This Issue Highlights
● Table of Contents for Global Education Outlook, 2020 Issue 10● Zhang Xiaolei, Wang Xiaoxiao, Xu Leping | Exploring Technology-Supported Innovative Driven Learning – An Interview with Professor François Tardy
Abstract: The taxonomy of educational objectives implies a layered approach to deep learning, analyzing the classification of objectives from two domains. This study interprets and synthesizes Bloom’s taxonomy, Anderson’s revised version, and Biggs’ SOLO taxonomy in the cognitive domain, summarizing two levels of deep learning objectives: connection and transfer, and critical thinking and creation. The study also deconstructs and integrates Hornstein’s four domains and five levels, and Marzano’s three systems and six levels of comprehensive objectives, identifying three levels of deep learning objectives: adapting to contexts, analyzing and applying knowledge; executing monitoring, constructing self-awareness; maintaining beliefs, and demonstrating self. The hierarchical structure of deep learning objectives has guiding and indicative significance for classroom evaluation practices, specifically: it helps clarify the directional goals and levels of evaluation; it aids in selecting evaluation methods suitable for deep learning objectives; it assists in designing evaluation tasks that stimulate deep learning; it facilitates the development of evaluation techniques to observe the extent of deep learning.
Keywords: Educational objectives; Taxonomy; Deep learning; Classroom evaluation
Author Biography
Zheng Donghui / Professor at the College of Teacher Education, Ningbo University (Ningbo 315211)
1. Research Questions
In recent years, deep learning has been widely and deeply discussed in the field of education, with two research orientations becoming increasingly distinct: one stemming from early cognitive domain deep learning (deep learning), and the other viewing deep learning (deeper learning) as a comprehensive learning competence. As a cognitive approach, deep learning is distinguished from shallow learning, representing “a process of deep processing aimed at understanding the content of learning materials and the ideas the author intends to convey,”[1] primarily reflecting three aspects of depth: “First, attempting to understand the problem and critically comprehend specific learning content; second, emphasizing the connections between new concepts and existing knowledge experiences; third, focusing on the logicality of arguments and linking arguments with conclusions.”[2] In summary, deep learning is learning for understanding and transfer. Discussing deep learning from a competence perspective means regarding deep learning as a capability that students must possess to be competent in 21st-century work and civic life. These competencies enable students to flexibly grasp and understand disciplinary knowledge and apply this knowledge to solve problems in the classroom and future work, primarily including mastery of core disciplinary knowledge, critical thinking, solving complex problems, teamwork, effective communication, learning how to learn, and developing academic mindsets.[3] These competencies transcend and extend the original cognitive domain focused on deep learning, adding interpersonal and self-related fields necessary to cope with future society. Although the connotations and structures of deep learning advocated by these two orientations differ, to translate them into classroom practice for primary and secondary school teachers, it is essential to layer the goals of deep learning, providing teachers with a clear hierarchical direction for objectives; otherwise, “teaching for deep learning” and “evaluating for deep learning” will become mere slogans.
How to layer deep learning objectives? Bloom’s taxonomy of educational objectives serves as an excellent reference point. Some studies directly reference Bloom’s cognitive objectives taxonomy to distinguish four cognitive levels of deep learning: application, analysis, synthesis, and evaluation. Some scholars trace deep learning back to Bloom’s taxonomy, believing that Bloom’s ability hierarchy discusses the depth of learning or deep learning issues.[4] Other scholars have used Bloom’s educational objectives taxonomy to derive the essence and connotation of deep learning.[5] Therefore, one can start from Bloom’s educational objectives taxonomy to explore the layering of deep learning objectives. Of course, it should not stop at Bloom’s educational objectives taxonomy but rather follow the path it indicates, choosing educational objectives classifications with similar values for analysis. What constitutes similar value? According to Bloom, “we categorize different types of objectives hierarchically, … arranging educational behaviors in a sequence from simple to complex, based on the viewpoint that a certain simple behavior can be combined with other equally simple behaviors to form a more complex behavior. Thus, we can express our classification in this form: behavior of type A forms one category, behavior of type AB forms another category, and behavior of type ABC can form yet another category. It should be related to the order of difficulty; thus, questions requiring behavior A are often easier to answer correctly than those requiring behavior AB. … In other words, behaviors are classified hierarchically.”[6] That is to say, Bloom’s educational objectives taxonomy does not merely distinguish different categories of objectives but categorizes them hierarchically based on the complexity of behavior and the difficulty of the problems. This also becomes the basis for our choice of educational objectives classification, which is classification and layering, not only categorizing objectives according to a certain standard but also distinguishing categories hierarchically based on certain rules.
Over the past 60 years, more than 20 new classifications have emerged to update and improve Bloom’s taxonomy.[7] From the perspective of the domains to which these objectives belong, these classifications can generally be divided into two major areas: cognitive and comprehensive, corresponding to the two orientations of deep learning. This study selects five objective classifications based on the standards of classification and layering, considering the self-consistency and influence of objective classifications, namely, Bloom’s cognitive objectives taxonomy and its revised version by Anderson (L.W. Anderson), Biggs’ SOLO classification in the cognitive domain, and Hornstein’s (A.D. Hauenstein) four domains and five levels, as well as Marzano’s (R.J. Marzano) three systems and six levels in the comprehensive domain. This study, starting from the cognitive and competency connotations of deep learning, interprets the corresponding categories of these objective classifications, distinguishing the deep learning objective levels they point to and discussing their significance for classroom evaluation.
2. Deep Learning Layering in the Cognitive Objective Perspective
(1) Bloom’s Deep Learning Levels Based on Outcomes
Bloom classified objectives related to knowledge, cognitive abilities, and intellectual skills from the simplest to the most complex behaviors according to classification principles that distinguish student behaviors, maintain internal consistency, and align with psychological phenomena, summarizing them into six levels: Knowledge (knowledge), Comprehension (comprehension), Application (application), Analysis (analysis), Synthesis (synthesis), and Evaluation (evaluation). Among them, “knowledge” is distinct from other categories because the primary psychological process involved in this category is memory, while in other categories, memory is merely part of a more complex process of association, judgment, and reorganization.[8] This memory is obtained through recognition or recall of concepts, materials, or phenomena, primarily referring to “specific knowledge,” “knowledge of methods for handling specific matters,” and “universal principles and abstract concepts in the discipline.”[9] It is evident that this level of “knowledge” focuses primarily on the recognition and representation of foundational knowledge, while from the comprehension level onward, it emphasizes the transformation, interpretation, and inference of knowledge, progressively ascending to form intellectual skills. Based on the initial understanding of deep learning, it is “the meaningful learning process of what students have learned (what is learned)”[10], where “knowledge” merely represents what has been memorized and has not yet reached the level of deep learning. The hallmark of comprehension is that when explaining the use of abstract concepts, students can use that abstract concept. The hallmark of application is that, without specifying a problem-solving model, students will correctly apply that abstract concept in appropriate contexts.[11] Hence, “comprehension” also does not meet the requirements of deep learning, as it merely serves as a “porter of knowledge,” meaning that “individuals do not need to connect certain materials with others, nor do they need to clarify its fullest meaning, but merely know what is being communicated and can apply this material or concept being communicated”[12], without reaching the level of internalizing knowledge, while “application” becomes the starting point of deep learning, where students can use knowledge to solve definite or uncertain problems. Following this reasoning, in Bloom’s cognitive objectives taxonomy, the pursuit of deep learning objectives ranges from “application” to “evaluation” because, in Bloom’s view, these six levels are progressively ascending and cumulative, meaning that “as long as relatively simple behaviors are seen as components of more complex behaviors, we can view the educational process as one built on the foundation of simpler behaviors”[13]. In terms of the content indicated by “application, analysis, synthesis, evaluation,” “application” refers to using abstract concepts in specific and concrete contexts, “analysis” refers to breaking materials down into their constituent elements, clarifying the relationships between parts and their composition, “synthesis” refers to combining various elements and components into a whole, and “evaluation” involves making judgments about the value of materials and methods for specific purposes. All of these are descriptive specifications of learning outcomes, indicating that for Bloom, deep learning is “deep” in the behavioral outcomes of the four levels.
(2) Anderson’s Deep Learning Levels Based on Process
In contrast to Bloom’s cognitive objectives taxonomy, which is a single-dimensional and outcome-oriented framework, Anderson et al.’s revised framework includes two dimensions: cognitive processes and knowledge. The cognitive process includes six categories: Remember (remember), Understand (understand), Apply (apply), Analyze (analyze), Evaluate (evaluate), and Create (create), forming a continuous unity determined by the complexity of cognition. In other words, “understanding” is more complex than “remembering,” and “applying” is more complex than “understanding,” and so forth.[14] For each category of cognitive process, Anderson et al. constructed a detailed explanatory framework: remembering refers to “retrieving relevant information from long-term memory systems,” such as recognition and recall; understanding refers to “constructing meaning from orally, written, and pictorially communicated instructional information,” such as explaining, exemplifying, classifying, summarizing, inferring, comparing, and clarifying; applying refers to “executing or implementing a procedure in a specific context,” such as execution and implementation; analyzing refers to “breaking materials into several components and judging the relationships between parts, as well as the relationship between parts and the overall structure or overall goal,” such as distinguishing, organizing, and attributing; evaluating refers to “making judgments based on criteria or standards,” such as checking and assessing; creating refers to “combining elements to form an organic whole or developing new models or structures,” such as proposing (hypotheses), planning, and creating.[15] Here, the cognitive processes differ significantly from Bloom’s behavioral outcomes: first, the parts of speech change; Bloom uses nouns focusing on changes in behavioral outcomes, while the revised version uses verbs emphasizing the cognitive processes of acquiring different types of knowledge. Second, the expressions of the six levels change; the lowest level of knowledge is replaced by an independent knowledge dimension, a new “remember” level is set, “understanding” replaces “comprehension,” and new connotations are added, with “synthesis and evaluation” changing to “evaluation and creation.” Third, the underlying learning perspective changes from an outcome-oriented behavioral performance to a constructivist learning perspective, emphasizing that “when students are actively engaged in meaningful learning, they know what they know (knowledge) and how they think about what they know (cognitive process)”[16]. Therefore, one cannot simply apply Bloom’s deep learning levels but must reanalyze the levels of the above six categories.
From the specific connotations of the six categories and their subcategories, it is clear that “remembering” has not reached the cognitive level of deep learning, as the content identified or recalled is only temporarily stored, whereas starting from “understanding,” the situation differs. “When the primary teaching goal is to promote retention, the focus is on emphasizing memory objectives; when the teaching goal is to promote transfer, the focus shifts to the other five cognitive processes from understanding to creating”[17], which aligns with the assertion of deep learning that “learning is for transfer,” where “deep learning is the process by which individuals apply what they learn in one context to another new context (i.e., transfer)”[18]. Here, “understanding” differs from Bloom’s “comprehension.” In Bloom’s view, comprehension is the lowest level of understanding, not requiring the interconnection of knowledge, allowing for unidirectional output of a certain knowledge point in a predetermined manner, without considering the specific context, while “understanding” emphasizes “establishing connections between new knowledge to be learned and existing knowledge”[19], and constructing meaning within context. In this regard, “understanding” should be the starting level of deep learning in cognitive aspects. The progression from “understanding” to “creating” thus becomes the five levels of Anderson’s deep learning, as “the six main categories of cognitive processes are arranged in increasing order of complexity”[20]. It should be noted that these levels differ in their cumulative nature from Bloom’s; “the six categories are allowed to overlap in terms of complexity judgment”[21], indicating a spiral ascent.
(3) Biggs’ Deep Learning Levels Based on Learning Quality
Biggs et al. pointed out the fatal weakness of Bloom’s taxonomy, stating, “Its true intention is to guide the selection of test questions rather than evaluate the quality of students’ responses to a particular question. … Bloom’s hierarchy is a priori, designed in advance by teachers”[22], and thus shifted to assessing the quality of students’ responses in real situations, distinguishing different cognitive levels among students. Biggs et al. are interested in “how good learning psychological processes reflect at different stages of cognitive development, … hoping to use materials to understand the naturally embodied levels”[23]. From this perspective, based on Piaget’s developmental stage theory, they constructed the SOLO (Structure of the Observed Learning Outcome) taxonomy, which structurally classifies observable learning outcomes (the results of students solving tasks or answering questions), distinguishing five levels of learning quality (learning quality), from low to high: prestructural, unistructural, multistructural, relational, and extended abstract. From the perspective of relevant cognitive operations (relating operation), the responses in the “prestructural” level include four situations: refusal, tautology, conversion, and jumping to individual details; the responses in the “unistructural” level summarize based on a single material; the responses in the “multistructural” level summarize based on a few limited, isolated elements; the responses in the “relational” level inductively use relevant knowledge to summarize within a specified context or range of experiences; the responses in the “extended abstract” level involve deduction and induction, summarizing experiences not previously encountered.[24] These five cognitive levels do not accumulate linearly like Bloom’s deep learning levels but present a progressive relationship, with qualitative differences between different levels. From the perspective of cognitive requirements, the first three levels of prestructural, unistructural, and multistructural have not reached deep learning, as they all exist in an isolated and fragmented state of thinking, while the latter two structures have reached the corresponding levels, “because students will use thinking processes such as comparison, explanation, analysis, application, conceptualization, hypothesis reasoning, and reflection, causing qualitative changes in learning outcomes rather than merely an increase in quantity”[25]. In terms of response structure, “relational” involves linking relevant materials to form a conceptual framework, while “extended abstract” synthesizes all relevant materials and their interrelationships into a hypothetical abstract structure that can be applied to instances or materials not initially included.[26]
(4) Deep Learning Levels in Cognitive Objectives Classification: A Comprehensive Perspective
By sorting and comparing the above three cognitive objectives classifications, it is found that each classification contains different levels of deep learning objectives, such as Bloom’s “application, analysis, synthesis, evaluation,” Anderson’s revised version’s “understanding, application, analysis, evaluation, creation,” and Biggs’ “relational, extended abstract.” Although the starting points of the three classifications differ and their category expressions are not entirely the same, there are still commonalities underlying their cognitive perspectives and states of thinking. One is the cognitive and constructivist orientation. Bloom’s cognitive objectives taxonomy’s assumptions about the types, hierarchies, and cumulative nature of learning and transfer demonstrate a psychological orientation that differs from the dominant behaviorism of the time, showing a clear view of information processing and cognitive development, as noted in the retrospective comments on Bloom’s educational objectives taxonomy after 40 years, “How bold the authors of the taxonomy were in their pre-assumptions in psychology, and how modern they are in many respects… Their modernity is evident in that decades later they participated in various developments in psychological theory”[27]. Anderson et al.’s revised version and Biggs et al.’s psychological constructivist tendencies are even more pronounced. Two, the essential connotations of different hierarchical categories tend to be consistent. Analyzing the actual direction of these categories indicates that they all point to advanced cognitive activities of verbal thinking, which integrate and transcend perception and memory, rationally thinking about and acting on knowledge or problems. From commonalities, integrating the three classifications, it becomes possible to delineate different levels of deep learning objectives that are distinct and non-overlapping.
In Bloom’s taxonomy, the four levels are specifically divided, and each level is further subdivided into different graded subcategories, thus forming a linear hierarchical goal based on categories. Due to being overly detailed and emphasizing cumulative hierarchical structures, it has received varying degrees of criticism, with several authorities suggesting that “evaluation is inherently contained within synthesis, so there is no subordinate relationship between them” and “evaluation in the objectives classification should not be placed above synthesis, but at most parallel to it”[28]. Some tests and surveys have also found that the order of “synthesis” and “evaluation” has been reversed.[29] If a qualitative distinction is to be made, it may be better to integrate synthesis and evaluation, emphasizing the judgment, arrangement, and combination of knowledge or materials to form new cognition.
In the revision process, Anderson et al. conducted empirical research on Bloom’s cumulative hierarchy, finding evidence supporting the cumulative levels of “understanding, application, analysis,” but insufficient evidence to arrange “synthesis” and “evaluation,” leading to the replacement of “synthesis” with “creation” as the highest level based on complexity.[30] At the same time, they distinguish between “learning for retention” and “learning for transfer,” emphasizing the “transfer” goals of the five levels of “understanding, application, analysis, evaluation, creation” through examples, clarifying the progressive relationships between these five levels and ensuring that the classifications within each level are parallel and mutually exclusive. However, the revised version has not formed a cumulative hierarchy, primarily reflected in the “understanding” level, where subcategories (such as explanation) are allowed to overlap with other levels. If “understanding” and “application” could be combined to form one level, it may resolve such overlaps and weak distinctions.
Biggs et al.’s “relational, extended abstract” fundamentally distinguishes the levels of deep learning based on the actual cognitive development levels achieved by students in their responses or problem-solving, rather than on point-like cognitive hierarchical distinctions, as in the “relational” level, cognitive objectives like understanding and application may be invoked to complete tasks, while in the “extended abstract” level, analysis, evaluation, synthesis, and creation may be required for cognitive activities to participate.
Thus, Biggs’ layering has a certain inclusivity, fundamentally covering the cognitive levels of the first two classifications based on cognitive development, while avoiding issues of overlapping levels and subcategories. It may be beneficial to use this as a reference point to differentiate the hierarchical distinctions of deep learning from a cognitive state perspective. In fact, deep learning in the cognitive domain focuses on holistic cognition, meaning it does not engage in fragmented learning but integrates knowledge to form a complete understanding, which is so-called “deep” in terms of cognitive processing levels.[31] Based on this, this study takes knowledge or problems as cognitive objects, integrating the different levels of the three cognitive objectives into two states of deep learning, named “connection and transfer” and “critique and creation.”
“Connection and transfer” involves contextualizing explanations, reasoning, categorizing, and summarizing established knowledge or problems, internalizing them into one’s knowledge through the interconnections between knowledge points before and after, and between new and old, and then being able to transfer this knowledge to similar or different contexts to solve related knowledge or problems. Here, understanding serves as the basis for connection, application as the manifestation of transfer, and relationality synthesizes the transfer state exhibited by understanding and application. What is described here is not a general knowledge objective detached from specific contexts, such as “students can use Ohm’s law to calculate voltage under specified current (in amperes) and resistance (in ohms) conditions”[32], but rather a cognitive complex demonstrated in response to a specific task or problem, such as “students can use Ohm’s law to solve practical problems of voltage calculation,” which requires students to understand Ohm’s law and be able to flexibly apply related knowledge in different contexts.
“Critique and creation” involves analyzing, commenting on, and synthesizing uncertain knowledge or open-ended problems (poorly structured problems), forming new understandings or reconstructing new knowledge through inductive or deductive abstract forms. It is a continuous system thinking process, starting from analysis and generating or constructing new schemas through activities such as verification, reasoning, argumentation, judgment, and reflection, producing creative solutions, innovative works, etc., not merely seeking correct answers or certain knowledge to problems.
3. Deep Learning Layering in Comprehensive Objective Classification
(1) Hornstein’s Four Domains Deep Learning Levels
Hornstein was dissatisfied with Bloom’s approach of separately developing educational objectives for the cognitive, affective, and psychomotor domains, believing that “(the three domains) as independent entities lack a unified context, have different quantities of categories, and their classification intentions are not parallel, with varying terminological expressions within categories and across categories as well as subcategories,” and “have not formed a comprehensive behavioral domain to promote the development goals of a complete person”[33], necessitating a comprehensive classification that reflects the combination of cognitive, affective, and psychomotor domain objectives. To this end, Hornstein introduced a systems perspective of teaching, constructing a conceptual framework consisting of inputs, processes, outputs, and feedback based on human integrity.[34] In this framework, “input” refers to information or content (external or others’ knowledge) that students need to master in learning a subject, including symbolic, descriptive, prescriptive, and technical information. “Process” refers to the learning experiences that learners undergo, manifesting the fulfillment of cognitive, affective, psychomotor objectives, or integrating them into behavioral domain objectives. “Output” refers to the results of learning, making learners knowledgeable, cultured, and capable complete individuals. “Feedback” monitors results, determining the differences between expected and actual achievement levels, and adjusts information (content) and objectives. In this framework, educational objectives are redefined as the learning process or experience, where the cognitive, affective, and psychomotor domains can be used independently or combined into a behavioral domain.[35] Each domain has five levels of interrelated objectives (see Table 1), while Bloom’s cognitive objectives classification’s “knowledge” is transformed into four types of information or content in the “input.”
Table 1 Hornstein’s Educational Objectives Classification[36]
Analyzing the specific meanings of the verbs listed in each of the five levels in each domain, the cognitive domain’s “conceptualization” and “comprehension” have not reached the requirements of deep learning; these two levels are not much different from the first two levels of Bloom’s cognitive objectives. Only at the “application” level does it emphasize “being able to clarify problems or situations and use appropriate principles and procedures to solve specific problems or situations”[37], thus entering deep learning. For the affective and psychomotor domains, most studies on deep learning emphasize students’ active participation and full engagement rather than passive responses, simple reactions, or imitation. In the affective domain, “acceptance” requires students to be willing to accept and pay attention to certain stimuli, representing the lowest requirement of interest and possessing considerable passivity; “reaction” refers to students responding to certain stimuli and being able to evaluate their feelings; “value judgment” refers to expressing an opinion or view on the value of certain things; “commitment” refers to trusting and committing to a certain value and regarding it as a guiding principle; “self-expression” refers to demonstrating and adapting behavior according to values and beliefs.[38] In the psychomotor domain, “perception” refers to the ability to accept and recognize stimuli related to specific concepts, ideas, and phenomena; “imitation” refers to activating, emulating, and coordinating natural potentials, thereby shaping behavior consistent with general patterns or situations; “generation” refers to integrating ability tendencies and behavioral performances, applying the acquired qualities and characteristics to skill recognition; “creation” refers to effectively maintaining and adjusting related skills to achieve predetermined functions; “mastery” refers to innovating and perfecting various abilities.[39] In terms of the respective connotations of the five levels and the specific details they include, the affective domain’s “value judgment,” “commitment,” and “self-expression,” along with the psychomotor domain’s “generation,” “creation,” and “mastery,” exhibit characteristics of deep learning in the self domain. If the three domains are integrated into a behavioral domain (which is Hornstein’s creatively comprehensive objective type), “adaptation, performance, and aspiration” become the three levels of deep learning performance.
(2) Marzano’s Three Systems Deep Learning Levels
Marzano addresses the theoretical dilemmas of Bloom’s educational objectives taxonomy, stating, “From a logical or empirical perspective, Bloom’s taxonomy merely provides a simple, linear behaviorist model, whose hierarchical structure has not been well integrated”[40], proposing a learning behavior model involving three intellectual process systems based on knowledge. Consequently, he constructed a two-dimensional objectives classification framework. One dimension represents three domains of knowledge: information, mental procedures, and psychomotor procedures; the other dimension represents three systems of intellectual processes with six levels, namely, retrieval (information extraction), understanding, analysis, knowledge application in the cognitive system, as well as levels in the metacognitive system and self-thinking system. In summary, educational objectives are the six levels that intellectual processes aim to achieve across the three knowledge domains, emphasizing the integration of cognition and affect, transcending Bloom’s unidimensional framework and unidirectional focus. Each level encompasses different types of performance, as detailed in Table 2. These six levels are distinguished based on processing flow and levels of awareness, differing from Bloom’s and Anderson’s orders based on cognitive complexity. In Marzano’s view, self-thinking is the first procedure, determining students’ motivation to learn a certain knowledge. Once self-thinking has established the importance of knowledge, the next system must be activated: metacognition, whose task is to clarify learning objectives related to knowledge and formulate learning plans. Finally, the cognitive system operates, primarily responsible for processing information, from simple extraction to complex knowledge application.[41] In this process, the cognitive system’s level of awareness is the lowest, followed by the metacognitive system, while the self-thinking system has the highest level of awareness. Within the cognitive system’s four levels, “retrieval” remains at the unconscious automatic level, while the other three awareness levels progressively elevate. Therefore, level 1 in Table 2 has not met the requirements of deep learning, while level 2, which “is responsible for converting knowledge into appropriate forms for storage in long-term memory”[42], does not differ fundamentally from Bloom’s “understanding” level, nor does it belong to the realm of deep learning. Level 3 onward begins to enter the perspective of deep learning. Among them, levels 5 and 6 integrate cognitive and affective elements, emphasizing holistic learning with full engagement, achieving higher levels of deep learning. From the perspective of processing flow, both metacognition and self-thinking become important foundations for deep learning, giving rise to two levels of performance: analysis and knowledge application. In this sense, the deep learning in Marzano’s educational objectives classification resembles a four-layered interconnected complex.
Table 2 Marzano’s Objectives Classification and Its Psychomotor Processes[43]
(3) Deep Learning Levels in Comprehensive Objective Classification: An Integrative Perspective
Hornstein’s deep learning levels across four domains and Marzano’s six levels of deep learning are not entirely the same, but a careful analysis of their specific directions reveals many similarities. If we reference Hornstein’s behavioral domain, the goal of “adaptation” is that “students can proficiently use their knowledge and abilities to solve problems in real or simulated situations similar to or different from those they have initially encountered”[44], while Marzano’s levels of “analysis” and “knowledge application” aim to resolve specific problems by categorizing, summarizing, and exploring information, mental procedures, and psychomotor procedures. It can be said that “analysis” and “knowledge application” are equivalent to the “application” level of “adaptation,” but lack the emotional experiences found in “analysis” and “knowledge application.” “Performance” refers to “evaluating different situations and taking practical actions, that is, analyzing, verifying, evaluating, and integrating knowledge, values, and beliefs according to the situation and taking action”[45]. This shares many similarities with metacognition, as “metacognition involves monitoring, evaluating, and regulating different types of thinking, responsible for executive control”[46], and both can be categorized into a single category or level. “Aspiration” encompasses the values and beliefs expressed in knowledge, skills, and emotions, which lead individuals to mastery, excellence, perfection, and success, and solve complex problems or situations, involving students obtaining higher-level practical knowledge, enhanced skills, values, and expertise, as well as greater sensitivity, artistry, creativity, and wisdom, leading to rational judgments and decisions.[47] Self-thinking comprises attitudes, beliefs, and emotions, which are interrelated and determine the intensity of motivation and attention. Specifically, self-thinking determines whether to engage in a specific task and how much effort to invest in completing that task.[48] It is evident that both “aspiration” and “self-system” emphasize the importance of beliefs, emotions, and values in task decision-making, knowledge construction, and cognitive development, requiring students to refine these non-cognitive factors in task learning and better utilize them to complete tasks.
Due to the use of different terminologies to express fundamentally the same deep learning levels in the two classifications, we attempt to integrate them, distinguishing three levels of deep learning objectives: adapting to contexts, analyzing and applying knowledge; executing monitoring, constructing self-awareness; maintaining beliefs, demonstrating self. These three levels present different relationships between knowledge and self in deep learning: the first level focuses on knowledge mastery and application, accompanied by positive self-experiences such as patience, responsibility, perseverance, and cooperation, adapting to different knowledge contexts to solve definite or uncertain problems. The second level involves monitoring, evaluating, and regulating the knowledge learning process and outcomes, achieving self, forming new cognitions and value concepts, and emotional responses. The third level centers on self, confidently completing challenging tasks or complex problems, gaining higher levels of wisdom, values, and achievement experiences. The classification of these three levels can be seen as a certain analysis of the connotation of deep learning from the perspective of competencies, and it is also a graded explanation of the deep learning competencies in the cognitive and self domains.
4. The Significance of Deep Learning Layering for Classroom Evaluation
Clarifying the different levels of deep learning based on educational objectives taxonomy is of great guiding significance for teachers conducting classroom evaluations. Why focus on the significance of deep learning objective layering for classroom evaluation rather than for teaching or other aspects? This primarily stems from three considerations. First, the most direct use of educational objectives taxonomy is to guide test preparation and classroom evaluation, and the deep learning objective levels derived from it should prioritize how to be applied in evaluation. Second, classroom evaluation has both instrumental and liberating significance for teaching.[49] Prioritizing the role of classroom evaluation can better play its catalytic role in facilitating the transformation of teaching and learning. There is an increasingly accepted view that “teachers should learn to evaluate before learning to teach”[50]. Clarifying the significance of classroom evaluation provides clearer direction for “teaching for deep learning.” Third, it clarifies the objects of classroom evaluation that promote learning. Since the 1990s, evaluations that promote learning have become a new paradigm widely discussed and practiced, yet what kind of learning is promoted remains unclear. Now that the hierarchical objectives of deep learning are clarified, it can provide a foothold for the specific objects promoted by evaluation. As an informal evaluation, classroom evaluation accompanies the entire learning process, aiming to promote students’ genuine learning, rather than simply distinguishing and judging learning outcomes and quality. From the perspective of promoting deep learning, the layering of deep learning provides a basis for the objectives, methods, tasks, and techniques of classroom evaluation.
(1) Clarifying the Direction and Levels of Evaluation Objectives
The theory of classroom evaluation holds that clear learning objectives are the foundation of high-quality evaluation. Once the deep learning objectives and their levels are clarified, teachers and students can design and practice evaluations based on this. Of course, teachers need to understand the deep learning objectives in different domains, distinguish the differences between different levels, and clarify the actual performances expected from students to achieve the objectives. Deep learning objectives in the cognitive domain focus on the development of higher-order thinking, while deep learning objectives in the competency domain emphasize the integration of self, metacognition, and knowledge. It is important to note that whether cognitive objectives or comprehensive objectives emphasize the integrity of the same level, it is not a decomposition and combination of objectives at the same level; for example, if the target to be achieved is “critique and creation,” it should be designed according to holistic thinking, rather than decomposing it into elements like “analysis, evaluation, creation, and abstraction” and then stacking them. In other words, deep learning’s hierarchical objectives should not be presented in a fragmented manner but should focus on the core of the objectives and be narrated comprehensively.
(2) Facilitating the Selection of Evaluation Methods Suitable for Deep Learning Objectives
Based on the analysis of deep learning objectives in the two domains, the methods used for classroom evaluation for deep learning should differ from traditional paper-and-pencil exercises, exams, and simple classroom questioning under the measurement perspective, emphasizing situationality, authenticity, and process, such as performance evaluation, portfolio assessment, dialogical evaluation, self-evaluation, and peer evaluation. Of course, it is not to apply all these methods to a certain type of objective but to select evaluation methods that are compatible with deep learning objectives based on practicality. For the cognitive domain’s “connection and transfer,” dialogical evaluation and performance evaluation are more suitable. Dialogical evaluation fosters deeper dialogues between teachers and students, helping students better connect new and old knowledge and construct knowledge maps; performance evaluation designs challenging tasks that guide students to apply and transfer knowledge maps to task situations. The “critique and creation” level should primarily use performance evaluation and self-evaluation, while also considering portfolio assessment and peer evaluation, inspiring students’ critical and self-reflective awareness through portfolios that cumulatively present students’ learning outcomes. For the three levels of comprehensive objectives, self-evaluation should be consistent throughout, as “self-evaluation helps improve learning motivation, participation, and efficacy”[51]. On this basis, the first level should primarily use dialogical evaluation, supplemented by performance evaluation; the second level can comprehensively employ performance evaluation and dialogical evaluation; the third level should utilize more portfolio evaluation and performance evaluation.
(3) Aiding the Design of Evaluation Tasks that Stimulate Deep Learning
Once the objectives and methods are clear, the design of tasks to achieve these objectives becomes crucial. Generally speaking, the evaluation tasks designed should enable students to demonstrate the knowledge or skills they have mastered. For higher-order thinking, the materials used in tasks should be guiding and novel.[52] Guiding means that the materials provided can provoke and stimulate students’ thinking; novelty refers to the newness of materials, which have not been used in classroom teaching and possess a certain level of difficulty. From the specific manifestations of the two cognitive levels of deep learning and the three comprehensive levels, in addition to guiding and novelty, evaluation tasks should also emphasize situationality and openness, strengthening the connection with students’ life worlds, prompting students to explore new knowledge amid uncertainty and enhance self-reflective awareness. For the two levels in the cognitive domain, “connection and transfer” focuses on designing knowledge tasks that require guidance from materials, providing some challenging materials, allowing students to make predictable constructive responses in relevant situations. The “critique and creation” level emphasizes designing completely open practical tasks that are fully open-ended, focusing on examining the depth of thinking and the innovation of outcomes. For the three levels of comprehensive objectives, the first level focuses on knowledge tasks that require guidance from new materials, designing situations that integrate and reflect children’s lives. The second level focuses on complex tasks that require the participation of reflective awareness, designing situations that are abstract or historical worlds above real life. The third level emphasizes long-term, high-difficulty tasks that require willpower, with situations designed to focus on the imaginative world that requires deep self-involvement.
(4) Assisting in Developing Evaluation Techniques to Observe the Extent of Deep Learning
The ultimate goal of classroom evaluation is to promote deep learning, but it is also necessary to understand the extent of deep learning achieved by students during the process to better monitor and adjust the learning process. In both the cognitive and comprehensive domains, the clear hierarchical division of deep learning provides a reference point for developing evaluation techniques to observe learning achievement. These reference points mainly materialize in evaluation tasks targeting learning objectives, and the techniques to be developed will also be task-specific. Since evaluation tasks for deep learning are not simply about obtaining a correct answer, they often require gathering evidence in knowledge, methods, thinking, emotions, etc., to make judgments; thus, the evaluation techniques employed primarily involve evaluation criteria. Evaluation criteria, also known as “scoring rubrics,” serve as guidelines, standards, or principles for assessing the quality of students’ responses, outcomes, or performances. Specifically, they break down a task into several components and provide detailed quality descriptions of acceptable or unacceptable performance for each component. Generally, evaluation criteria will describe the different performances of the deep learning objectives that a task aims to achieve, at least dividing them into three levels to distinguish the overall performance of different levels of deep learning or the performances of each level at different deep learning points, such as the emotional points, analytical points, and knowledge application points of the first level of comprehensive objectives. Developing such evaluation criteria should involve students’ right to know or participation, with teachers inviting students to discuss and share results together. In this way, evaluation criteria not only become tools for teachers to evaluate tasks but also scaffolds for students’ self-evaluation, helping students clarify what extent of deep learning they have achieved and what efforts are needed to reach higher levels.
References:
Long press to follow for more exciting content
Global Education Outlook
Submission website:
http://www.kcs.ecnu.edu.cn/globale