-world problems. In other words, deep learning does not require students to simply replicate or apply knowledge but requires them to integrate existing information and concepts to generate new ideas and solutions.
At this stage, research on deep learning has broken through existing patterns and categories, shifting researchers’ focus from exploring the connotations and characteristics of deep learning to how to promote students’ realization of deep learning and what dimensions to stimulate the development of students’ deep learning abilities. Researchers have constructively proposed the basic theories and operational frameworks for students’ deep learning abilities. A consensus has gradually been reached that enhancing students’ deep learning capabilities can be achieved by updating teaching goals, strategies, school culture, and other aspects. A prominent achievement of this stage is the deep learning framework developed by the William and Flora Hewlett Foundation, which also provides a basis for the next stage of deep learning projects.
In recent years, there has been a high degree of consensus in the U.S. regarding the expectation to improve the educational level of all students. However, in the education sector, American schools tend to provide two-tiered curricula for students from different backgrounds. One tier is mainly set for middle-class white students and relatively affluent families, allowing them opportunities for deep learning; the other tier is primarily for low-income and minority students, focusing almost exclusively on basic skills and knowledge. Policymakers have tracked the learning status of students from low-income and minority families, revealing that they have not mastered basic skills such as reading or math calculations well. In response, a movement supporting “deep learning” has emerged among researchers, policymakers, and teachers, aiming to improve students’ future success in college, careers, and civic life.
Since 2013, research in the field of deep learning has shifted from theoretical analysis to practical exploration, with the “deep learning” movement aimed at improving educational quality and promoting educational equity gaining momentum. Among the existing deep learning research, the “Study of Deeper Learning” (SDL) project initiated by the William and Flora Hewlett Foundation and implemented by the American Institutes for Research (AIR) is unique in both theoretical development and practical innovation, marking a milestone in U.S. deep learning research.
The WFHF is a leader in promoting the deep learning movement in schools, defining “deep learning” as “a set of competencies that students must master to form a keen understanding of academic content and apply that knowledge in classrooms and workplaces.” In 2014, the WFHF funded AIR to conduct a comprehensive and in-depth “Deep Learning Study” project. The SDL project focused on 19 high schools that adopted or partially adopted deep learning teaching, primarily using quasi-experimental research methods that combined quantitative and qualitative approaches. The study began in 2014 and lasted for three years. All 19 experimental schools in the project adopted classroom teaching models focused on deep learning and school systems and cultures that support the successful implementation of these teaching strategies. The results showed that learners participating in the deep learning project performed at higher levels in academic achievement, self-efficacy, and teamwork skills compared to 11 schools that did not implement deep learning projects. Based on experiential learning, the WFHF conducted a detailed analysis of the deep learning capability framework, categorizing competencies and summarizing the roadmap of competencies and corresponding implementation strategies. AIR ultimately established a six-dimensional deep learning practice framework that includes mastering core academic content and effective communication.
Compared to the SDL project, the “Deep Learning Network” (DLN) project in the U.S. is more focused on educational equity and quality improvement for minority citizens. The DLN consists of alliances such as the New Technology Network, Asian Society, and Vision Schools, aiming to cultivate students’ essential skills for adapting to 21st-century social life through deep learning. The DLN covers more than 500 charter schools and public schools across 41 states, with most students from low-income and minority families. The deep learning goals of the DLN include: (1) Master core knowledge; (2) Critical thinking and complex problem-solving; (3) Effective communication; (4) Collaborative learning; (5) Learning how to learn; (6) Developing a positive learning mindset.
The DLN primarily focuses on the curriculum design for deep learning. In terms of goals, there are no unified curriculum standards set among alliance schools, and the principles for curriculum development mainly focus on deep learning goals while aligning with state curriculum standards and requirements for student graduation. In terms of development, the main bodies for curriculum expansion can be subject experts as well as frontline teachers and students; in terms of content, the curriculum includes not only subject courses but also project-based courses, interdisciplinary integrated courses, and comprehensive practical activity courses. For example, Avalon School, as a member of the “Vision Schools” alliance, has most of its project-based courses designed and led by students; Codman Academy has collaborated with a local theater to develop a literature integration course, allowing students to learn literature and history through comedic performances, gaining corresponding competencies.
From the perspective of alliance schools’ positioning, these schools aim to create communities of professional learning where students can freely explore. The alliance schools uphold a shared belief that professional learning communities are a powerful form for students to engage in deep learning. To this end, the DLN employs three core concepts in community planning: first, providing environments for all students to engage in deep learning; second, allowing all students to enjoy the joy of deep learning through collective cooperation; third, ensuring that the final outcomes promote the growth of everyone, including teachers and students.
From the perspective of school culture, teachers and students in alliance schools can learn in an atmosphere of strong motivation and high efficacy orientation, and can autonomously construct a humanistic environment of equality, respect, trust, cooperation, and sharing. In the DLN, schools are not only places with educational functions but also resemble a family. Here, teachers and students, as well as students among themselves, can freely form learning communities, establishing deep collaborative relationships based on interests and abilities. In practical terms, schools have established deep learning groups, with teachers acting as advisors responsible for providing immediate and long-term technical support to students. During this period, teachers and students collaboratively negotiate and explore learning plans, learning forms, and learning progress. For example, in Codman Academy’s teaching plan, advisors and students meet for at least 30 minutes each time; in public school alliance schools, learning groups must hold at least one hour of video meetings each week.
Regarding the evaluation of educational quality in alliance schools, data obtained by the American Institutes for Research indicate that students in SDL experimental schools demonstrate higher levels of learning motivation, collaboration skills, and academic achievement. Overall, students participating in SDL and DLN schools outperform those in non-participating schools in academic performance, and they also show better high school graduation and college admission rates.
At this stage, research in the field of deep learning has completed its practical shift. Relevant studies have transitioned from singular theoretical analysis and interpretation to focusing on empirical research and practical exploration. The deep learning movement and the educational reforms in schools have further promoted the development of deep learning research, making deep learning no longer merely a theoretical exploration and educational experiment but a real learning goal that individual students can achieve.
In the field of education, macro and micro research cannot be neglected. If there is only micro exploration without macro planning, it is easy to fall into blindness; conversely, if there is only macro oversight without micro details, it often becomes superficial. The micro needs to be guided by the macro, while the macro requires validation from the micro. For decades, the development of deep learning research in the U.S. has effectively balanced macro aspects, such as the formulation of policies related to deep learning in the context of knowledge economy and globalization, making deep learning a research hotspot in the education field. Meanwhile, research on deep learning has also considered specific strategies at the micro level, focusing on how deep learning penetrates specific educational contexts, such as how school classroom teaching, campus culture, and educational technology influence the development of students’ deep learning abilities. Macro research leads micro research, while the results of micro research provide necessary support for macro research. Over the past forty years, U.S. deep learning research has successfully integrated macro and micro research.
In the field of education, theoretical research and empirical research are essentially interdependent. Progress in educational theoretical research can promote empirical research, while empirical research often pushes theoretical research to a more rigorous level. Early deep learning research primarily focused on theoretical analysis, with significant contributions in clarifying the connotations and characteristics of deep learning. However, deep learning is not purely speculative; as deep learning integrates deeply with school education, scholars have increasingly entered real classroom settings, utilizing empirical methods such as educational experiments to explore the formation processes and influencing factors of students’ deep learning, achieving fruitful results that greatly promote the development of deep learning theory and practice. It is evident that the interdependence and close connection between the two can promote their mutual development.
The concept of deep learning originated from research in multi-layer neural network machine learning in the field of artificial intelligence. In this field, deep learning is an algorithmic thinking process, the core of which is the computer’s simulation of the complex logical thinking and information processing of the human brain. New achievements from other disciplines have prompted researchers in the education field to deeply reflect. Deep learning research in the U.S. not only absorbs research methods from disciplines such as computer science and brain science but also further transfers their research outcomes into specific teaching strategies aimed at cultivating students’ deep learning abilities, such as the proposal and application of online deep learning frameworks in remote education. In recent years, achievements in areas such as software engineering and data engineering in the U.S. have gradually been applied to deep learning projects and have become important means for schools to conduct teaching activities.
Overall, U.S. deep learning research has the following deficiencies in terms of content, subjects, and methods: First, in terms of research content, existing studies mostly describe the concept and generative mechanisms of deep learning from the student perspective, revealing the conditions for the occurrence and development of deep learning through specific viewpoints or operational systems. However, there has been insufficient research on the “deep teaching” aspect guiding students’ deep learning. In the basic education phase, since students do not yet possess the ability to independently engage in deep learning, “Deep Learning” aimed at students’ core competencies should be a bilateral activity jointly undertaken by teaching and learning. Existing studies have not deeply explored issues such as teachers’ teaching motivations, teaching objectives, teacher-student interactions, and reflective learning in the deep teaching process, and there is a lack of research aimed at designing deep teaching projects and models for teachers. Second, in terms of research subjects, there is an excessive focus on primary and secondary education students, with relatively little research on university students. As a country deeply influenced by multiculturalism, U.S. deep learning research should possess a multicultural perspective, yet existing studies are overly concentrated on domestic and English-speaking countries, with research on deep learning in other cultural contexts still awaiting development. Third, in terms of research methods, U.S. deep learning research has overly focused on students’ deep learning in specific projects, with insufficient research on deep learning in students’ daily natural states.
In China, concepts such as “Deep Learning” and “Deep Teaching” have only begun to emerge in the educational field over the past decade. Currently, research on deep learning in China is still in its preliminary exploration stage, with both the quantity and quality of research needing improvement. By sorting through the historical evolution and characteristics of U.S. deep learning research over the past forty years, the author believes that research in China’s “Deep Learning” field should be improved in the following aspects.
In summary of the U.S. understanding of deep learning, it differs from superficial learning primarily characterized by mechanical memorization. Deep learning is a learner-driven, critical learning method, and an effective way to achieve meaningful learning. American scholars distinguish it from superficial learning that does not involve higher-order thinking processes, repeatedly demonstrating its differences and innovations from other forms of learning in experiments. They argue that while other learning types reflect learners’ understanding to some extent, deep learning pays more attention to critical higher-order thinking, active knowledge construction, effective knowledge transfer, and the resolution of real problems.
In this regard, Chinese scholars have also conducted related research. In their studies, scholars have directly introduced foreign definitions of deep learning to some extent, considering deep learning as a critical learning method for new ideas and knowledge, integrating them into existing cognitive structures to make decisions and solve problems. Some scholars also view deep learning as a learning approach that emphasizes critical thinking. However, when advocating a new type of teaching or learning method, the first step should be to clarify and analyze the concept. However, in reviewing China’s deep learning research, few scholars distinguish it from various learning methods advocated in the new curriculum reform when analyzing its connotation. What exactly distinguishes deep learning from meaningful learning, heuristic teaching, problem-based learning, and even cooperative learning in terms of its essential characteristics? What specific abilities and qualities should be focused on cultivating in practice? Such inquiries should become urgent issues to be addressed in China’s deep learning research and directions for future research.
In terms of overall development strategy for education, when research on deep learning in the U.S. was basically mature, it was timely regarded as a fundamental learning ability that American citizens should possess in the 21st century. To address the challenges posed by rapid changes in the 21st century to education, the “Partnership for 21st Century Skills” developed the “Framework for 21st Century Skills,” which specifies an educational goal system involving core subjects, themes, and basic skills related to students’ deep learning. Over decades, U.S. deep learning research has continuously adjusted and deepened, aligning with the evolving requirements for talent development and the core objectives of national education reform, emphasizing top-level design in research, and gradually establishing a deep learning goal system focused on cognitive, interpersonal, and personal domains as an important theoretical basis and practical tool for guiding deep learning practices.
China has already proposed relevant requirements for talent cultivation goals from a strategic height in the “National Medium- and Long-Term Education Reform and Development Plan Outline (2010-2020),” stating that “efforts should be made to enhance students’ sense of social responsibility in serving the country and the people, their innovative spirit of daring to explore, and their practical ability to solve problems.” Since 2005, research on deep learning in China has undergone more than ten years, but a macro, overall goal for deep learning directed at all students has yet to be constructed, and specific curriculum and teaching objectives combined with various age groups still need to be developed. In recent years, with the formal introduction of the “Core Competencies for Chinese Students,” which emphasizes “ability + character,” core competencies are leading educational and teaching reforms. It can be said that China is facing a development opportunity similar to the U.S. proposing the “Framework for 21st Century Skills.” Therefore, we should draw on the experiences of the U.S. in designing the “Deep Learning Framework” within the “Framework for 21st Century Skills” to align deep learning with core competencies, constructing a deep learning goal system in top-level design to promote the exploration and development of deep learning and deep teaching practices.
Valuing empirical research on deep learning is a lesson learned from the U.S. for later researchers. Since 2012, research in the field of deep learning in the U.S. has shifted from theoretical analysis to practical exploration, focusing on utilizing empirical research to analyze the learning processes and methods of deep learning while also emphasizing evaluation research on deep learning. The implementation of “Deep Learning” research projects has promoted the integration of deep learning theory and practice. However, in the current research on deep learning in China, there is little empirical research on deep learning classroom practices, lacking clear teaching direction.
Future domestic research on deep learning should develop in-depth, deepening empirical studies on deep learning, designing and conducting a series of educational experiments on deep learning. In practice, teaching design aimed at deep learning is undoubtedly a priority, meaning that implementers must comprehensively analyze and coordinate elements such as teaching goals, content, implementation, and assessment, fully leveraging the implicit curriculum role of campus culture, allowing students to achieve deep learning in the classroom and ultimately learn how to learn. Deep learning research experiments can adopt a combined approach of “top-down” and “bottom-up,” where educational theoretical researchers enter the field of deep learning and teaching to test theories based on theoretical discussions, while frontline teachers actively conduct action research based on deep teaching practices, collecting data and refining theories in empirical research to promote the in-depth development of deep learning theory and practice in China.
The paper is sourced from “Foreign Education Research” 2019, Issue 1
