I’ve blogged about AI in genealogy before, and the latest tool offered by Ancestry makes me even more cautious about the current use of artificial intelligence in genealogy research.
When searching the 1911 England and Wales census, there is a button by the side of the search results which says “Learn what this record has to say with Ancestry AI”. I looked up a someone with my uncommon surname to see what it was all about.
Ancestry AI Assistant Example
Evelyn Constance Arnaud was born in New Brunswick, Canada in about 1876 and married into the Hewat family. By 1911 she was widowed, and living in Leamington Spa, Warwickshire in the household of a distant relative, Richard Warner Brooks, and gave her occupation as “lady housekeeper”. Richard was listed as a retired Army major, and an import/export merchant. There was a nurse, and three other servants living in the household.
So let’s see what Ancestry’s AI has to say about Evelyn and the 1911 census. Some of the content the AI generates is reasonable. Statements such as “She was a relative of the head of the household, Richard Warner Brooks, a 65-year-old gentleman. The nature of their relationship was not specified, but it hinted at a close familial bond” are probably reasonable, as is “The presence of a nurse, Lillia Margaret Freeman, suggested that someone in the household required special care or attention”.
However, some of the assumptions and conclusions drawn by the AI are frankly staggering. It says:
· Martha Brown, a 30-year-old servant, diligently attended to the needs of the household.
· the ensemble of dedicated staff members.
· Evelyn, with her experience and expertise, managed the daily affairs of the household.
· relied on Evelyn's capable hands to ensure everything ran smoothly.
· a testament to the compassion and consideration shown by Richard and Evelyn towards their family members or guests.
· Evelyn's presence, along with the dedicated staff members, created an atmosphere of order and care
There is no evidence in the census for ANY of this. How do we know that the servants were diligent and dedicated? We also don’t know whether Evelyn had expertise and experience, or whether she just left all the household organisation to the staff. The presence of a nurse in the household definitely does not allow us to draw any conclusions whatsoever about “compassion and consideration”, or that Evelyn’s presence created “an atmosphere of order and care”. These are value judgements, subjective opinions which can never be drawn from words on a page.
This is the real danger of these AI tools being pushed as the next big thing in genealogy. Users are going to see the button, click on it, and take what they are being told as truth, despite the little warning at the bottom that responses may be inaccurate. The AI model is not sophisticated enough to look at other records to draw conclusions, or the use Ancestry’s own DNA database. I have another AI generated report which states confidently the parentage of a child on the 1911 census, a direct contradiction from the DNA test done and hosted on Ancestry’s own site.
We all know that a large number of user-generated trees hosted on Ancestry are complete nonsense and this AI “assistant” is just going to add to the problem. Do your own research. Use the records. Yes draw conclusions about someone’s social standing and income levels from details given on a census. But we can NEVER attribute personality traits from this sort of data. Take it all with a massive, massive pinch of salt.
Comments