Identity Resolution

Identity resolution is the process of deciding when separate records, mentions, profiles, or database entries refer to the same underlying person, household, customer, patient, company, or other entity. In practice it is the work of reconnecting fragmented evidence without collapsing genuinely different entities into one mistaken match.

Why It Matters

Real-world data is messy. The same person may appear under variant spellings, changed surnames, partial dates, inconsistent addresses, multiple IDs, or different supporting documents. In genealogy that might mean one ancestor appears across a baptism register, two censuses, a marriage record, and a newspaper obituary with slightly different details. In customer data it may mean the same person exists as several profiles. In healthcare it may mean one patient is split across systems.

Good identity resolution helps AI systems build more complete histories, reduce duplicates, improve retrieval, and make downstream recommendations or analyses more trustworthy.

How AI Helps

Older systems often relied on brittle exact-match rules. Modern AI and machine learning can weigh partial matches more flexibly by considering names, dates, places, relatives, co-occurring entities, source quality, and graph structure together. The strongest systems do not hide the uncertainty. They expose confidence, supporting evidence, and review paths so humans can approve difficult matches.

Identity resolution overlaps with record linkage, but the term is slightly broader. It often includes cross-profile deduplication, entity clustering, and long-lived identity management rather than only pairwise matching between records.

What To Watch

The biggest risk is the false merge. A missed match can slow research. A wrong match can corrupt a tree, a customer profile, or a patient history. That is why high-stakes identity resolution usually needs provenance, thresholds, and human review instead of silent automatic merging.