The Intellectual Sources and Contemporary Meaning of Cognitive Autonomy

The first four commentaries discussed, in turn, the historical significance of AI general education, the changing task of education in the age of intelligence, the formation and defining features of the intelligent society, and the concept of cognitive autonomy. One thread runs through all of them: as social structures and processes grow ever more complex, humans keep handing part of their cognitive activity to external tools and systems; AI goes further and lets machines take part in posing problems, organizing knowledge, reasoning, judging and arranging action. Under these conditions, how people keep the lead over their own cognitive activity becomes a fundamental question that education must answer.

The answer given by the Redbook "Artificial Intelligence General Education" is cognitive autonomy. The Redbook defines cognitive autonomy as "the process by which an individual, in a complex cognitive environment, actively regulates their own belief formation, use of evidence, allocation of trust, boundaries of judgment, and responsibility for action." Correspondingly, the capacity for cognitive autonomy is the individual's ability to initiate, sustain and adjust this process so that it can continue across different tasks and situations ^[1].

Cognitive autonomy has a deep intellectual background. Reflection, the demand for evidence, rational autonomy, self-regulation and the scrutiny of epistemic authority have long been present in philosophy and educational research. These traditions intersect and hold different theoretical positions, and together they provide intellectual resources that cognitive autonomy can inherit and develop.

1. Reflection and the Demand for Evidence: Early Sources of Cognitive Autonomy

How people take responsibility for their own cognition is an old question. Knowledge may come from tradition, classics, teachers and social authorities, but whether accepting it requires reasons, and whether the recipient should examine those reasons, has always been bound up with a person's place in the activity of knowing.

In Plato's Apology, Socrates reflects on how his wisdom differs from that of others: about things he does not know, he is able to admit that he does not know ^[2]. This self-examination reveals a basic requirement of the cognitive process: the knowing subject must distinguish what they actually know from what they merely think they know, and keep their judgments open to continued examination and testing. It became an important intellectual resource for later discussions of reflection, the boundaries of judgment and intellectual humility.

Chinese thought also has a tradition of reflecting on and scrutinizing knowledge and authority. Mencius said that "to believe everything in the Book of Documents would be worse than having no Book at all," and he critically examined records in the Shangshu ^[3]. This embodies a reflective, enlightening attitude toward knowledge: a classic's standing is no reason to accept it wholesale; the recipient must make their own judgment.

A long historical distance separates these ideas from today's cognitive autonomy, yet they share the same question: when knowledge comes from outside oneself, does one still bear the responsibility to reflect and to seek evidence? These reflections provide early intellectual resources for understanding cognitive autonomy.

2. Rational Autonomy: The Human Being as Cognitive Subject

In Enlightenment thought, autonomous cognition became more closely tied to the standing of the human being as a subject. In his 1784 essay "An Answer to the Question: What Is Enlightenment?", Kant made the courage to use one's own understanding a central aim of enlightenment. The difficulties he attended to included people's habitual dependence on the guidance of others, and the social conditions needed to escape that dependence, above all the freedom to make public use of one's reason ^[4].

This discussion gave autonomous cognition a clearer social meaning. People should be able to examine the claims they accept, understand the reasons that support them, and take part in public discussion of those reasons. Epistemic authorities can help, but the individual's responsibility for judgment must be preserved. Autonomy and responsibility, reason and public life were thereby linked.

This thought provides an important value foundation for cognitive autonomy: people need to become subjects who can use knowledge, revise their judgments and take responsibility for their actions. At the same time it leaves an educational question: how can education cultivate rationally autonomous citizens?

3. Modern Educational Research: The Scholarly Background of Cognitive Autonomy

Since the twentieth century, education and psychology have studied human judgment, reflection and self-regulation from different directions. Together these studies form the scholarly background of "cognitive autonomy" in the Redbook.

The theory of reflective thinking offers a practical path for cultivating independent thought. Dewey's theory of reflective thinking stresses that cognitive autonomy must be realized as the actual activity of forming and revising judgments. Learners act, experience the consequences and adjust what they do next through reflection; these are necessary processes in developing cognitive autonomy, and educational activity should be organized accordingly ^[5].

Research on critical thinking attends to the evidential basis of judgment. The expert consensus study led by Facione characterized critical thinking as purposeful, self-regulatory judgment involving interpretation, analysis, evaluation and inference, and also emphasized an open, prudent attitude and a disposition to actively revise one's understanding ^[6].

Research on metacognition attends to how people know and regulate their own knowing. Flavell discussed cognition about cognition, making people's understanding of themselves an object that can be studied and cultivated. It concerns how people assess their degree of understanding, the effectiveness of their strategies and the limits of their cognition, and adjust accordingly ^[7].

Research on self-regulated learning attends to how learners actively organize their learning. Zimmerman emphasized that, in a cycle of forethought, performance and reflection, learners must continually regulate their goals, strategies, motivation and behaviour, and also actively seek help and draw on resources in their environment ^[8].

Research on epistemic cognition attends to how people understand knowledge and the grounds on which it holds. Hofer and Pintrich discussed people's views about the sources of knowledge and the grounds of its justification ^[9].

Research on information literacy and media literacy attends to discernment and participation in the information environment. From Zurkowski's 1974 discussion of information literacy to contemporary information and media literacy frameworks, this field has addressed information sources, judgments of authority, ethical participation and media power ^[10–12].

Each of these lines of research has its own emphasis, and they also overlap. Reflective thinking stresses forming and revising judgments through inquiry; critical thinking stresses the scrutiny of evidence; metacognition supports self-monitoring; self-regulated learning supports the organization of activity; epistemic cognition supports judgments about the grounds of knowledge; and information and media literacy support discernment of the cognitive environment. The Redbook brings these resources together under a single question: in an environment made up jointly of people, institutions and intelligent systems, how does an individual continually take part in, and actively regulate, their own process of knowing? That is cognitive autonomy ^[1].

4. The Intelligent Society: Cognitive Autonomy Becomes the Fundamental Goal of Education

If discussion of cognitive autonomy has a long history, why should cognitive autonomy become the fundamental educational goal of the intelligent society? The root reason is that the relationship between dependence and autonomy has changed structurally in the intelligent society.

In earlier commentaries we discussed how, over humanity's long development, outsourcing part of our cognitive capacities to tools and systems can raise individual productivity and also make society run more efficiently ^[1,13]. Yet when necessary cognitive practice is replaced over long periods, the formation and maintenance of the corresponding capacities may be affected ^[14]. For a long time, the tension between dependence and autonomy appeared mainly as a problem of particular tools and specific tasks.

The arrival of the intelligent society widens this tension. As AI gradually enters study, work and public life and individuals use it to connect to larger systems of knowledge and collaboration, dependence on external intelligence will become a pervasive and lasting condition of life. AI is beginning to replace human cognitive activity in a wider range of settings; if necessary cognitive practice keeps shrinking as a result, substantial cognitive loss may follow ^[15].

Two situations must be distinguished here. For people who are still learning, the problem may show up as capacities that should have formed never being adequately practised and developed; for people who already have some capacity, it may show up as existing capacities going unused and unmaintained. Both can be masked by the appearance of tasks being completed smoothly: the quality of the output improves, while people's actual understanding and judgment do not necessarily improve with it ^[14,15].

It should be noted that this tendency has social drivers. Demands to save time, cut costs and raise output continually encourage handing more steps to systems. Even individuals willing to think may face pressure for efficiency.

The reduction of cognitive practice under the pressure of efficiency may therefore become an important factor in the erosion of cognitive autonomy.

It is precisely here that cognitive autonomy acquires its fundamental place in education. Autonomy, reflection and a critical spirit have always been important aims of education; the intelligent society goes further and makes maintaining the capacities of the subject itself a task that must be explicitly organized and continuously safeguarded. If education judges its success mainly by the quality and speed of immediate output, it may overlook learners' actual participation in the cognitive process, and so weaken the foundation in the subject on which its other goals depend.

Cognitive autonomy should therefore become the fundamental goal that integrates the content, methods and assessment of AI general education. Knowledge learning, tool use and task training all need to answer the same question: do these activities strengthen learners' capacity to pose questions, understand grounds, calibrate trust, revise judgments and take responsibility? Education needs to make clear which cognitive activities learners must experience for themselves, and to discuss what kinds of first-hand experience help cognitive capacities to form ^[1].

5. Related Research and the Redbook's Theoretical Advance

"Cognitive autonomy" appeared in existing research before the Redbook. Identifying these usages accurately helps to show what the Redbook inherits and where it advances the theory.

In their 2004 study of autonomy support in the classroom, Stefanou and colleagues distinguished organizational, procedural and cognitive autonomy support, with cognitive autonomy support encouraging students to explain their reasoning, compare solutions and evaluate their own work ^[16]. This already touches on how education can support students' independent thinking. In a 2007 study, Beckert used the concept of "cognitive autonomy" to carry out a conceptual analysis and develop a measurement instrument covering adolescents' evaluation of thinking, voicing of opinions, decision-making and self-evaluation ^[17]. These studies show that autonomy at the cognitive level already has a clear basis in education and developmental psychology.

In discussions of AI, research has also used the related concept of "epistemic autonomy". Harris's 2024 study of AI deepfake detectors discussed the relationship between reliance on technological judgment and people's epistemic autonomy and self-trust ^[18]. This term differs from the specific definition of "cognitive autonomy", but it shows that the way reliance on AI may change people's epistemic standing has already become an explicit subject of discussion.

The Redbook's theoretical contribution is to place cognitive autonomy against the background of the formation of the intelligent society as a whole, focusing on how individuals, once external intelligence intervenes continually, actively regulate the relations among belief, evidence, trust, boundaries and responsibility. More importantly, it establishes this active regulation as the overarching goal of AI general education for all citizens.

Under this goal, the Redbook further links conceptual definition, capacity building and educational implementation: it characterizes the process of cognitive autonomy through six analytical dimensions, organizes educational training around four capacities (induction, generalization, judgment, and self-awareness), and on that basis discusses knowledge content, curriculum systems and developmental progression. Its advance takes the form of an educational theory and practical framework for the intelligent society, giving the ideas of autonomy, reflection and self-regulation a shared organizing goal and a path to realization under new social conditions.

References

1. [1] Wang D, Ma S, eds. Artificial Intelligence General Education: Cognitive Autonomy in an Intelligent Society[R/OL]. v2.3.3 (draft for comments). Center for Artificial Intelligence General Education (AIGE), Department of Computer Science and Technology, Tsinghua University, 2026-08 [2026-09-26]. https://aige.cs.tsinghua.edu.cn/redbook/
2. [2] Plato. Apology[M/OL]. Jowett B, trans. 21b–23b (Stephanus pagination) [2026-09-26]. https://classics.mit.edu/Plato/apology.html
3. [3] Mencius. Mencius, Jin Xin II[EB/OL]. Chapter 3 [2026-09-26]. https://zh.wikisource.org/wiki/%E5%AD%9F%E5%AD%90/%E7%9B%A1%E5%BF%83%E4%B8%8B
4. [4] Kant I. An Answer to the Question: What Is Enlightenment?[EB/OL]. 1784 [2026-09-26]. https://sourcebooks.web.fordham.edu/mod/kant-whatis.asp
5. [5] Dewey J. How We Think[M/OL]. Boston: D. C. Heath & Co., 1910. https://www.gutenberg.org/ebooks/37423
6. [6] Facione P A. Critical Thinking: A Statement of Expert Consensus for Purposes of Educational Assessment and Instruction. Research Findings and Recommendations[R/OL]. 1990. ERIC ED315423. https://files.eric.ed.gov/fulltext/ED315423.pdf
7. [7] Flavell J H. Metacognition and Cognitive Monitoring: A New Area of Cognitive-Developmental Inquiry[J]. American Psychologist, 1979, 34(10): 906–911. https://doi.org/10.1037/0003-066X.34.10.906
8. [8] Zimmerman B J. Becoming a Self-Regulated Learner: An Overview[J]. Theory Into Practice, 2002, 41(2): 64–70. https://doi.org/10.1207/s15430421tip4102_2
9. [9] Hofer B K, Pintrich P R. The Development of Epistemological Theories: Beliefs About Knowledge and Knowing and Their Relation to Learning[J]. Review of Educational Research, 1997, 67(1): 88–140. https://doi.org/10.3102/00346543067001088
10. [10] Zurkowski P G. The Information Service Environment Relationships and Priorities: Related Paper No. 5[R/OL]. Washington, DC: National Commission on Libraries and Information Science, 1974. ERIC ED100391. https://files.eric.ed.gov/fulltext/ED100391.pdf
11. [11] Association of College and Research Libraries. Framework for Information Literacy for Higher Education[EB/OL]. 2016-01-11 [2026-09-26]. https://www.ala.org/acrl/standards/ilframework
12. [12] Grizzle A, Wilson C, Gordon D, eds. Media and Information Literate Citizens: Think Critically, Click Wisely![M/OL]. 2nd ed. Paris: UNESCO, 2021. https://unesdoc.unesco.org/ark:/48223/pf0000377068_eng
13. [13] Risko E F, Gilbert S J. Cognitive Offloading[J]. Trends in Cognitive Sciences, 2016, 20(9): 676–688. https://doi.org/10.1016/j.tics.2016.07.002
14. [14] Bainbridge L. Ironies of Automation[J]. Automatica, 1983, 19(6): 775–779. https://doi.org/10.1016/0005-1098(83)90046-8
15. [15] Bastani H, Bastani O, Sungu A, et al. Generative AI without Guardrails Can Harm Learning: Evidence from High School Mathematics[J]. Proceedings of the National Academy of Sciences, 2025, 122(26): e2422633122. https://doi.org/10.1073/pnas.2422633122
16. [16] Stefanou C R, Perencevich K C, DiCintio M, et al. Supporting Autonomy in the Classroom: Ways Teachers Encourage Student Decision Making and Ownership[J]. Educational Psychologist, 2004, 39(2): 97–110. https://doi.org/10.1207/s15326985ep3902_2
17. [17] Beckert T E. Cognitive Autonomy and Self-Evaluation in Adolescence: A Conceptual Investigation and Instrument Development[J]. North American Journal of Psychology, 2007, 9(3): 579–594. https://www.researchgate.net/publication/282023761_Cognitive_Autonomy_and_Self-Evaluation_in_Adolescence_A_Conceptual_Investigation_and_Instrument_Development
18. [18] Harris K R. AI or Your Lying Eyes: Some Shortcomings of Artificially Intelligent Deepfake Detectors[J]. Philosophy & Technology, 2024, 37: 7. https://doi.org/10.1007/s13347-024-00700-8