Korea Occupation Direct Crosswalk v1

What it is

All 1,270 KSCO-8 unit groups are linked directly, not via ISCO-08, to an occupation master of 1,023 nodes. The nodes are the 1,016 O*NET-SOC codes in the O*NET® 31.0 Database (O*NET-SOC 2019 taxonomy) and 7 Korea-specific "other occupations" (ACW-KR-NNNN) with no O*NET-SOC counterpart. Relations follow the definitions of the ESCO–O*NET crosswalk (2022) and always read from the O*NET-SOC side.

Three AI judges worked independently: two Anthropic Claude runs using different evidence, and one OpenAI Codex run. Their answers were combined by majority vote, and the operator reviewed and approved the 61 links where the judges did not agree.

This is AI Changing Work's judgment mapping. It is not a concordance from USDOL/ETA, Statistics Korea, or the Ministry of Data and Statistics.

• The links do not pass through ISCO-08 (International Standard Classification of Occupations). Going national classification → ISCO → O*NET-SOC code crosses two steps, and the approximations of each step compound. The occupation descriptions of the two classifications were compared directly.
• Korea-specific occupations with no O*NET-SOC counterpart were not forced into a match; they were added to the master as other occupations (ACW-KR-NNNN).
• Relation definitions follow the technical report of the ESCO–O*NET crosswalk (European Commission × U.S. Department of Labor, 2022), and relations always read from the O*NET-SOC occupation side.
• Linking other countries' classifications directly to the same master would turn it into a cross-country network.

Key figures

Occupation master

1,023

1,016 O*NET-SOC codes + 7 other occupations (Korea)

KSCO-8 unit groups linked

1,270 / 1,270

0 unit groups left unlinked

Links

2,075

match 1,620 · related only 455

Status

1,467 · 547 · 61

confirmed · provisional · reviewed

Of the 1,016 O*NET-SOC occupations, 867 have a match with a Korean unit group, 148 have related links only, and 1 has none (Coroners).

How it was made

1. Forward mapping — 1,016 O*NET-SOC occupations → KSCO-8 unit groups (up to 3 per occupation)
2. Relation review — the 467 pairs labelled related in step 1 → exact / narrower / broader / close / related
3. Orphan reverse mapping — the 495 KSCO unit groups that no O*NET-SOC occupation pointed to in step 1 → O*NET-SOC occupations (up to 3 per unit group). Where a majority of judges found no fitting occupation, the unit group was added as other occupation ACW-KR-NNNN
4. Review — links without judge agreement were reviewed and approved by a person

Step 3 is the same step as the "orphan validation" of the ESCO–O*NET crosswalk. The forward mapping starts from the O*NET-SOC side, so unit groups where KSCO is more finely divided (private-academy instructors by subject, jobs specific to Korean institutions, and so on) are never chosen and are left over (495 of 1,270, 39%). Judging them once more from the opposite direction connects every KSCO unit group to some node of the master.

• The three judges did not see each other's answers or the labels from earlier steps.
• The reference for judgment was the English titles and definitions on the O*NET side and the Korean source text on the KSCO side. In step 1, O*NET tasks were also shown as a supplementary AI Changing Work Korean translation, with an instruction not to use it as the reference. In steps 2 and 3, only the English O*NET source text (definitions, reported job titles, tasks) was shown.
• The published files contain codes and titles only. They do not contain KSCO explanatory-manual text or O*NET definition and task statements.

• In step 3, Codex judged all 103 unit groups where A and B disagreed, plus 60 randomly drawn unit groups where they agreed (fixed seed for reproducibility). Reliability of provisional links: in that random 60, Codex chose a different top occupation from the A–B consensus in 22% (13).
• Cross-model check: agreement between two runs of the same model was not strong evidence. Of the 341 pairs where Claude A and B agreed in step 2, Codex gave a different relation for 115 (34%).
• Human review: the links without judge agreement (16 split · 45 partial) were reviewed and approved by the AI Changing Work operator on a side-by-side source comparison screen (2026-09-25).
• Rule checks: 0 violations of "at most one exact per O*NET-SOC occupation". 0 of the 1,270 KSCO unit groups left unlinked.

Limitations

• This is a judgment mapping. It has not been validated against employment statistics. The two containment shares (o_in_k, k_in_o — the judges' estimate, made during relation review, of what share of workers in one occupation fall within the other side's unit group) are the judges' estimates, not measured headcount ratios.
• The 547 provisional links did not go through cross-model validation. Based on the sample, about 22% may differ from the cross model in their top choice.
• The judges are AI models. Human review was limited to links without agreement.
• KSCO task information is mostly at the 4-digit level, so boundaries between 5-digit unit groups rely heavily on descriptions and example jobs.
• Military occupations (KSCO major group A) are divided by rank, while O*NET-SOC occupations are divided by branch, so close and related links are common and confidence is lower.

Attribution required when you share this data

AI Changing Work (2026). Korea Occupation Direct Crosswalk v1 (KSCO-8 ↔ O*NET-SOC 2019). https://aichanging.work/data/kr-direct-crosswalk. Licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/).

Terms of use and sources

Below is the full text of the package's LICENSE file, reproduced as published. In short: AI Changing Work's contributions (links, relation labels, match class, stage, consensus and status values, and the ACW-KR-* nodes with their English names) may be used under CC BY 4.0 to the extent AI Changing Work holds rights in them. O*NET-SOC codes and titles remain under the O*NET® 31.0 Database license (CC BY 4.0) with the attribution below, and KSCO codes and titles are items of a Korean government notice that is not protected by copyright.

The English names of the other occupations (ACW-KR-*) are AI Changing Work's translations, not English names set by the Korean government.

Download

These files are byte-identical to the published package. Use the SHA-256 values to check a downloaded file.