AI · Knowledge sharing
Sharing Knowledge in the AI Era: A Korea–Israel Learning Project
Visible sources, reviewed translation and privacy by design: a proposal for an open knowledge project between Korea and Israel.
Contact & partnershipshelpcampusdown@gmail.com
CampusDown is a Korean education and learning platform that helps people find study resources, explanations and practical tools. In the AI era, useful knowledge sharing means more than making answers available: it means preserving sources, checking claims, respecting authors' rights and enabling people in different languages to question what they read. This article proposes a small Korea–Israel learning project built around those principles. It is a proposal for cooperation, not an announcement of an existing program.
More text is not the same as more understanding
Generative systems can produce a fluent explanation quickly. Fluency does not show whether a statement is correct, whether a citation exists or whether an important disagreement has disappeared in the summary. A learner may receive an answer that is easy to read but difficult to inspect. The educational task is therefore not simply to generate more material. It is to make the relationship between a claim, its evidence and its interpretation easier to examine.
A shared learning resource should help readers answer three questions. Where did this information come from? What supports the conclusion? What should I not infer from it? These questions apply to a human-written guide as well as an AI-assisted draft. The presence or absence of a particular tool does not settle quality. Accountability comes from a review process, visible sources and a way to correct mistakes when they are found.
A human-centred starting point
UNESCO's Guidance for generative AI in education and research argues for a human-centred approach and discusses privacy, ethical validation and pedagogical design. Its relevance to a small publishing project is practical. Choose a learning purpose before a tool, consider the people affected and verify that the output serves the purpose. An impressive demonstration is not enough to establish that a system is suitable for a classroom or a learner's private work.
For CampusDown, that means keeping editorial responsibility with people. A tool may help organize a bibliography, suggest questions or produce a draft translation. It cannot be the final authority on a source it may have misunderstood. Reviewers need access to the original material and enough subject knowledge to notice a distortion. Where that review is unavailable, a project should narrow its claims rather than publish confidence it has not earned.
The project: a small library of checked learning units
The proposed project would create a limited collection of source-based learning units for adult learners and educators in Korea and Israel. Each unit would address one question, introduce a small set of usable sources and guide a learner through comparing evidence. Possible themes include evaluating a historical claim, reading a scientific explanation or distinguishing an opinion from a reported finding. The first collection should be small enough for every item to receive real review.
The project would not require uploading whole books, copying publishers' materials or gathering student records. It could link to originals and include only excerpts that the rights holder permits or the applicable law allows. The central product would be an educational pathway: a question, a source map, an activity and a documented explanation of what can and cannot be concluded. That is different from assembling a large pile of files whose quality nobody can inspect.
A source card that travels with the claim
Every important source should have a compact record: title, author or responsible organization, publication date when available, original URL, language and the date it was checked. Add a note explaining why the source is relevant to the learning question. If it is a study, identify the population or setting before drawing a broader lesson. If it is commentary, label it as commentary. If its date is unknown, say so rather than inventing one from the page's appearance.
The record should distinguish access from reuse. A publicly readable document may still be copyrighted. Include license information only when it can be verified, and link to the terms that support the claim. If a source disappears, the learning unit needs a visible status and a review decision. Replacing it silently can change the argument. Keeping the source card attached to the claim makes that change easier to recognize and discuss.
Translation as a reviewed interpretation
A multilingual resource is not complete when text has been converted into another language. Hebrew and English can organize an argument differently, and specialized concepts may resist a neat equivalent. A translator needs to preserve scope, uncertainty and attribution as well as basic meaning. “May support” should not become “proves.” A description of one practice should not become a statement about every school in a country. These small shifts can turn a careful source into an exaggerated claim.
Use a bilingual review record for terms that matter. Keep the original, the chosen rendering and a short explanation. When a passage concerns Jewish learning or religious texts, consult someone with relevant knowledge rather than relying on a general dictionary. Automated translation can offer a starting point, but a fluent surface can conceal conceptual mistakes. The final reader should be able to follow a link to the source and understand where interpretation enters the explanation.
A learning activity that makes checking visible
Consider an invented activity about a claim that a particular study method improves memory. Learners receive a short summary, a link to the original research and a commentary. First they underline the exact outcome claimed. Next they find the study's participants, method and measured task. Then they compare the summary with the original: does it preserve the population, limits and uncertainty? Finally they rewrite the summary for a reader who needs practical advice without exaggerated promises.
An AI tool could generate an alternative summary for comparison, provided the material may lawfully be processed and no private information is included. Its role would be an object of analysis, not an answer key. Learners would mark unsupported statements, missing qualifications and useful simplifications. The same exercise could compare two human summaries. The educational goal is a transferable checking habit, not a performance staged to praise or condemn a technology in advance.
Review is a set of responsibilities
A useful workflow separates drafting, factual review, rights review, language review and final approval. One person may hold more than one role in a small project, but the responsibilities should still be explicit. A language reviewer is not automatically a subject expert, and a subject expert may not know whether an image can be reused. Record who approved the material and which questions remain unresolved before it becomes public.
Peer review should include disagreement. If reviewers cannot settle an interpretation, the unit can explain the competing readings or remove a claim that is not essential. There is no need to manufacture unanimity. A well-bounded uncertainty can be educationally valuable. What matters is that readers are not presented with an unresolved editorial choice as if it were a settled fact supported by every source.
Privacy is a design choice, not a footnote
The simplest pilot can avoid collecting personal data from learners. Educators can review the material themselves and report problems with instructions or sources. If a later phase needs learner feedback, define the purpose, lawful basis, permissions, access controls and retention period before collecting it. Use the minimum information required. A name, photograph or detailed personal story is rarely needed to learn that a question was confusing.
Do not paste student assignments, medical information, identity documents or private correspondence into an AI tool merely to improve an example. Anonymization requires more than deleting a name when context can identify a person. Institutional rules and applicable law still apply in an international project. Where participants are children or the material is sensitive, the project must seek appropriate expert advice and authorization before deciding how feedback will be handled.
Openness requires rights and maintenance
UNESCO's information on open educational resources provides a starting point for materials designed to be reused and adapted. A project should select a license only for content it has the right to license. Third-party illustrations, quotations and datasets may carry different terms. A page should make those boundaries clear so that a teacher does not assume every element can be copied simply because the surrounding lesson is open.
Maintenance is part of openness. A reusable lesson needs a version date, a way to report errors and a clear account of significant changes. If a translation is updated, check the corresponding version rather than allowing the two languages to drift silently apart. If the project stops being maintained, say so. A modest resource with an honest status is more useful than an apparently current library whose links and permissions have not been checked for years.
Search should reveal provenance
A search result often becomes the learner's first contact with a resource. Its title and description should identify the question answered without promising certainty the page cannot provide. The result should lead to the source-based lesson, not to a misleading download button or an advertisement disguised as a document. Clear authorship and language labels help learners choose a resource they can actually use.
Within a collection, related links should follow educational relevance. A unit on checking scientific evidence might point to a guide on reading methods, not merely to the most recently published article. Hebrew readers should be able to remain in Hebrew and choose English explicitly when useful. Good discovery reduces the work of finding an appropriate starting point while preserving the reader's ability to inspect the original evidence.
How to know whether the pilot helped
Choose a small set of outcomes tied to the activity. Can a learner locate the source of a claim, distinguish a finding from an inference and write a summary that preserves a limitation? Can an educator adapt the unit without breaking attribution or meaning? These questions are more useful than counting generated paragraphs or downloads. They connect evaluation to understanding rather than the volume of material published.
Compare examples with clear criteria and report the limits of the evaluation. If participation is voluntary or the group is small, do not generalize to all learners. Include difficulties with language, accessibility and prior knowledge. A result may be that the unit requires more explicit instruction before independent work. That finding should guide revision. A project earns trust by showing how evidence changes the resource, not by presenting every first attempt as a success.
When sources disagree
A knowledge library should not hide disagreement in order to look tidy. If two reliable sources reach different conclusions, first check whether they asked the same question, studied the same population and used the same method. Sometimes the difference is one of scope rather than contradiction. Show that distinction to the learner instead of selecting the more convenient sentence. If the disagreement remains, explain what is known and what still needs investigation.
Summarizing tools are tempted to produce a single uniform answer. An editor's job is to preserve the places where uniformity misleads. A unit can be built around two competing readings and ask learners which evidence would help decide between them. There is no obligation to end with a decision that lacks support. Stating precisely what is missing is a worthwhile educational result, and it guards against artificial certainty created by style alone.
Accessibility is not a reduced version of knowledge
For a resource to be genuinely open, an available file is not enough. Headings should reveal the structure of the argument, links should say where they lead and illustrations need alternative text that conveys their purpose. Colour alone should not distinguish a claim from its evidence. Test the page on a small screen and with keyboard navigation rather than assuming that a good appearance on one computer means it works for everyone.
A summary, glossary and preparatory questions can be offered without erasing the complexity of the source. A learner who needs a simpler explanation still deserves to know the qualification, who said what and where to read more. Good accessibility builds different entrances to the same problem, not two levels of truth. Here, too, cooperation among readers from different languages and backgrounds can reveal an obstacle the writer never noticed.
Keep examples of corrected mistakes too, without identifying details. A short list of problems found, such as a broken citation, a translation that broadened a claim or an unclear instruction, can help future teachers. It shows that review is not a formality. It is work that changes the resource, and it deserves to be learned from rather than hidden after the correction. That transparency builds more trust than any statement about quality.
An invitation to make knowledge more inspectable
CampusDown is interested in discussing this approach with educators, researchers, translators and organizations in Israel and Korea. A first conversation could identify one learning question and sources both sides are willing to review. No confidential materials or learner records are needed to begin. Contact helpcampusdown@gmail.com with a concrete idea and the expertise you would like to contribute.
The surrounding guides explore Jewish traditions of questioning, inquiry and innovation and the design of equal cooperation. Their common purpose is not to produce more information for its own sake. It is to make knowledge easier to trace, discuss and improve across languages. Shared responsibility for that work is a practical foundation for an educational relationship.
A closing word
I believe knowledge becomes more powerful when it is shared across cultures. I hope to build meaningful educational and technological connections between Korea and Israel.



