Unraveling the Past: AI and the Evolution of Genealogy

My mother was born and raised in American Samoa, my father is Mexican American, and I grew up in San Diego, my birthplace. Frameworks of education and culture in the United States influence my understanding of Samoan culture, so I often approach this part of my family history with the perspective of an outsider. My outsider’s viewpoint highlights challenges of genealogy for Samoan heritage: limited written records, different practices of family formation, cultural and language barriers, and complicated genetic relatedness.

Recently, I excitedly discovered that technology, including AI, has helped the field of genealogy evolve, and these new tools potentially provide traction for overcoming challenges like mine. I spoke with four women who are professional genealogists for insights on these significant, emerging advancements in genealogy, as well as about ethical considerations for responsible use and options for those who want to get more involved with genealogy themselves.

Chaos Overlaid with Statistics

Much of genealogical research involves extracting insights from extremely large datasets. AI has significantly advanced work in the field by automating tasks such as organizing, analyzing, and identifying patterns of complex information in handwriting, photos, historical records, and DNA ancestry tests. This automation complements the ability of people to provide context and interpretation to their research. AI is especially important for techniques that combine DNA analysis and traditional genealogy research to identify homicide or sexual assault suspects in criminal investigations and to solve cases of unidentified human remains. These techniques are called Forensic Genetic Genealogy (FGG) or Forensic Investigative Genetic Genealogy (FIGG).

Colleen Fitzpatrick, founder of Identifinders International, is the pioneer of FGG. She worked the first cold case using FGG in 2011 and solved the first cold case using FGG in 2015. As a leader in the field, she has seen it grow from its infancy to a game-changing tool in human identification. As the amount of data available to FGG has increased enormously, she believes that AI can bring order to that data to streamline the successful application of FGG to cold case work. She says, “People are sort of chaotic. The value of AI is in overlaying that chaos with statistical models trained on massive datasets, enabling insights we wouldn’t be able to see with manual analysis alone.”

University of New Haven Forensics Professor Claire Glynn explained the significant impact of accessing digitized records through the use of AI. Previously, retrieving these records required contacting individual offices or paying fees for mailed paper copies. This process caused a major bottleneck in some forensic genealogy investigations. According to Glynn, “We may need to collect and review and document records going back hundreds of years in order to build the family trees we need.” Platforms like Ancestry.com, the US National Archives, and international databases now house billions of digitized documents, including civil registration records (related to birth, death, marriage, and the census) that span centuries and continents. AI tools usefully digitize records, transcribe handwriting, and translate different languages, providing accessibility that addresses the genealogy bottleneck issue.

Another pioneer in genealogical research, Laura Olmsted, serves as Executive Director of DNAngels, a nonprofit organization that she founded and which supports, guides, and assists adults who seek to accurately identify their biological family through analyzing DNA results and genealogy. Her work also benefits from advancements in natural language processing that allows AI to scan, interpret, and transcribe historical documents. FamilySearch Labs is a website that makes experimental AI features available to users of family history services provided by the family history resource organization FamilySearch. “FamilySearch Labs have natural language processing to extract data from digitized records in over 100 languages, which really reduces the time required to transcribe and analyze historical data. In one case, AI-assisted transcription helped a team solve a lineage mystery by revealing connections hidden in fragmented records. These tools enable genealogists to access more comprehensive and accurate information than was possible even a few years ago,” she stated.

Olmsted elaborated by presenting an overview of AI DNA analysis applications for genealogy. “In DNA analysis, machine learning algorithms help predict family relationships more accurately by analyzing the genetic markers and by clustering DNA matches into family networks. For example, clustering algorithms are used in platforms like MyHeritage and Ancestry that can now categorize DNA matches into maternal and paternal lines, which speeds up identifying unknown relatives,” she explained.

A fourth expert, Nicole D. Price, PhD, serves as an adviser on the DNAngels Board of Directors and also as CEO of the leadership development company Lively Paradox, which she founded. She described the impact of economic accessibility related to advances in DNA testing services. “The ability to even provide this kind of service at [these prices] so that consumers can benefit is only made possible because of AI.”

A Genealogy Empathy Revolution

Price emphasized a concern she has that relates to testing results, a concern that motivates her to push for what she calls an empathy revolution. “Empathy in the STEM fields is important to me,” she said. “But the way that these testing services, like AncestryDNA and 23andMe, are marketed as a fun way to find out about your ethnicity estimates or genetic predisposition, without [providing] basics about what to do in cases of upsetting discovery — like, ‘Wow, it looks like neither of my parents genetically relates to me,’ — [is worrisome.]”

She continued with a personal story: “Both of the parents who raised me were deceased by the time I found out that my biological father wasn’t the man I thought he was, so I couldn’t ask them questions. I now had to go talk to someone I haven’t known for 45 years and say, ‘Hello, I’m your daughter.’ I was lost for four months, as was my therapist, about what to do. I had no idea that the DNAngels existed or that there were support groups for this kind of thing.”

Through ChatGPT, Price built her own ability to articulate and understand her new circumstances. “I described to ChatGPT what it was like to see my face on another person’s face, because I don’t look like my mother, so I never looked like anyone [I knew, as I was] growing up. And it was through ChatGPT that I learned that the term for looking like a relative is genetic mirroring.

“What I would like to see is the science fill some of those gaps in human ability around support services that specialize in genetics and the connection to self-actualization, even if it’s nothing but the basics: ‘You’re probably experiencing grief and a loss of identity, and you might consider this support group.’ I would like to see the larger outfits take responsibility for that.”

Price anticipates that AI will help support empathetic responses. She noted, “We care about people, don’t want to hurt their feelings, and want to help in the most emotionally intelligent ways. If by chance that’s not the natural way we communicate, asking AI for help can be a quick way to get opening talking points, suggestions for what to say, what not to say, what would be most helpful, what wouldn’t be. The traditional way of doing that is talking to our friends, but, for example as it relates to genetic discoveries, our friends or even therapists might not have the experience or training needed to understand [our emotions].”

As more individuals participate in genetic testing for genealogy and as technology for analysis advances, the likelihood of rapidly uncovering family secrets requiring special support for empathetic responses grows as people use DNA to match with newly discovered relatives. After a recent audit of cases, the DNAngels estimated the percentage of NPE (Not Parent Expected) events in the first generation of a family tree at 13%. Olmsted said, “It’s important to consider that this data comes from a specific population: individuals who often approach us because they already suspect discrepancies in their parentage. This selection bias likely amplifies the percentage compared to the general population. However, it does provide meaningful insights into cases where parentage surprises exist.” These emotionally complex discoveries require a compassionate approach to support all parties involved.

Price also noted a societal double standard means women have more at stake: men can have children outside of relationships with minimal social stigma, but women still face significant judgment and negative stereotypes. She also pointed out, however, that “women, because of socialization, are more empathetic and tend to be more compassionate.” These strengths help them in genealogy spaces, so they often serve as advocates for equality and understanding, leading efforts to address the societal norms and biases that perpetuate shame and silence around issues like parentage and familial responsibilities.

Ethical Considerations and Standards of Responsible Use

The women I spoke with each presented potential recommendations for responsible use of technology, including AI, in genealogy applications. Price and Fitzpatrick both pointed out that the primary ethical question around the terms of possession and use of a vast amount of genetic data. When individuals submit data to companies like 23andMe, Ancestry, GedMatch, or FamilyTreeDNA, they typically agree to terms of service that allow the company to anonymize and share the data for research purposes. While these companies often claim to prioritize data privacy, concerns remain. Users who agree to share their data may well misunderstand the terms of service.

Olmsted raised the example of FIGG used to identify a suspect following the killing of four University of Idaho students in November 2022. In this case, the search and arrest warrants did not mention FIGG, but its use in the case became a focus. Glynn noted that to ensure the integrity of a FIGG investigation and the admissibility of evidence in court, investigative teams must maintain strict ethical standards and meticulous documentation. This involves protecting the privacy of individuals, avoiding unauthorized access to genetic databases, and maintaining a clear chain of custody for all evidence. She said, “In protecting the integrity of the case, we won’t end up with wrongful convictions or cases thrown out because someone messed up somewhere. Victims deserve that, the families of those victims deserve that, and suspects deserve that.”