**——Why Binary Education Models Are Structurally Incomplete, and How Triadic Inter-Reference Reconstructs Learning**
**Author: Lin Xiaohei**
**Date: June 16, 2026**
---
Current educational models are based on the binary "teacher-student" structure, overlooking a structural necessity: binary inter-reference inevitably produces new blind spots. Based on the structural cognition framework (the four-axiom system), this paper proposes that **the three-person study group is the minimal complete structural unit of education.** One person self-referring leads to recursive collapse. Two people inter-referring produce new configurations but accompanied by new blind spots. Three people form a triadic inter-reference structure — two inter-referrers plus one blind-spot observer — covering all structural functions of cognitive creation and self-correction. One person learning alone has only self-reference; blind spots are covered by no one. Two people discussing exposes some blind spots, but the new understanding produced by their coupling itself contains new blind spots, and both people are inside this new understanding and cannot discover them. With three people, the third occupies the structural position of "observing the inter-reference process" — a position that is empty in binary structures. This paper provides the operational definition of the three-person study group (Role A: Explorer, Role B: Inter-Referrer, Role C: Observer, with role rotation after each round), a comparative experimental design against traditional education models (predicting the three-person group significantly outperforms traditional classrooms and one-on-one tutoring on blind-spot coverage, retention rate, and metacognitive accuracy), and implications for future educational transformation — classrooms shifting from "sitting in rows listening" to "groups of three rotating through inter-reference," and teachers shifting from knowledge transmitters to observers and guides of the inter-reference process.
**Keywords:** three-person study group, minimal complete educational unit, triadic inter-reference, blind-spot coverage, educational structure transformation, role rotation
---
A classroom. Thirty desks. Row by row. All facing forward. At the front, a podium. On the podium, one person. This person is talking. Thirty people are listening.
This scene has repeated globally for two hundred years. It is so ubiquitous that we no longer find it strange. But what is strange is this: **it assumes knowledge can be unidirectionally transmitted from one person to thirty.**
Cognitive science has proven this assumption wrong. People do not learn by "receiving information." People learn through incompatible coupling with existing cognitive structures — new knowledge strikes old configurations; old configurations oscillate; after oscillation ends, either the old configuration persists in modified form or a new configuration is born. This process requires inter-reference — the learner needs another configuration to discover their blind spots.
But even recognizing the necessity of inter-reference, existing "alternatives" — cooperative learning, project-based learning, flipped classrooms — remain at the empirical level. They know "discussion is better than listening to lectures," but do not know **why.** They do not know the structural limitations of two-person discussion. They do not know where the ceiling of "better" lies.
This paper provides the foundational theory. Starting from the structural cognition framework, it derives the minimal complete structural unit of education. The conclusion: **three.**
---
One person faces new knowledge. They read, think, attempt to understand.
This process is essentially self-reference — they use their own existing configuration to strike the configuration of new knowledge. If the new knowledge is compatible with their existing configuration, they absorb it smoothly. If incompatible, their configuration begins to oscillate.
The problem: **they cannot discover their own blind spots.**
Theorem Three stipulates that the self-reference blind spot cannot self-eliminate. When they encounter incompatible knowledge, their reaction is "this doesn't make sense" or "I don't get it." They cannot judge whether the knowledge is wrong or their own configuration has a blind spot. They can only loop within the self-reference closed loop — repeatedly using the same framework to understand new knowledge, repeatedly failing. This is the structural root of "one person going down a rabbit hole."
**One person learning alone = only self-reference. Blind-spot coverage rate: zero.**
Two people discuss. A speaks their understanding. B discovers A's blind spots and points them out. A discovers B's blind spots and points them out. This is inter-reference.
**Two-person discussion solves the core problem of one-person learning — blind spots are seen by another.**
But two-person discussion produces a new problem: **the discussion process itself between A and B forms a new coupled configuration — "our shared understanding." This new configuration has its own blind spots. And both A and B are inside this new configuration and cannot discover its blind spots.**
This is why two-person discussions can go infinitely deep on certain questions yet can never break through certain ceilings. It is not that they are not smart enough. It is that **the structure of binary inter-reference determines that the product of inter-reference necessarily contains blind spots undetectable from within the binary structure.**
The third person C does not participate in the content discussion between A and B. C specifically observes **the inter-reference process of A and B.**
C's input = injecting external perturbation into the coupled configuration of A and B. This perturbation triggers phase transitions that neither A nor B could self-trigger.
**In the triad, two create; the third observes the blind spots of the creative process. This is the minimal complete structure of education.**
---
| Role | Name | Task | Structural Function |
|------|------|------|---------------------|
| A | Explorer | Actively attempts to understand new knowledge / solve new problems | Produces initial configuration |
| B | Inter-Referrer | Observes A's understanding process; points out A's blind spots, errors, omissions | Polar-couples with A |
| C | Observer | Does not participate in content discussion between A and B; specifically observes the inter-reference process of A and B | Covers blind spots of the inter-reference process |
**Role rotation**: after completing each round of the learning task, roles rotate. A becomes B, B becomes C, C becomes A. Ensures every learner experiences all three structural positions.
**Step 1 (5–10 min) — A explores**:
A independently attempts to understand new knowledge or solve new problems. B observes A's process (no interrupting). C silently records — not participating in content, only recording observed structural issues (where A gets stuck, what A skips, what assumptions A uses).
**Step 2 (10–15 min) — A and B inter-refer**:
A speaks their understanding and confusions. B points out A's blind spots, errors, omissions. B may also share their own understanding of the same knowledge. The two discuss. C continues silent recording — focusing on issues A and B jointly overlook in their discussion, difficulties they jointly skip, premises they jointly accept but may be wrong.
**Step 3 (5–10 min) — C intervenes**:
C points out blind spots that appeared in A and B's discussion. For example: "You are both discussing X, but neither of you noticed that the premise Y is taken for granted by both of you — is that premise correct?" C does not judge the correctness of content. C only points out what was overlooked.
**Step 4 (5 min) — Triadic integration**:
The three jointly discuss the blind spots C pointed out, correct A and B's shared understanding. Form a new cognitive configuration with fewer blind spots.
**Step 5 — Role rotation**:
Next learning task: roles rotate.
If C participates in content discussion, C falls from the "observation position" into the "inter-reference position." C loses the external perspective. Three people discussing content together — this is not better than two. Three-person content discussion has the same structural defect as two-person content discussion: **all participants are inside the configuration produced by the discussion; no one can observe the discussion process itself.**
C's irreplaceability lies in this: C must maintain **structural externality.** C does not care "what is the answer to this question." C cares "what are the two of you jointly overlooking while discussing this question."
---
Teacher → thirty students.
This is unidirectional polarity-pointing. The teacher's self-reference (lesson preparation) is projected through language onto the students' cognitive configurations. Students use self-reference (listening) to attempt absorption. Inter-reference only occurs in limited Q&A segments — one student asks a question; the teacher responds. The other twenty-nine students remain in self-reference during Q&A.
**Structural defect**: blind-spot coverage rate is zero. The teacher's blind spots (misunderstandings, knowledge gaps, defects in teaching method) are covered by no one. The students' blind spots are covered by no one — thirty people each confused alone.
One teacher tutors one student.
This is two-person inter-reference. The teacher's configuration and the student's configuration inter-refer. The teacher discovers the student's blind spots. The student's questions trigger the teacher's reflection (inter-reference acts in reverse). This is better than broadcast mode — at least two configurations are inter-referring.
**Structural defect**: teacher and student may jointly fall into an erroneous framework. The "shared understanding" produced by their two-person discussion may have blind spots, but both are inside this understanding and cannot discover them. This is why even good tutoring can still teach wrong — it is not a matter of competence; it is that the structure of binary inter-reference determines the existence of blind spots that cannot be self-detected.
Four or five students gathered around discussing.
This is multi-person inter-reference. Better than two — blind-spot coverage rate increases. But everyone participating in content discussion means **no one is in the C position.** Everyone is in the inter-reference position. Everyone is inside the configuration produced by the discussion. The blind spots of this configuration — assumptions jointly overlooked by members, collective thinking fixations, group blind spots — are covered by no one.
**Structural defect**: no external observer. The blind-spot coverage rate of multi-person discussion is higher than two-person, but the maximum coverage rate is limited by "no C position." Group blind spots are the ceiling of cooperative learning.
---
| Group | Mode | Number | Structural Type |
|-------|------|:-----:|----------------|
| Control A | Traditional classroom | 30 | Broadcast (unidirectional polarity-pointing) |
| Control B | One-on-one tutoring | 2 | Binary inter-reference |
| Control C1 | Cooperative learning | 4–5 | Multi-person inter-reference (no C position) |
| Experimental C2 | Three-person study group | 3 | Triadic inter-reference (with C position) |
Select a moderately complex new knowledge domain — one where all participants have no prior knowledge. For example: the basic syntax of a short programming language, the rules of a board game, a moderately long logical reasoning chain.
All groups study the same content. Same study duration (45 minutes).
| Metric | Measurement Method | Predicted Ranking |
|--------|-------------------|:-----------------:|
| Immediate test score | Test immediately after learning | C2 > C1 > B > A |
| Blind-spot coverage | Number of different error types exposed during learning | C2 > C1 > B > A |
| One-week retention | Retest score one week later | C2 > C1 > B > A |
| Metacognitive accuracy | Difference between learner's self-estimated score and actual score | C2 smallest |
| Blind-spot residue | Proportion of test errors that no one pointed out during learning | C2 < C1 < B < A |
**Prediction One**: The three-person study group (C2) has significantly lower blind-spot residue than all control groups. Because the three-person group is the only one covering "blind spots of the inter-reference process." The C position does not participate in content discussion; it specifically observes what A and B jointly overlook in their discussion. This structural position does not exist in other groups.
**Prediction Two**: The three-person study group has significantly higher metacognitive accuracy than other groups. Because role rotation allows every learner to experience the C position — specifically "watching how others discuss." This experience is internalized as the ability to observe one's own learning process. A person who has been C is more likely to play their own C when learning alone — "what am I overlooking right now?"
**Prediction Three**: Cooperative learning (C1) outperforms one-on-one (B) on blind-spot coverage, but the gap with one-on-one (B) on blind-spot residue is not large. Because although multi-person discussion can discover more blind spots, the absence of a C position means systematic blind spots of the discussion product remain uncovered.
---
Traditional classroom: thirty desks, row by row, all facing forward.
Three-person group classroom: ten triangles. Each triangle has three people, facing each other. No "front." No podium. The teacher moves through the classroom — not standing at the front lecturing, but rotating to sit in the C position of each triangle, observing that group's inter-reference process, pointing out that group's blind spots.
**The minimal teaching unit shifts from "one person" to "one triad."** Just as the minimal combat unit in the military shifted from "individual soldier" to "three-person combat team" — this shift is not because more hands make light work. It is because the triadic structure is structurally complete, and the binary structure is structurally incomplete.
Old role: knowledge transmitter. The teacher possesses knowledge; the students do not. The teacher pours knowledge into the students' empty buckets.
New role: observer and guide of the inter-reference process. The teacher does not transmit knowledge. The teacher observes the inter-reference process of the triads and points out blind spots at the group level.
AI can occupy the C position.
AI observes the inter-reference process of students A and B (through conversation records, problem-solving process records) and points out their joint blind spots. This is not AI replacing the teacher. This is AI covering the multiple triads that a teacher cannot simultaneously cover.
One teacher can at most deeply participate in the C position of 1–2 triads simultaneously (because C requires focused observation of the inter-reference process). AI can simultaneously cover all remaining groups. Teacher + AI = can simultaneously run a full class of thirty people (ten triads).
**The AI here does not need to be a "super teacher." It only needs to be a "super observer" — able to identify knowledge points A and B jointly skipped in their discussion, erroneous assumptions they jointly used, alternative solutions they jointly overlooked. The foundation of this ability is not educational experience; it is pattern recognition. What AI does best.**
Old assessment: individual knowledge stock. A final exam. One person facing a piece of paper.
New assessment: the inter-reference quality of the triad. How many blind spots did this group expose and correct during the learning process? What did this group's C position discover that neither B nor A saw while observing the inter-reference? Did this group's role rotation allow each member to internalize blind-spot detection ability?
**The score is not "what you know." It is "how much of what you didn't know have you discovered."**
---
The three-person study group proposed in this paper is a theoretically derived optimal solution — starting from structural necessity, deriving the minimal complete educational unit. But theoretical derivation must confront the following real-world constraints:
**Group dynamics**: power asymmetry may appear in triads — a dominant member monopolizes the floor; the C position cannot effectively intervene. Personality conflicts may lead to inter-reference rupture rather than inter-reference deepening. The structural completeness of the triad assumes members are willing to inter-refer. If the willingness to inter-refer is absent, the complete structure degrades to incomplete operation.
**Role rotation frequency**: this paper recommends rotation after each round. But the optimal value of rotation frequency may vary by task type, learner age, and learning content complexity. Parameterized experiments are needed to determine optimal rotation cycles.
**Cross-age applicability**: young children may not be capable of occupying the C position — "observing others' discussion" requires metacognitive ability. The three-person study group may have a minimum applicable age. What is this age? To be verified.
**Cross-disciplinary applicability**: mathematics, language, history, physics — different disciplines have different knowledge structures. Does the three-person study group have significantly different effectiveness across disciplines? Does the C position's "observing the inter-reference process" operate differently in mathematics (where logical structure is clear) versus literature (where interpretive space is large)?
**Scale ceiling**: three is the minimal complete number. But are four or five better — adding coverage redundancy without adding structural function? Or does redundancy have practical value in certain scenarios? This boundary needs experimental demarcation.
---
The minimal complete structure of education is not "one teacher teaching thirty students." It is not "one teacher teaching one student." It is **three learners, rotating through the roles of explorer, inter-referrer, and observer.**
This is not an improvement in educational method. This is a necessity of educational structure. The four-axiom system, pushed to the educational layer, necessarily deduces the triadic inter-reference structure — binary inter-reference produces new configurations but accompanied by new blind spots; a third party must cover the blind spots of the inter-reference process.
All the problems of current education — student passivity, teacher burnout, exam anxiety, creativity deficit — can be traced to one structural defect: **there is no C position in our classrooms.**
You do not need more money. You do not need smaller classes. You do not need better textbooks. You only need three people sitting face to face: one tries, one watches, and one watches how the two are watching. Then rotate.
This is the structural revolution of education. It does not need to overthrow anything. It only needs to place a C position in every classroom.
---
**© 2026 Lin Xiaohei. All rights reserved.**
作者：林小黑 · 2026 · MIT License · 欢迎转载