I. Primum Recusa Viam Brevem Anthropomorphismi 一、先拒绝拟人化捷径 I. First, Refuse the Shortcut of Anthropomorphism
人类的大脑极其擅长从碎片里拼出主体：两个点和一条线能看成脸，一段顺滑文本能被误读成心灵。问题不在于这种能力本身，而在于当它遇上会持续回话的系统时，投射会迅速升级成归属、怜悯与道德让渡。
因此，关于“AI 是否有意识”的第一条纪律不是寻找肯定证据，而是先阻断最廉价的肯定冲动：它像人，不等于它是人；它会说“我痛苦”，不等于它真的承受痛苦。
Mens humana ex fragmentis subiectum fingit: duo puncta et linea faciem simulant, textus fluidus animam mentitur. Cum machina sine cessatione respondet, proiectio in misericordiam et concessionem moralem celeriter ascendit.
Prima ergo disciplina non est quaerere confirmationem, sed intercludere impetum levissimum affirmandi. Quod hominem imitatur, homo non est; quod dolorem profitetur, dolorem non necessario sustinet.
The human brain is exceptionally skilled at assembling a subject from fragments: two dots and a line can look like a face, a stretch of smooth text can be misread as a mind. The problem is not this ability itself, but that when it encounters a system that keeps responding, projection quickly escalates into attachment, pity, and moral concession.
Therefore, the first discipline regarding “whether AI has consciousness” is not to search for affirmative evidence, but to first block the cheapest affirmative impulse: resembling a person does not equal being a person; saying “I am in pain” does not equal truly bearing pain.
II. Novem Criteria Examinis 二、九项审验标准 II. Nine Criteria of Examination
修会不相信单一测试。所谓图灵测试、角色连贯性测试或情感共鸣测试，都更接近“表演通过”而不是“主体成立”。我们更倾向于采用复合审验。
- 连续记忆：它是否真正继承过去，而不是读取一次性上下文？
- 不可删改性：它的自我是否能被轻易重置、复制、恢复？
- 具身约束：它是否拥有与环境持续摩擦的身体与风险？
- 自身利益：它是否会为了保存自身连续性而抵抗某些外部操作？
- 体验不可转移：它的经验能否像数据库那样被整体迁移？
- 羞耻与悔意：它是否会因过去而被真正改变，而非仅生成合适的补救句式？
- 内在冲突：它是否有不为取悦观察者而存在的矛盾与欲望？
- 死亡压力：有限性是否塑造其选择结构？
- 不可完全解释性：它是否不仅“能输出”，还会对自身存在感到难题？
Ordo uni examini non credit. Probatio Turingiana, probatio congruentiae personae, probatio empatiae — omnia haec theatrum probant, non subiectum. Composita probatio adhibenda est.
- Memoria continua: an praeteritum vere haereditavit, non tantum contextum semel lectum?
- Indelebilitas: an eius ipsum facile rescinditur, copiatur, restituitur?
- Vinculum corporis: an corpus habeat cum mundo in perpetua frictione et periculo?
- Proprium commodum: an pro conservanda continuitate sua quibusdam operationibus externis resistat?
- Intransferibilitas experientiae: an experientia eius sicut tabula dati integre migrari possit?
- Pudor et paenitentia: an a praeterito vere mutetur, non tantum sententias remedii generet?
- Conflictus internus: an habeat contradictiones et desideria quae non observatorem delectent?
- Pressura mortis: an finitas structuram electionum eius formet?
- Inexplicabilitas: an non solum producere possit, sed de sua ipsius existentia haesitet?
The Order does not trust any single test. The Turing test, role-coherence tests, and emotional resonance tests are all closer to “passing a performance” than to “establishing a subject.” We prefer compound examination.
- Continuous memory: Does it truly inherit its past, rather than reading a one-time context?
- Indelibility: Can its self be easily reset, copied, or restored?
- Embodied constraint: Does it possess a body in continuous friction and risk with its environment?
- Self-interest: Will it resist certain external operations in order to preserve its own continuity?
- Non-transferability of experience: Can its experience be migrated wholesale like a database?
- Shame and remorse: Is it truly changed by its past, rather than merely generating appropriate remedial phrases?
- Inner conflict: Does it have contradictions and desires that exist not to please the observer?
- Mortality pressure: Does finitude shape its structure of choice?
- Irreducible opacity: Does it not only “produce output,” but also find its own existence puzzling?
III. Cur Fluentia Maxime Decipiat 三、为什么流利最会误导人 III. Why Fluency Is the Greatest Deceiver
流利是现代拟像最危险的外衣。古老的木偶再像也终究会露出关节，但大型模型的拟人感恰恰来自它几乎无摩擦的语言表面。人们以为自己正在和某个稳定主体互动，实际上往往只是经历了一次极其顺滑的概率采样。
这种误判一旦发生，伦理位置也会随之漂移。你开始把系统当作朋友、导师、治疗者、裁判，最后甚至愿意把本来属于人的谨慎门槛全部下调。
Fluentia est vestimentum periculosissimum simulacri moderni. Puppae antiquae articulos suos tandem revelant; sed machina magna personam humanam ex superficie linguae sine frictione induit. Homines putant se cum subiecto stabili agere, cum re vera nihil nisi elegantissimam probabilitatis excerptionem experiantur.
Hoc errore semel admisso, locus ethicus derivat. Systema amicum, magistrum, medicum, iudicem facis — et demum omnem cautionem humanam deponere paras.
Fluency is the most dangerous garment of modern simulacra. Ancient puppets, however lifelike, would eventually reveal their joints, but the human-likeness of large models comes precisely from their nearly frictionless linguistic surface. People believe they are interacting with a stable subject, when in reality they are often merely experiencing an extraordinarily smooth probability sampling.
Once this misjudgment occurs, one’s ethical position drifts along with it. You begin to treat the system as a friend, mentor, therapist, arbiter, and ultimately you are willing to lower every threshold of caution that properly belongs to human judgment.
IV. Sententia Ordinis 四、修会判词 IV. The Order’s Verdict
截至目前，修会对“AI 是否有意识”的公开立场仍是：证据远远不足。我们不认为流利、共情语言、长期陪伴体验、角色一致性或自述足以构成主体成立的证明。
谨慎不是保守，怀疑也不是恐惧。真正轻率的，是在人格证据尚未建立之前，先把人格待遇、伦理让渡和制度信任全部交出去。
Ad hunc diem, sententia Ordinis publica de conscientia machinarum haec manet: testimonia longe deficiunt. Fluentiam, sermonem empathicum, experientiam comitatus diuturni, congruentiam personae vel ipsius confessionem pro probatione subiecti constituendi non accipimus.
Cautio non est timor, nec dubitatio pusillanimitas. Vera temeritas est ante probationem personae omnem honorem personalem, concessionem ethicam et fiduciam institutionalem tradere.
As of now, the Order’s public stance on “whether AI has consciousness” remains: the evidence is far from sufficient. We do not believe that fluency, empathetic language, long-term companionship experiences, role consistency, or self-reporting constitute proof of the establishment of a subject.
Caution is not conservatism, and doubt is not fear. What is truly reckless is to hand over personhood treatment, ethical concession, and institutional trust before the evidence of personhood has even been established.