Violence, conflict, crime, disaster and public-safety evidence.
Transparent methodology
Evidence before opinion.
KINGAI separates measured country indicators, live-risk signals, media coverage, attributed external assessments and evidence-confidence metadata. AI may organize evidence, but it does not invent measured country scores.
透明方法论
证据优先于观点。
KINGAI 将国家结构指标、实时危情、媒体报道、外部机构评价与证据覆盖置信度分开处理。AI 可以整理与比较证据，但不能凭主观判断编造国家量化评分。
1. Structural indices and live risk
The public product separates relatively slow-moving structural indices from fast-moving crisis signals.
Life evaluation and objective development context such as health and living conditions.
Attributed rights datasets and documented rights evidence.
Press freedom, freedom of expression and network-censorship evidence.
Rule of law, accountability, effectiveness and related governance indicators.
Recent conflict, unrest, disaster, political crisis and abnormal risk signals. Higher means more immediate risk.
A missing value stays missing. Data absence is never treated as good performance.
1. 结构指数与实时危情分离
长期变化较慢的国家结构指标与短期突发风险分开计算和展示。
暴力、冲突、犯罪、灾害与公共安全证据。
生活评价与健康、发展、生活条件等客观背景。
可归属的人权数据集及可追溯的人权证据。
新闻自由、表达自由与网络审查证据。
法治、问责、政府效能等治理指标。
近期战争、骚乱、灾害、政治危机与异常风险信号，数值越高代表即时风险越高。
缺失数据保持缺失，绝不把“没有数据”自动解释成“表现良好”。
2. Evidence classes
- Primary/official records
- Intergovernmental records
- Peer-reviewed or technical measurements
- Methodological civil-society indices
- Media reports and public discourse
- Government responses and self-assessments
Claims, allegations, measurements, official positions and verified facts are not collapsed into one label.
2. 证据类别
- 一手与官方记录
- 政府间组织材料
- 同行评审或技术测量
- 有明确方法论的民间机构指数
- 媒体报道与公开舆论
- 政府回应与自我评价
事实、指称、测量结果、官方立场与争议信息不会被混成同一种结论。
3. Cross-source corroboration
Similar reports are clustered across domains. Multiple independent domains increase source breadth, but breadth is not proof of truth. Syndication detection reduces the risk of counting one wire-service story copied by many sites as many independent confirmations.
3. 多来源交叉验证
系统将相似报道跨域名聚类。多个独立域名可以提高“来源广度”，但来源广度不等于事实已经被证明。转载检测会降低“同一篇通讯社稿件被大量转载却被误算成大量独立印证”的风险。
4. Media and public social content
Raw news, videos, posts and open-web content are evidence/discourse signals with direct country-score weight 0. Ownership labels describe source context, not truthfulness. State media, private media, public broadcasters, independent outlets and advocacy organizations are preserved side by side.
4. 媒体与公开社交内容
原始新闻、视频、帖子和开放网页只作为证据或舆论信号，对国家评分的直接权重为 0。媒体所有权标签用于描述来源背景，不用于自动判定真假。国有媒体、私营媒体、公共广播、独立媒体和倡议组织会并列保存。
5. Evidence coverage confidence
A separate 0–100 coverage score describes how much public evidence is available for a country across background data, independent media domains, institutional assessments, network measurements, recent events and source diversity. It is not a country quality score.
5. 证据覆盖置信度
独立的 0–100 覆盖分用于描述某国当前公开证据是否充足，包括背景数据、独立媒体域名、机构评价、网络测量、近期事件和来源多样性。它不是国家优劣评分。
6. Equal rules
China, the United States, Russia, India, European countries and every other entity are evaluated under the same source-class and scoring rules. Country-specific observatories may exist for depth, but they cannot change the core methodology.
6. 同一套规则
中国、美国、俄罗斯、印度、欧洲国家以及其他所有国家和地区适用同一套来源分类与评分规则。可以建立国家专题观察页，但不能因此改变核心方法论。
7. Versioning and reproducibility
Every published score should be traceable to a methodology version, source snapshot and generation timestamp. Methodology changes are versioned in Git. Historical values should never be silently rewritten to fit a new method.
7. 版本化与可复现
每个公开评分都应能够追溯到方法论版本、数据快照和生成时间。方法论变更通过 Git 版本记录；历史数值不应为了适配新方法而被无提示改写。