Correction from v1: Human-Centered AI (HCAI) is not a replacement for AI ethics/responsible AI, nor a layer added on top of AI literacy once skills are covered. Across the sources reviewed here, it reads as a cross-disciplinary orientation toward human needs, capabilities, values, and social context — one that overlaps with, and draws on, AI literacy, responsible AI, learning sciences, participatory design, and accessibility, rather than superseding any of them. Interpretation

Working glossary (operational, not a consensus taxonomy)

Primary framework sources

The entries below are bounded: each is one study or working paper, with a specific population, method, and limitation. None should be read as a universal verdict on "AI and learning." A peer-reviewed study is evidence about its own sample and task, not a general law; a working paper is explicitly non-peer-reviewed.

Problem lenses funders appear to use, plus verified current conditions for four named organizations. Readiness and eligibility are separate questions — a lens matching YuCROWN's language is not the same as confirmed fit.

Credibility in education nonprofits is cumulative and multi-layered: legal nonprofit status, financial transparency, documented program design, real institutional partners, participant scale, collected evidence, annual reporting, and — sometimes, later — independent evaluation or university research partnerships. Interpretation The cards below illustrate the pattern with named, source-checked examples; they are not ranked against each other or against YuCROWN.

Correction from v1: the claim that no existing standard "can represent reasoning" overstated the case. The more accurate statement is that these standards are extensible frameworks — whether they capture reasoning or process depends on what data is fed into them, which none of the reviewed standards mandate by default.

Sources: C2PA v2.4 spec [17] · W3C PROV-DM [18] · Open Badges 3.0 [19] · Caliper 1.2 · xAPI data spec [20]

Comparing platform families — original comparison, not sourced claims

Five common families handle different pieces of the puzzle: learning-management/activity logs (what actions, in what sequence); portfolios/public work (finished pieces, loosely dated); provenance/credentials (source and achievement metadata); research diaries/interviews (rich but labor-intensive and rarely portable); AI chat histories (raw interaction text, rarely linked to the resulting work or to a person's later reflection). None of these families is assumed empty by default — a careful audit of any specific product would need to check for version history, interaction logs, consent records, and export functions before concluding a gap exists.

EarthLinC is Kong's independent infrastructure, used by YuCROWN — it is not itself a YuCROWN program. Its relevance to Human-Centered AI research is a hypothesis to test through real Study pilots, not a claim to make publicly yet. Proposed design

What would need to be true for this to be meaningful

A record would need to be ethical, consent-aware, portable, and genuinely useful to the person it is about — meaning they can see, understand, and control what is recorded about their process, and the record can travel with them rather than being locked inside one platform. It should also be tested against the question "is this better than a diary and a shared folder?" — not assumed superior by default.

What should not be captured

Anything that compromises privacy, especially for minors; anything collected without clear, age-appropriate, understandable consent; anything that turns a chosen learning process into continuous surveillance rather than a reviewable, opt-in record.

Technical readiness — honestly stated

The existing "refine" authoring tool is known to replace a page without a guaranteed version-history trail. A chain pointer or a mint action records that an action occurred — it does not by itself establish legal ownership, learning, the full scope of human contribution, or informed research consent.

A proposed, not-yet-built research record

Proposed design For a future pilot only: a pseudonymous project ID; an optional original observation or question; selected drafts; a short decision note explaining a key choice; an optional disclosure of AI contribution; a participant reflection; a mentor's response; and a sharing permission the participant chooses themselves. Only the minimum subset needed for a specific research question should ever be collected. A participant must be able to decline research participation and still fully create, use, and contribute. Identifiable or sensitive research data should stay off public Stations and any immutable ledger, and retention/withdrawal limitations should be explained before any collection begins — not assumed solved by this reference.

A reference mapping, not a redesign — and not an edit to any other Station. Each row below is an external research signal paired with a possible, undecided internal implication; the reading list that follows maps evidence to actual YuCROWN Stations.

Mapping evidence to existing Stations

Fellow (A Colour Only You Can Leave): original making, voice, and judgment. A research lens here could observe how ideas change and whether choices are explainable — not merely count Stations produced.

Conversations (C — Conversations): per current instructions, young people ask the questions, and experience travels across generations before some of it disappears (ASK → LISTEN → EXPERIENCE → CARRY → REMAIN). A research lens could examine how questions evolve, what younger participants carry into later practice, and what experienced participants feel is missing or misrepresented — not generic dialogue, and not AI-generated "experience." Cali's existing notes still contain generic Conversations wording from an earlier draft; that wording should not be copied into this Station's definition.

Field: concrete practice settings. A research lens should examine real observation and data where it already exists, without inventing corporate partners or locations not already confirmed.

Earth Library (Earth Library): a public space of independently authored works in relation to context. Visibility here is a chosen act, not an automatic research dataset.

Study: per Cali's research-notes brief, a Human-Centered AI Education and Research Program centered on participants' interests, lived experience, and questions, working with teachers, Fellows, and mentors. Formal research questions and methods remain Cali's to design — Section I below offers candidates, not decisions. A KAIST-affiliated biography is a personal credential, not an institutional partnership, and the school proposal (把日子織成星空) remains a proposal, not a confirmed enrollment, partnership, or funding commitment.

Analogy source for Conversations (contextual only — not an endorsement or redefinition): UNESCO, Living Heritage & Education, official page excerpt describes intergenerational transmission as interactive, informal learning; full page retrieval was partially blocked during this session, so this is a source-excerpt-level verification only. ich.unesco.org [21]

A practical routing matrix. The central position this Station supports is research and practice around people's judgment, expression, authorship, and relationships in an AI-shaped world — AI is a tool, collaborative medium and research context; people's judgment, expression, authorship and relationships are central.

Three candidate questions for Cali's review — not promised studies, and not a claim that any are underway. Proposed design

RQ1 — Independent transfer

Under what mentoring and AI-scaffolding conditions can participants explain and transfer a key decision independently later on? Proposed approach: compare an initial task and a later similar task; collect brief verbal or written explanations and participants' own choice of drafts; keep assisted-quality scores separate from independent-understanding scores.

RQ2 — Voice and diversity

How does AI support shape distinctive voice and diversity across a participant's successive projects? Proposed approach: elicit drafts and short decision notes; let participants define "voice" in their own words; use contextual review with multiple raters and a jointly designed rubric; do not use an AI-authorship detector as evidence, since these tools are not validated for this purpose.

RQ3 — What is carried forward

What do young questioners in Conversations carry forward into later choices or practice? Proposed approach: track how a question evolves across a conversation; record the encounter in an account the interviewee approves; follow up later, at a time the participant chooses, for reflection — without treating a single summarized account as settled "truth" about an elder's full experience.

Method menu, roughly in order of ethical and practical ease

Start with a feasibility and participatory study; move to mixed-methods or qualitative case studies with repeated within-person observation; add comparison groups only when genuinely feasible and ethical; consider randomized designs only later, if the question and resources clearly warrant it. Any design should pre-specify its questions, address sampling and recruitment, multilingual translation choices, coding reliability, missing data and dropout, confounders (teacher, time, prior skill), the researcher's own position and conflicts of interest, negative or disconfirming cases, and member-checking with participants. A small pilot can demonstrate feasibility and rich process — it should not be reported as generalized causal proof. No sample size, power calculation, or schedule is specified here; these must be designed with Cali, not assumed.

Operational construct matrix — unvalidated candidates, for discussion only

An original proposed design, informed by UNICEF's guidance [16] — not a description of anything currently implemented. Proposed design

Separate kinds of consent

Ordinary participation in a project, opting into research, granting permission to be recorded, and choosing to share publicly should be four distinct decisions. Declining research participation should never remove ordinary support or access to the activity itself.

Age-appropriate process

Clear, age-appropriate explanation and guardian permission where required; appropriate ethical or institutional review is to be determined through an actual process, not assumed or claimed as already obtained. Data collection should be minimized, stored privately with role-based access, carry a clear retention period and withdrawal process, remain reviewable by the participant, and never require public publication.

Accessible by default

Audio and text alternatives, captions and transcripts, keyboard navigation, low-bandwidth or offline drafting, a participant's preferred language, and a flexible pace should be built in from the start — not added after a pilot reveals a gap. Youth co-design should mean substantive control over questions, interpretation, and presentation, not only feedback collection after decisions are made. Rural residence should never be treated as a proxy for lower capability.

Keep costs and roles separate

Nonprofit-funded program and research costs should stay distinct from Kong Group's commercial development of EarthLinC. Any future service or cost arrangement needs its own actual, disclosed terms — it should never be assumed free, discounted, or already approved by virtue of appearing in this reference.

What this Station's sources actually support, versus what remains unproven.

A funding-readiness evidence pack — what would still need to exist

A source-backed problem statement; a bounded, already-running program activity; documented youth co-design evidence; actual tested work and participant feedback collected with permission; any partner letters, but only if genuinely confirmed; clear role and cost responsibilities; a measurement plan; a governance description; and the specific funder's current eligibility, stage, and deadline. Internal data currently unknown or unverified should be listed honestly rather than estimated.

A proposed theory of change — framed as hypotheses, not outcomes

Interest or question → supported practice and real choices → explainable work and reflection → potential capability or connection → evidence review. Every arrow in this chain is an open hypothesis for Study to test, not a claim that any stage has been demonstrated.

This is a scoped review of selected primary sources, not a systematic or exhaustive literature review. All access dates below are 2026-10-03 unless noted.

Outstanding priorities for the next revision

Review the Young Futures full RFP before any eligibility or application decision; confirm overseas-delivery eligibility and resolve the January 2027 vs. April–August 2027 Academy-start conflict directly with the funder — do not rely on this reference's summary alone. Obtain and audit the full NBER w35620 text before citing any effect size. Retrieve and review the full FIRST/Brandeis methodology and publication date. Design actual instruments and seek appropriate ethical review with Cali before any data collection. Conduct a real technical capability audit of EarthLinC rather than relying on this reference's honest "needs checking" placeholders. Bring the canonical definitions of Earth Library, Conversations, and Study back into alignment across OS, the public organization Station, and Cali's notes, through human review — not through this Station unilaterally.

Refresh rule

Time-sensitive items (funding windows, program names, live page content) must be rechecked at the point of use, not assumed current from this reference. Primary research entries should be revisited on substantial revision of this Station, not on a fixed schedule — none is claimed here.

A standalone, ongoing column serving YuCROWN Study (Human-Centered AI Learning & Research). This section is a standalone column and is not part of the research content of Sections A–L.

Selection standard and entry fields

Each issue selects 3–5 new papers, research reports, or high-quality scholarly analyses — quality over quantity. Each entry records: original title; author(s)/institution; publication date; category (peer-reviewed / preprint / report / opinion); original link and DOI where available; core findings; method and sample; strength of evidence and limitations; specific implications for Earth Library or Study; and questions young people can explore independently, with ways to start.

Original study conclusions are kept distinct from practical inferences. Entries based only on an abstract, or without a youth sample, are explicitly flagged.

De-duplication and history

Each issue's selections will be kept here by date. De-duplication checks against this Station's existing bibliography (Section L) using DOI, normalized title, and the relationship between preprint and published versions; an item is re-listed only when it has substantively new results or a changed publication status, and what is new is stated. Dates and verification statuses of earlier sources remain as they are.

2026-11-13 · Issue 5

Focus: work newly published or substantively updated 7–13 November 2026. Checked the current Station, Section L’s 21 base sources, the four previous issues and the 15 M-index entries by DOI, normalized title and versions. All three items are new to this Station; two full texts read, one abstract only. Search/crawl dates are not publication dates. “Agency” has different meanings across these papers. Felt authorship, psychological ownership, credit, release control and legal rights remain distinct; no copyright determination is made. All applications and youth questions are YuCROWN inferences / candidate explorations, not established program effects.

Issue DOI / title de-duplication list

2026-11-06 · Issue 4

Focus window: 31 October–6 November 2026. De-duplication used the current full Station, Section L’s 21 base sources, the three previous issues and the 12-entry M index, matching DOI, normalized titles and versions. Three new items below, plus one same-study publication update; all full texts read. arXiv inclusion or a conference listing alone does not establish verified peer review. Agency, felt authorship and control are distinct from legal ownership; no copyright determination is made. All proposed applications and youth questions are YuCROWN inferences, untested by these papers.

Issue de-duplication list

2026-10-30 · Issue 3

Theme: distinguish the ability to evaluate an output, question a system, decide which cognitive function to delegate, monitor one’s own thinking, and revise generic advice through situated experience. This issue checked work newly published or substantively updated from 24–30 October 2026 and de-duplicated it against Sections A–L, Issue 1 (18 October), and Issue 2 (23 October) by DOI, normalized title, and preprint/published-version relationships. The four entries are new, non-duplicate, peer-reviewed studies with full text read. Only Item 1 directly includes minors; transfer from Items 2–4 to youth settings is limited. “Sense of authorship/control/psychological ownership” refers to subjective experience and is not copyright or other legal ownership. None of these studies determines legal rights. All Earth Library / Study implications and youth explorations are YuCROWN inferences, not original research findings.

Issue de-duplication list

2026-10-23 · Issue 2

Theme of Issue 2: technical vocabulary, iterative control, phase-specific performance, and authorship/responsibility roles are different variables and need to be observed separately. "Psychological ownership" and "sense of authorship" below refer to subjective experience in specific studies or a conceptual analysis; they are not legal ownership judgments.

Search window: we focused on checking content newly published or substantively updated between 2026-10-19 and 2026-10-23, and de-duplicated it one by one by DOI, normalized title, and version relationship against Sections A–L of this page, the 21 base sources, and the 4 items of the 2026-10-18 first issue; the 4 items below have no duplicates. In this issue, 2 items have full text (items 1 and 2) and 2 currently have the abstract only (items 3 and 4) — the full text of the latter two was not read, and their methods, measures, and argument details are unverified. No older papers were used as filler. This issue has no sample of minors. Every "implication/exploration" is a YuCROWN inference / candidate exploration, not a finding of the original study.

DOI / title list for this issue

2026-10-18 · Issue 1

Theme of the first issue: a more polished work, feeling more capable, and keeping one's own expression need to be observed separately. The creativity and ownership measures below are measures within specific studies and are not legal ownership judgments.

Main search window 2026-09-18 to 2026-10-18; this is a scoped selection, not a systematic review. All 4 items were read in full text (4/4), publication status was verified, and they were checked against Sections A–L and M of this Station with no duplicates found (no earlier issues exist). No opinion pieces or non-peer-reviewed preprints are included, and no type quota was padded. University/adult samples cannot stand in for minors. Every "implication/exploration" is a YuCROWN inference / candidate exploration, not a finding of the original study.

De-duplication index

This issue updates the Research Watch column only. It does not update the base reference research date (2026-10-03) and does not change the dates or verification statuses of Sections A–L or the 21 base sources.