Boundaries, a Research Agenda, and Conditions for Revising the Framework

A complete structure is not evidence of an effect. The preceding articles establish a contestable conceptual scheme: autonomous human development, four dimensions, a learning system, one design cycle, and distinct responsibilities at four scales. They report no primary data. This article does three things. First, it sorts the sentences of the series into definitions, normative claims, and hypotheses, so that sentences about effects are not upgraded in retelling. Second, it states one core research question and five groups of issues: selfhood, views of knowledge, and agency; pathways and long-term capabilities; artificial intelligence and cognitive autonomy; family, school, and society; assessment and governance. Validation is layered—design research, then comparison, then experiment or quasi-experiment, then transferability—and the layers are not skipped. Third, it states the ethical floor once: human purpose and responsibility; data minimization and authorization for minors; growth data are not an object of trade; no fixed labels from a single performance or an algorithmic portrait; human review of consequential judgments; a minimum version under scarce resources; no anxiety marketing, no promises of unproven outcomes, and no credential barriers without a research basis; limits and failures are reported. Whether unequal resources are amplified remains to be tested. The minimum-version principle is applied without waiting for that test. Revision is open to argument and to counterexamples. The text changes when the evidence changes. Parallel terms are not added in order to avoid a counterexample. The pace of extension follows the honesty of the evidence and whether others can use the method. It does not follow the scale of dissemination. Skill reports take the secondary skill as the unit and present a skill profile. Secondary skills are not combined into a composite grade for a primary skill. Research does not treat such a composite grade as a finding. The propositions of this article are the sentences the series has already written. 中文摘要 结构完整不等于效果已经得到支持。前四篇确立的是一套可以争论的概念:人的自主发展、四个维度、学习系统、一个设计循环和四个尺度上的不同责任。它们没有报告原始数据。本文做三件事。第一,把系列中的句子分成定义、规范性主张和有待检验的命题,并列出需要研究加以处理的问题,以免效果句在转述中被升格。第二,给出一个核心研究问题,以及五组研究议题:自我性、知识观与能动性;学习路径与长期能力;人工智能与认知自主;家庭、学校与社会;评价与治理。验证按设计研究、比较、实验或准实验、可迁移性分层,不跳步。第三,把伦理底线集中陈述一次:人的目的与责任、未成年人数据的最小必要、成长数据不是交易对象、不用一次表现或算法画像写下固定标签、重要判断保留人工复核、低资源下仍有最小版本、不贩卖焦虑也不承诺未证明的效果、如实报告限制和失败。资源差距是否被放大,仍待检验;最小版本的原则不等待该项检验完成才执行。修订对论证和反例开放。证据改变时修改文本。不靠增加平行术语来回避反例。扩展的速度服从证据是否诚实、方法能否被他人使用,不服从传播规模。技能报告以二级技能为单位,给出技能剖面,不合成一级综合等级。研究不把一级综合分当作发现。本文的命题就是各篇已经写出的句子。 The English PDF is the primary file. The Chinese PDF is a parallel version of the same text (zho).