I think it is wonderful that My Heritage (and others who extract annual subscriptions from family history researchers) seek new and exciting uses of information technology to reveal previously undiscovered insights that will justify the expense. But the sad reality is that occasionally the cutting edge of technology can leave scars.
Today I received another exciting revelation of a "new" Theory of Family Relativity. Sadly, its chain of reasoning depends upon the same invalid assumption (that my great-grandmother who migrated to Australia in 1881 was recorded in the UK 1891 Census, presented with 27% confidence) as the previous effort. Feeling that it would be unfair to damn a whole program on the basis of two tiny misteps (well technically, the same misstep twice), I resolved to review all twelve offerings.
Only three of the other ten "theories" included that same logical break (but one of those did offer three internal options, all of which fell at the same obstacle). That is, 41% of the output of this program (in my case) is invalidated by exactly the same flawed reasoning.
Five other theories each relied upon a match for Jane in an earlier census (although not a correct match) to bridge back to a parent who was apparently buried in the United States (although he definitely died in Australia). For those keeping score, that leaves two "theories" (16%) that match connections that I had already made through old-fashioned document analysis.
No doubt the experience of others will differ. I felt terribly guilty about having not drawn these shortcomings to the attention of My Heritage when the first message arrived; in order to save them the time, expense and embarassment of continuing to rely upon false reasoning.
Even in My Heritage does not care (although "perhaps" their view will change), I still feel that the blame for this mess rests with me. If I had uploaded a more comprehensive tree, their algorithm could select different starting points and hence have the opportunity to make different errors instead of repeating the same nonsense over and over. There is surely a minimum level of data below which the type of extrapolation being undertaken has a very high probability of error. Since I have not met that threshold, I ought not be entitled to these high-level services.
Why not place a warning on each member's tree that is likely to generate erroneous "theories" to give the owner the option to improve the tree to the required minimum or to lose the privilege of receiving (what will almost certainly be) speculative works of family fiction.
As an interim measure I have taken the only path offered and "updated my preferences" to stop the email notifications of each new Theory. I imagine that they will still be generated and attached to my DNA "matches" but at least I can ignore them there and reduce the impact on my blood pressure.
1 https://faq.myheritage.com/en/article/what-if-a-theory-of-family-relativity-is-wrong
No comments:
Post a Comment