Every case in the JusticeLens dataset comes from at least one of three sources: (1) Korean Supreme Court 종합법률정보 case search results, (2) Sentencing Commission 양형위원회 published documents, or (3) major-outlet news coverage from 한겨레, 경향신문, 조선일보, 중앙일보, 동아일보, 연합뉴스, MBC, JTBC, KBS, BBC Korean, and equivalent. Where we have a case number, we cite it. Where we have a news URL, we cite it.
We do not collect statements from defense counsel, prosecution offices, or trial transcripts that are not public. We do not interview the defendants or their families. We do not have access to unpublished court records.
This means the dataset is **survey-quality, not adjudicatory**. We pair cases that public reporting describes; we cannot rule out factual nuances that the reporting omitted. Where we believe the reporting may be incomplete or contested, we say so on the case page.
We do not argue that every individual case was decided incorrectly. Many are within guideline ranges. Many can be defended on facts we don't know. The thesis of the project is statistical and structural — that across many cases, the same mitigators flow systematically toward the same kinds of defendants. We hold that thesis up to the data; readers are invited to disagree.
We especially welcome corrections. If a case is mischaracterized, if a number is wrong, if a citation is broken — submit it through /take-action and we will review every submission.