摘要
面对深度学习成果的纷繁以及深度学习本身的复杂,文章以CSSCI和SSCI文献为代表,基于复杂性理论的宏观框架梳理研究现状,并按复杂性理论的基本范式提出研究展望。研究表明:复杂系统的概念、条件、机制、过程、标度等方面构成了复杂性理论的宏观框架和基本范式;既有深度学习界定均源自以往的学习理论和框架,特别是依托布卢姆认知目标分类,但这难以反映复杂学习现象的一切特性;不同理论视角的深度学习内涵,加之对机制理解的差异,导致深度学习机制在内容、形式和表述上较为悬殊;深度学习的优化过程在宏观层面缺乏较一致的原则,在中观层面未能有效结合深度学习机制,在微观层面欠缺较明确的统整架构;深度学习的评价体系分为框架类、维度类和指标类,但均以定性为主且局限于认知范畴,无法定量反馈其余因素的影响;由于评价体系的不完善,深度学习的实证研究结论缺乏足够的教育意义和普适性。因此,“深度学习如何表征(内涵)”“深度学习如何实现(机制、过程)”“深度学习如何评价”是深度学习研究值得继续探索的三个核心问题。基于复杂性理论的基本范式,从学习系统视角将深度学习的内涵重新概念化、深度学习的机制和过程科学化、深度学习的评价多维动态化,是今后深度学习研究颇具价值和意义的深化方向。
Facing the numerous achievements of deep learning and the complexity of deep learning itself,this paper takes CSSCI and SSCI literature as representative,combs the research status based on the macro framework of complexity theory,and puts forward the research prospect according to the basic paradigm of complexity theory.The research shows that the concepts,conditions,mechanisms,processes and scales of complex system constitute the macro framework and basic paradigm of complexity theory;the existing definitions of deep learning are derived from the past theories and frameworks of learning,especially from Bloom's Taxonomy of educational objectives,but this can hardly reflect all the characteristics of complex learning phenomena;the connotations of deep learning from different theoretical perspectives and the differences in understanding of mechanisms lead to great differences in the content,form and expression of the deep learning mechanism;the optimization process of deep learning lacks unanimous principle at the macro level,fails to effectively integrate the deep learning mechanism at the middle level,and lacks a clear integrated framework at the micro level;the evaluation system of deep learning is classified into three categories:framework,dimension and indicator,but they are mainly qualitative and limited in cognitive domain,which is difficult to quantitatively reflect the influence of other factors;due to the imperfect of evaluation system,the empirical research conclusions of deep learning lack sufficient educational significance and universality.Therefore,uhow to describe deep learning(connotation)uhow to realize deep learning(mechanism,process),and“how to evaluate deep learning”are three core problems that need to be further studied in deep learning research.Based on the basic paradigm of complexity theory,reconceptualing the connotation of deep learning from the perspective of learning system,making the mechanism and process of deep learning scientific,and making the evaluation of deep learning multidimensional and dynamic from the perspective of learning system are the valuable and meaningful expansion directions of deep learning in the future research.
作者
段茂君
郑鸿颖
DUAN Maojun;ZHENG Hongying(School of Educational Science,Sichuan Normal University,Chengdu,Sichuan,610066,China)
出处
《教育与教学研究》
2022年第12期1-24,共24页
Education and Teaching Research
基金
教育部基金项目“思维可视化视角下高中英语教学模式研究”(编号:17XJA8800009)。
关键词
深度学习
复杂性理论
内容分析
理论范式
学习系统
deep learning
complexity theory
content analysis
theoretical paradigm
leaning system
