A careful approach to unusual information

Large collections of public information may contain relationships that are difficult to notice by hand. Software can help people examine those relationships, but it can also create false confidence when calculations and interpretations are presented as facts.

AnankeNet is designed around a stricter idea: preserve what was observed, explain what was derived, and keep judgment human.

What the research explores

The prototype studies whether an evidence-first system can:

• preserve source material and its provenance;
• make unusual patterns easier to examine;
• keep facts separate from calculated and probabilistic output;
• expose the reasons and uncertainty behind analytical results;
• support accountable human review and disagreement.

What it does not decide

AnankeNet does not determine guilt, establish criminal conduct, or identify victims or offenders. It is not designed to contact, confront, or investigate marketplace users. It does not replace qualified reviewers or competent authorities.

An unusual result is a candidate for examination. It is not a conclusion about meaning, intent, or conduct.

Why evidence comes first

Analysis can be repeated or improved. A lost or altered source cannot be reconstructed with the same confidence. The research therefore treats preservation, integrity, and provenance as prerequisites for useful analysis—not as administrative work added later.

Human accountability

Reviewers must be able to inspect the source, understand the reasoning, see uncertainty, and disagree with the system. Automation may organize information and raise questions. It does not make the final judgment.

Find patterns. Preserve facts. Explain uncertainty. Let qualified investigators determine what happened.