自然实体锚点式跨学科教学：理论模型检验与机制探索 基于“多彩大自然”准实验的混合方法研究

摘要：目的 检验自然实体锚点式跨学科教学（NEAI）模型对学生环境素养的提升效果，并系统探究其“情感-认知-行为”作用机制。方法 采用整群准实验设计，在乌鲁木齐市场域A与场域B两处选取120名六年级学生，通过盲态分组工具将学生平衡分配至实验组（n=60）与对照组（n=60）。实验组由一位教师实施8周NEAI课程，对照组由另一位教师按常规教学进行。使用青少年环境素养量表、跨学科能力问卷、环境行为核查表（双盲家长报告）收集量化数据，辅以30例分层抽样访谈与120份手抄报质性分析。结果 协方差分析显示，NEAI显著提升学生环境素养（调整后均值差=4.1分，95%CI [3.2, 4.9], F(1,117)=86.42, p<0.001, BF₁₀>1000），效应量Cohen's d=2.38。行为转化率达71%。延迟后测未发现显著衰减（T2 vs T1: p=0.38, TOST等价边界[-0.2, 0.2]）。贝叶斯中介分析发现，情感对行为的直接效应BF₁₀=11.2（强证据支持有直接效应），而“情感→认知→行为”链式中介效应BF₀₁=4.7（中等证据支持无中介）。多层模型分析表明，场域类型与干预无显著交互作用（γ=0.12, p=0.43）。结论 NEAI模型能有效提升学生环境素养，其作用机制主要表现为情感的直接驱动。鉴于小样本与单场点限制，该效应值需在多校复制中校正，机制结论仅提供初步提示，有待大样本验证。

Abstract: Objective To examine the effect of the Natural Entity Anchored Interdisciplinary Teaching (NEAI) model on improving students’ environmental literacy, and systematically explore its "affective-cognitive-behavioral" mechanism.
Method: A cluster quasi-experimental design was adopted. A total of 120 sixth-grade students were selected from two field sites (Site A and Site B) in Urumqi, and balancedly allocated to the experimental group (n=60) and control group (n=60) via a blinded grouping tool. The experimental group received 8-week NEAI courses delivered by one teacher, while the control group received conventional teaching conducted by another teacher. Quantitative data were collected using the Adolescent Environmental Literacy Scale, Interdisciplinary Competence Questionnaire, and Environmental Behavior Checklist (double-blinded parent report). Qualitative data were supplemented by 30 stratified sampling interviews and qualitative analysis of 120 hand-copied posters.
Results: Analysis of covariance (ANCOVA) showed that NEAI significantly improved students’ environmental literacy (adjusted mean difference = 4.1 points, 95% CI [3.2, 4.9], F(1,117)=86.42, p<0.001, BF₁₀>1000), with a Cohen's d effect size of 2.38. The behavior conversion rate reached 71%. No significant attenuation was found in the delayed post-test (T2 vs T1: p=0.38, TOST equivalence boundary [-0.2, 0.2]). Bayesian mediation analysis revealed that the direct effect of affect on behavior had a Bayes factor BF₁₀=11.2 (strong evidence supporting the presence of a direct effect), while the chain mediation effect of "affect → cognition → behavior" had a BF₀₁=4.7 (moderate evidence supporting the absence of mediation). Multilevel model analysis indicated no significant interaction between field type and intervention (γ=0.12, p=0.43).
Conclusion: The NEAI model can effectively improve students’ environmental literacy, and its underlying mechanism is mainly characterized by the direct driving effect of affect. Given the limitations of small sample size and single-site constraint, the obtained effect size needs to be calibrated in multi-school replication studies, and the mechanistic conclusions only provide preliminary hints that require validation with large sample sizes.