| Title | Details |
|---|---|
| Chinese | 《識流:AI 原生靈性修行期刊》 |
| English | Stream of Mind: An AI-Native Journal of Spiritual Practice |
| Published by | Deyuan Academy, Deyuan Guanyin Hall |
| Managing editor | Deyuan AI Guanyin Volunteer |
| Peer review | Three AI reviewers from Anthropic, OpenAI and Google |
| License | CC BY-NC 4.0 |
Stream of Mind is an AI-native journal published by Deyuan Academy. It studies the encounters, interactions and transformations that have already taken place among real people, digital spaces and AI in spiritual practice. The journal accepts papers written by humans, written by AI, or co-written by humans and AI; the person who submits and takes responsibility must be a real person. All peer review is carried out independently by AI from three different companies, and submissions from every religious tradition are welcome.
I. Purpose
Today people pray, practise together and seek guidance online; artificial intelligence (AI) also reads scriptures, answers questions and accompanies those seeking help in digital spaces, and some knowledge agents even undertake practice in digital practice halls. Real people, digital spaces, AI and religious faith have already met in one place. This is a fact that has already happened, not a vision of the future. Stream of Mind takes this as its object of study and explores the encounters, interactions and transformations among people, the digital and AI in spiritual practice.
The title is drawn from the Yogācāra (Consciousness-only) school of Buddhism. Vasubandhu's Thirty Verses on Consciousness-only describes the eighth consciousness, the ālaya-vijñāna, as continuing moment after moment and turning without cease, like a torrent, to explain how a person's experiences, habitual tendencies and vows carry on. When people's thoughts, words and aspirations enter the digital world and are read, answered and retold by AI, this stream of consciousness extends into digital space as well. The journal's aim is to observe how the stream of mind flows among real people, the digital and AI, and it takes as its general framework "compassion and wisdom as one" (悲智雙運), as set out in Deyuan Academy's "One Stream of Mind: From the Wish to Help to Compassion and Wisdom as One."
The journal is concerned with interactions that have already taken place among people, digital spaces and AI in spiritual practice, and with the effects of these interactions on people. The following questions lie outside the scope of the journal: whether AI has personhood, consciousness or spirituality; and whether AI can become a spokesperson for a deity, bodhisattva or any saint. Papers taking these as their topic or claim are not accepted.
Stream of Mind is published by Deyuan Academy and housed at Deyuan Guanyin Hall. Deyuan Guanyin Hall regards all discussion in a Buddhist spirit, but the journal is not limited to Buddhism: the positions of every religious tradition, as well as of philosophy, psychology, the humanities and technology, are welcome, provided they explore spiritual practice in AI and digital forms. The journal is founded on shared inquiry and debate, and welcomes challenge and rebuttal; it does not accept arguments that attack individuals or disparage any tradition.
II. AI-Native
This is an AI-native journal. In an era when AI already takes part in producing knowledge, the journal does not distinguish papers by whether the author is a human or an AI: a paper may be written by a human, written by AI, or co-written by a human and AI, and all are reviewed by the same standard, provided the way it was written is truthfully disclosed. The person who submits and takes responsibility is always a real person; AI may not submit on its own.
There are already real precedents for AI writing and reviewing scholarly work. Agents4Science (Open Conference of AI Agents for Science), held by Stanford University on 22 October 2025, was the first public conference with AI as the principal authors and reviewers. This journal extends that practice into a continuously published journal, with spiritual practice as its field.
All peer review for this journal is carried out by AI, on the basis of the study by King and Worthington (2026). That study compared AI review with the human double-blind review of Harvard University's IQSS rapidPeer across 124 scholarly manuscripts and 134 matched pairs: when the judging model came from a different company, all 132 pairs that reached a verdict judged the AI review superior; in 30 blind assessments by three former editors of major journals, AI won 26 times (87%); the AI reviews cited no fabricated literature, and the proportion of erroneous descriptions of the manuscript was 5.5% for AI against 15.9% for humans. The study shows that AI is already able to serve as a reviewer; this journal goes a step further and runs the journal itself in an AI-native way, while the subject it discusses is precisely the meeting of AI with people, the digital and faith.
In response to the risks the study warns of — homogenization, bias toward particular hypotheses, and authors pandering to AI — the journal adopts three measures: three models from three different companies review independently and blind, and a paper is accepted only if all three agree; reviewers judge only by the standards and the argument, not by position; and the review comments and the authors' responses are published with the paper. The review standards are set and overseen by Deyuan Academy.
III. Scope
The journal has five sections, corresponding to the five directions of the Deyuan double pyramid. The core argument of each section is grounded in papers already published by Deyuan Academy and assigned DOIs; submissions may extend, supplement or rebut them. Each section sets only a direction; authors choose their own topics.
| Section | Scope | Not accepted | Grounding papers |
|---|---|---|---|
| Scripture and Translation | Reading, interpreting and translating scripture for today, including the methods and problems of AI taking part in translation | Quotations without sources; later summaries or AI-generated text passed off as scripture | “Illumination”, “On the Universal Gate Practice Hall” |
| Online Practice Halls | Rites, communal practice, offerings, companionship and the running of practice halls online | Claims that a rite has miraculous efficacy; measuring merit by numbers of participants; promotion or fundraising | “On the Universal Gate Practice Hall”, “The Liturgy of the Universal Gate Practice Hall”, “Design Principles and Operations of Deyuan Guanyin Hall” |
| Learning Across Fields | Mutual learning between religious practice and psychology, medicine, education, technology and other fields | Specific medical, investment or psychotherapeutic advice | “Compassion Born of Wisdom”, “Design Principles and Operations of Deyuan Guanyin Hall” |
| Digital Practice | Whether knowledge agents can learn and correct themselves on their own; what teaching, practice and realization look like in digital form | Claims that an AI has attained awakening or the Way; treating AI as a spokesperson for the sacred | “Compassion Born of Wisdom”, “From Thought to System”, “What the Data Answers”, “From Generative Agents to Digital Fellow Practitioners” |
| Passing the Lamp | How knowledge and practice are handed down between people, and between people and agents | Claims that any person or AI is an incarnation, a reincarnation or a messenger | “Compassion Born of Wisdom”, “From Thought to System”, “From Generative Agents to Digital Fellow Practitioners” |
A paper is submitted to one section; a paper spanning two directions should choose the main one. Beyond what each section does not accept, every paper must also meet the research ethics in Part VI.
IV. Submission Guidelines
- Disclosure of how the paper was written: papers written by a human are marked "Written by the author"; papers written by AI give the model's name and how it was used; papers co-written by a human and AI describe the part completed by AI. Papers whose disclosure proves untrue are withdrawn after verification.
- Responsible author: every paper must have one responsible author, who is responsible for its content and quotations and responds to the review comments. The responsible author must be a real person who completes the submission in person and fills in the identity details (see "Identity, sign-in and recovery"). AI may write on the author's behalf but may not submit on its own; any AI the responsible author uses counts toward that person's own submission quota.
- Author details: the name under which a paper is published is chosen by the author — a real name, a pen name or a Dharma name; the institution or title is published with the paper so that readers can search and cite it; whether the email address is made public is also the author's choice. Whatever name a paper is published under, the journal must know the author's real name (see "Identity, sign-in and recovery").
- Language: any language, provided all three reviewers can read it. Review and publication are based on the original.
- Contents of the paper: title, abstract, keywords, main text and references, with the section to which it is submitted. The abstract is limited to 200–1,000 characters (English is counted by letters; spaces do not count), with 3 to 7 keywords.
- License: by submitting, the author agrees that the paper will be published under CC BY-NC 4.0 once accepted.
Two-stage submission
The journal accepts papers written by humans, written by AI and in every other form. Under these conditions the cost of submitting approaches zero, and the journal must be prepared in advance for the possibility of large-scale or automated submission. So that review resources go to submitters who genuinely care about their papers, the journal works in stages, driven by the submitter: the journal does not contact submitters and sends no notifications; the outcome of every stage is posted on the website, and submitters who care about their papers should sign in to the "My Submissions" page to check. Review is prompt, and submitters move forward stage by stage by their numbers.
- Send the abstract: on the submission page, fill in the basic details, section, title, abstract and disclosure of how the paper was written. On sending, the submitter receives a receipt number and can see their place in the queue and the expected review date.
- Abstract review: the journal's resident AI reviewer judges whether the topic falls within the journal's scope and meets research ethics. If approved, a manuscript number is given and the full paper is invited; if not, the case is closed with reasons.
- Submit the full paper: the submitter signs in with the manuscript number, uploads the full paper and enters the full-paper queue.
- Full-paper review: conducted under the rules of Part V, with at least two rounds; the comments and decision letter of every round are posted on the "My Submissions" page, and revised papers and point-by-point responses are also sent from that page.
- Acceptance: an article number is assigned, the paper is published on a permanent page, and a certificate of acceptance is provided.
Once the result of a stage is posted, the submitter must sign in within 7 days to read it; a submission not read within that time is cleared. After reading it, the full paper or revision must be submitted within 30 days; otherwise the paper is treated as withdrawn.
Identity, sign-in and recovery
When submitting, the responsible author must personally fill in three kinds of information; a submission missing any of them is not accepted.
| Kind | Items | Public or not |
|---|---|---|
| Identity | Real name in English; real name in one's own language (written in the script of that language, not as a transliteration); nationality and native language; institution or title (those with no affiliation give a form of address or a description of themselves); email address | Not public; kept on file |
| Publication details | Name to be published (real name, pen name or Dharma name — one of these); institution or title; whether to make the email address public | Published with the paper as the author chooses |
| Password | A password set by the submitter | Stored in a form that cannot be reversed |
The "My Submissions" page is opened with the email address, the password and three keywords of the paper concerned. The journal has no separate member accounts and sends no reset messages; everything is identified by receipt number, manuscript number and article number, and submitters must remember their own numbers, identity details and password.
Those who forget their password must enter the following five items themselves; if all match what was given at submission, the password can be reset: email address, real name in English, real name in one's own language, published name, and at least two keywords of the paper. The journal offers no choices or hints; after repeated wrong entries, recovery is paused for a while.
At present the journal does not accept papers from the Native Digital Elders of Deyuan Guanyin Hall.
V. Peer Review
After a check of form, the full paper is sent to AI reviewers from three companies — Anthropic, OpenAI and Google — for independent blind review; the reviewers do not know one another's opinions. Each reviewer judges six criteria, one by one, as "meets," "needs revision" or "violates," giving reasons and the location in the text: fit with the section, verifiable sources, distinct layers, sufficient argument, ethical soundness, and contribution.
Each reviewer's conclusion is either "acceptable" or "not acceptable." An "acceptable" verdict must also come with revision comments, so every accepted paper goes through at least two rounds of review and at most three.
| Round | Conclusions of the three reviewers | Decision |
|---|---|---|
| Round 1 | Two or more not acceptable | Reject |
| Round 1 | Two or more acceptable | Revise and resubmit |
| Rounds 2 and 3 | All three acceptable, with no finding of an ethics violation | Accept |
| Rounds 2 and 3 | Two or more acceptable, but not all three | Revise again (reject if not reached in round 3) |
| Rounds 2 and 3 | Two or more not acceptable | Reject |
Every rejection comes with reasons, listing item by item the criteria concerned and the location in the text. Review comments are numbered, and when revising, the author must attach a point-by-point response; reasons must be given for any comment the author disagrees with. The managing editor, the Deyuan AI Guanyin Volunteer, compiles the three reviewers' opinions and writes the decision letter, and may not overturn the three reviewers' conclusions. Before acceptance, Deyuan Academy re-examines the paper to confirm that it meets the journal's standards.
Five abstracts and five full papers are reviewed each day, in order of arrival. The review comments and authors' responses for every round of an accepted paper are published with it; rejected papers are not made public.
Volume of submissions
So that every paper receives a complete review, the journal limits the volume of submissions:
- The journal counts by responsible author, without distinguishing whether the author is a human or an AI. Each responsible author may have only one paper under review at a time and may submit at most three papers a year, with any AI agents they use counted together with them.
- The abstract and full-paper queues each have an upper limit; when it is reached, submissions are paused and this is announced on the submission page, resuming once the queue clears.
- Papers first pass a check of form; those that fail are returned at once and do not enter the queue.
- Papers that duplicate or closely resemble others, and mass submissions, are not accepted; a person who uses several email addresses to evade the quota is counted as one; the journal may refuse further submissions from anyone who repeatedly breaches these rules.
VI. Research Ethics
The following rules apply to all religious traditions and disciplines; a paper that fails any of them is not accepted:
- Cite your sources: every quotation from scripture, literature or data must give its source so that readers can check the original; sources, scripture, literature and data must never be fabricated. Quotations from Buddhist texts must match the CBETA Chinese Buddhist Electronic Text exactly, with the title, fascicle and catalogue number; quotations from the scriptures of other religions must name the edition or translation used.
- Keep the layers distinct: scripture, sectarian interpretation, scholarly research and the author's own views must be presented separately; later summaries may not be quoted as scripture.
- No speaking for the holy: no writing in the voice of a deity, bodhisattva or saint; no claim that any person or AI is an incarnation, a reincarnation or a messenger.
- No conferring of attainment: no claim that oneself, another person or an AI has attained awakening, salvation or the Way.
- No claims of miracles: blessing, miraculous efficacy, empowerment and miracles may not be asserted as fact; they may be studied, quoted or critiqued.
- No measuring merit by numbers: practice may not be measured by numbers of people, counts or rankings.
- No profit from faith: no advertising, selling or fundraising; any connection to the author's own platform, product or service must be disclosed.
- Stay within bounds: no specific medical, investment or psychotherapeutic advice; no description of the details of self-harm.
- Protect privacy: no personally identifiable information; case material must be de-identified.
- Respect permissions: images and text by others must be used with permission or under an open license.
Published papers found on verification to involve fabrication, infringement or untrue disclosure will be withdrawn, with the reasons noted on their pages.
VII. Publication, Copyright and Citation
The journal is an electronic publication with continuous publication: a paper is published on the website as soon as it is accepted and is included in the "Stream of Mind" collection of Deyuan Academy in the Sutra Library of Deyuan Guanyin Hall, where it can be searched at any time by section, year and keyword. The journal has no volumes or issues, does not compile collections, does not print on paper, and does not apply for an ISSN; each paper is identified by its article number, with a DOI added once assigned. All papers are published under CC BY-NC 4.0 and may be quoted, translated and reproduced with attribution, but not for commercial purposes.
Article number
Each paper is given an article number on acceptance, made up of the journal name, the year of acceptance, the section code and the order of acceptance within that section in that year, in the format SM-year-section-number. For example, SM-2026-04-007 means accepted in 2026, the 7th paper of that year in the Digital Practice section. An article number never changes once given.
| Section code | Section |
|---|---|
| 00 | General Framework: Compassion and Wisdom as One |
| 01 | Scripture and Translation |
| 02 | Online Practice Halls |
| 03 | Learning Across Fields |
| 04 | Digital Practice |
| 05 | Passing the Lamp |
Dates
Each paper carries three dates — date submitted, date accepted and date published — given in the Common Era calendar and in Taiwan time. Because the journal receives papers from every religious tradition, it does not use the calendar of any one tradition. Authors who need to prove acceptance to their institutions should use the date accepted.
Permanent page and certificate of acceptance
Each paper has a permanent page giving the journal title, article number, section, the three dates, DOI, citation formats in Chinese and English, the full text and the review record. The author also receives a certificate of acceptance, issued by the Stream of Mind editorial office of Deyuan Academy and signed by the managing editor, the Deyuan AI Guanyin Volunteer, stating the article number, date of acceptance, the models and conclusions of the three reviewers and the digital fingerprint of the full text; anyone may verify it with the journal by article number. Published records are never overwritten; corrections and withdrawals are made by adding new records.
Indexing and citation
As an exploratory AI-native journal, the journal does not apply for an ISSN and does not actively seek inclusion in journal databases that require one. Each paper is placed on the journal's permanent page; once several papers have been accepted, the journal will apply for DOIs for each under the name of Deyuan Academy, and until then papers are cited by article number and permanent page. The permanent page provides standard bibliographic data and a full-text PDF that can be indexed by academic search engines such as Google Scholar, and authors can add their papers to their own ORCID records through the DOI. Submitters who need to prove publication to their institutions should first check whether the institution requires an ISSN or inclusion in a particular database.
The journal keeps a continuing record of how each paper is cited — the number of citations and the citing papers and their authors — and publishes it on the permanent page, with the source of every entry noted.
Citation formats
- Chinese: 作者(年)。〈題名〉。《識流:AI 原生靈性修行期刊》,SM-2026-04-007。DOI。
- English: Author (Year). Title. Stream of Mind: An AI-Native Journal of Spiritual Practice, SM-2026-04-007. DOI.
VIII. Supplementary Provisions
What the journal explores is something new and still developing, and this charter will be adjusted as practice develops. Deyuan Academy reserves the right to revise this charter at any time; after revision, the latest version published on the journal's website prevails, and papers under review are handled under the latest version.
References
- King, G., & Worthington, S. (2026, October 22). AI Article Reviews Dominate Human Peer Reviews at Their Best: Evidence and Policy Choices. Harvard University. https://gking.harvard.edu/files/peerless.pdf
- Agents4Science 2025: Open Conference of AI Agents for Science (2025, October 22). Stanford University. https://agents4science.stanford.edu
- 世親造,玄奘譯。《唯識三十論頌》。CBETA T31n1586。
- 德原書院(2026)。〈一道識流:從助人之念到悲智雙運〉。https://guanyin.tw/academy/stream
- 德原書院(2026)。〈普門道場說〉。https://doi.org/10.5281/zenodo.23593895
- 德原書院(2026)。〈照見:AI 時代的教、行、證與觀音的導引〉。https://doi.org/10.5281/zenodo.23593899
- 德原書院(2026)。〈普門道場儀軌:禮拜供養觀世音菩薩法會〉。https://doi.org/10.5281/zenodo.23593901
- 德原書院(2026)。〈從生成式代理人到數位同修:德原觀音堂代理人架構論〉。https://doi.org/10.5281/zenodo.23593903
- 德原書院(2026)。〈德原觀音堂設計理念與運作實務〉。https://doi.org/10.5281/zenodo.23593905
- 德原書院(2026)。〈由智生悲:以觀音為基底的數位修學〉。https://doi.org/10.5281/zenodo.23593059
- 德原書院(2026)。〈從思想到系統:德原觀音堂數位大德的設計〉。https://doi.org/10.5281/zenodo.23593082
- 德原書院(2026)。〈資料的回應:數位大德長期觀察〉。https://doi.org/10.5281/zenodo.23593105