设计科学视域下智能心理测量学的本体重构、范式演进与工程化整合
Ontological Shift, Paradigm Evolution, and Engineering Reconstruction of Psychological Assessment in the Age of AI: A Design Science Perspective
郑蝉金 1虞增艺2
作者信息
- 1. 华东师范大学上海智能教育研究院
- 2. 华东师范大学教育学部教育心理系
- 折叠
摘要
传统心理测量学长期受自然科学范式影响，强调实体论意义上的真值探寻。面对智能时代的挑战，探针式测量范式难以充分解释技术对人的重塑作用。基于设计科学思想，心理测量学向工程科学范式拓展的可能性得到论证；通过重构“内部环境—界面—外部环境”的三元架构，心理测量学的本体论转换得以展开。在此基础上，测量被重新界定为连接个体心理黑箱与任务环境的人工界面，界面从生物具身耦合向智能时代人智协同认知界面跨越的演进逻辑得到系统梳理。该视角不仅为化解智能测量的黑箱危机提供本体论依据，也在界面设计层面促进测量学、实验心理学及人因科学的整合，为心理测量学从测评工具转向认知工程系统、推动心理学工学化转型奠定理论基础。
Abstract
Traditional psychological measurement has long been shaped by the natural science paradigm and by an essentialist ontology that seeks to identify the true status of latent psychological variables. In this tradition, reliability, validity, statistical modeling, and error control provide the scientific basis for inferring psychological constructs from observable responses. However, in the age of artificial intelligence, psychological measurement is increasingly embedded in intelligent, interactive, and dynamically adaptive environments. These changes do not simply move psychological measurement away from the natural science paradigm. Rather, they create new explanatory pressures and practical requirements: psychological measurement must continue to address how psychological constructs can be validly inferred, while also explaining how artificial interfaces can be designed to generate stable, interpretable, fair, and ethically responsible construct evidence.Drawing on Herbert Simons theory of the sciences of the artificial, this paper reframes psychological measurement as the design of artificial interfaces that mediate between the inner psychological environment and the outer task environment. On this basis, a triadic architecture of inner environment, interface, and outer environment is proposed as the structural core of this reconstruction. In this framework, measurement tools are not treated as passive probes that merely record pre-existing traits, but as designed artifacts that organize tasks, interaction processes, feedback mechanisms, and evidence generation. The goal of measurement is therefore not replaced by interface adaptation, but extended from error control and construct approximation to include interface fit, process evidence, system transparency, and iterative improvement in intelligent measurement contexts.Building on this ontological reconstruction, the paper delineates four historical stages in the evolution of psychological measurement interfaces: biological-cognitive coupling, symbolic transformation, dynamic simulation, and intelligent evolution. The first stage is grounded in embodied oral and behavioral assessment, where biological and cognitive systems are tightly coupled. The second stage is represented by paper-and-pencil testing and standardized scales, in which psychological activity is transformed into stable symbolic records. The third stage emerges with computerized testing, virtual reality, game-based assessment, stealth assessment, and learning analytics, which enable dynamic simulation and process data collection. The final stage is characterized by human-AI collaborative cognitive interfaces, in which artificial intelligence participates in task construction, feedback generation, scaffolding, and evidence collection.In such contexts, interaction and dynamics are not merely external features of intelligent measurement, but become central mechanisms through which construct evidence is generated. Examples such as adaptive testing, multimodal evaluation, AI-assisted art creativity assessment, and AI-assisted learning diagnosis show that psychological measurement has begun to move from outcome-centered scoring toward process-based evidence construction. A learners response is no longer limited to a final answer or product; it may include explanation quality, strategy revision, feedback uptake, behavioral trajectory, multimodal expression, and changes across rounds of human-AI interaction. These forms of evidence require measurement designers to consider not only statistical model fit, but also whether the interface can elicit the intended psychological processes and sustain meaningful interaction across diverse populations.The paper further discusses the engineering integration of psychological measurement from a design science perspective. Rather than rejecting traditional evidence-centered design, an agile extension, namely Agile ECD 2.0, is proposed. It transforms the linear evidence chain into a continuous and versioned process of construct hypothesis, interface prototype, process evidence, validity feedback, and iterative revision. This framework is particularly relevant for dynamic item banks, intelligent scale development, multimodal diagnostic systems, and human-AI collaborative assessment. Interface fit is also discussed as an important condition for construct validity in intelligent measurement, because task design, interaction rhythm, evidence capture, cognitive load, accessibility, and population fairness all affect whether valid construct evidence can be generated.Ultimately, the design science perspective does not replace the natural science paradigm in psychological measurement. Instead, it provides a complementary framework for measurement situations marked by high contextuality, strong interaction, process orientation, and dynamic adaptation. Natural science continues to provide construct theory, statistical models, and validity evidence, while design science translates these foundations into operable artificial interfaces, iterative measurement processes, and adaptive intelligent systems. Social science further clarifies value commitments, ethical boundaries, and responsibility in the use of measurement results. In this sense, the future development of psychological measurement lies in the coordinated coexistence of natural science, social science, and engineering science, through which psychological measurement can move toward a more integrated system of explanation, regulation, and design.关键词
设计科学/心理测量学/人智协同认知界面/工程心理学/界面设计Key words
design science/psychological assessment/human-AI collaborative cognitive system/engineering psychology/Interface design引用本文复制引用
郑蝉金,虞增艺.设计科学视域下智能心理测量学的本体重构、范式演进与工程化整合[EB/OL].(2026-07-12)[2026-07-14].https://chinaxiv.org/abs/202607.00109.学科分类
科学、科学研究/计算技术、计算机技术
