Yesterday, the FamilySearch contributor made dozens of changes (and errors) to people I follow.
Yesterday, April 22 2024, the FamilySearch contributor ("Change made by authorized support staff or as part of an update") made dozens of changes to people that I follow, creating lots of problems. Typically, it would attach a source ("Computer Generated Trees," entry for Pedro Antonio Benedetti.) and then add children to the family. In the first case that I dug into, there was an existing child named Leon Benedetti (GM6R-1D8), born 3 August 1864. The FamilySearch contributor added two additional children, both named Leon Benedetti with no other information but a birth year (1865 and 1868). It then married one of these duplicate brothers to the first wife of Leon, and married the other duplicate to the second wife of Leon. So now there's a bit of a mess to clean up just on this one case. And there are dozens of these cases that I need to now go check and fix.
Automation is great until it goes awry, and that's what seems to have happened here. Does anyone have any insights on to what's going on here?
Update: Upon looking at more of these cases, in many instances, the only change was attaching the source ("Computer Generated Trees …"), or attaching the source and adding an alternate name. In these cases, no harm no foul. But in the cases in which children are added, I'm really concerned.
Best Answers
-
This appears to be a similar problem to that I am having with 'Search > Genealogies'. If you do a search you should be able to find the thread.
Go to:
Then scroll down a bit to see the Collections. The 3rd one is 'Computer-Generated Trees'. I think that is where that source came from. Both Sources and People are being added.
1 -
'A computer scanned more than 35 million records to create a family tree of people who lived in Argentina.' What records? Are they primary sources, and how good are they? Why aren't they identified on the citations? How legible were they? How well did the process cope with family relationships? So many questions.
I'm shocked that this information is being automatically pulled into the Tree (quantity over quality), especially if the de-duplication is as imperfect as it clearly is for gedcoms.
The computer generated tree as a separate thing does sound potentially useful in providing hints.
3 -
I have been trying to raise a separate query on the whole issue of Computer-Generated Trees, but due to the current state of the Community platform (can't paste screenshots and, intermittently, not even URLs) I had to give up and post here.
Would @Maile L or another moderator kindly get some background detail on what is going on here?
Firstly, from comments above, there appears be be a way, through a FamilySearch project, to add details from these trees directly to Family Tree IDs.
Specifically, can we kindly be told what this is all about? See https://www.familysearch.org/ark:/61903/2:5:7FZ8-SB8.
You will see in the Citation (possibly - the page crashed when a tried to view a further generation!) - "submitted 23-12-2007 by US CensusProject". So that "explains" the married-name surnames for some of the spouses, but where is the connection between Dorothy and Joseph Wrightson and their "children" (and their spouses), with surnames White and Kinsley?
And concerning FamilySearch's apparent new determination to "rate" its records, this particular collection is headed as having "MODERATE DATA ACCURACY". I find these ratings (together the ones being trialled as "Quality Scores" against profiles in Family Tree) to be very concerning, as I fear this will provide inexperienced users with a false sense of security about the quality of the data they are being presented with on the website.
FamilySearch users really need explanations of these newly-found behaviours and settings. In particular, of how we can all now attach records from at least one section of Genealogies** as sources to Family Tree profiles, whereas only a particular FamilySearch project team appears able to attach details from Computer-Generated Trees to FT profiles.
Perhaps there is an article on the whole issue somewhere on the website - if so please can a moderator please direct us to that? Otherwise, please help to clear up the current confusion on the apparent "new functionality" FamilySearch has decided to apply to collections under the Genealogies heading.
** That section is "Community Trees" (HIGH DATA ACCURACY rating) - not to be confused with "Computer-Generated Trees (MODERATE DATA ACCURACY rating). See https://www.familysearch.org/ark:/61903/2:2:3WQP-JFX, which illustrates ordinary users can attach these records as sources to Family Tree IDs.
4
Answers
-
@Tiffany Farnsworth Nash Yours were added by a normal user, though. These were added by FamilySearch themselves. @ShawnFReid may be right about the automation, because it isn't hard to see that the tree in question is wrong—there are two children with the same name, and must have lived within each other's lifetimes, because they both have a wife listed and were only born a few years apart.
Alternatively, the person who put them in could have trusted the regular users to do the merges.
1 -
I have no problems with computer generated trees. And receiving a hint that one of these generated trees may be connected to a particular person/family and allowing me to curate any potential change would be fine. But to have an automated system make the updates seems like a step too far. I can't message the FamilySearch contributor and ask them about the change they just made like I can other contributors. It just keeps making the same mistakes over and over.
1 -
WOW, good luck. So much for "Robotic Genealogy" or "Artificial Intelligence" I am glad you were "following" or otherwise you may not have of ever noticed.
0 -
I've just done a bit of work in my current set of extracted FS data.
I have just over 250,000 Sources (relating to around 40,000 Persons and their couple and parent/child relationships).
Of these, 28 point to Genealogies ('/61903/2:' in the link).
Obviously it's possible to create a Source pointing to any URL anywhere, so I can't know which of these 28 were created via the 'official' Source Linker route.
However, when I open the links, the 'Attach to Family Tree' button appears only (in my case) where the Genealogy entry is in a 'Community Tree' or an 'Oral Genealogy', but not otherwise (e.g. 'Ancestral File' entries don't get the 'Attach' button). Thus 'Attach' seems to be allowed only for those Genealogies that are considered to be of 'HIGH DATA ACCURACY'. Which would make sense.
Incidentally the 'Attach' button seems to appear, where allowed, whether or not the entry is already 'attached to Family Tree', but that's a separate problem.
1 -
"submitted 23-12-2007 by US CensusProject"
So we have spent hours, days, and weeks cleaning up the duplicates created by the USCensusProject only to have their "contributions" return.
3 -
Albeit they are appearing in a different form. As you have illustrated, there is a 2007 date recorded, so hopefully no "new" records are being added to the website.
I was always happy for the data to be published in a form referenced by Professor Joe in his YouTube video. That showed how spreadsheets were used to gather data on whole communities and appeared useful in providing family / neighbour connections. But this "Computer-Generated Trees - Census Trees" section is presenting data in a quite confusing manner. As illustrated in the example I reference in my previous post, we are given to believe that a couple named Joseph and Dorothy WRIGHTSON had children with surnames of WHITE, KINSLEY, and BARRIS or BOWERS. The pedigree chart is most unhelpful and probably completely inaccurate - yet we are invited to believe the detail under this heading contains "moderate data accuracy".
So, I'm not so much worried about the duplication factor here, but the way individuals are being incorrectly connected in family pedigrees.
3 -
I can only imagine that FS' Data Quality team have not been given the strategic teeth to stop these loose cannon projects in their tracks.
Maybe when all these bulk inserts start affecting PQS scores someone might start joining the dots. (Let's just hope that Genealogy data sources, especially 'MEDIUM DATA ACCURACY' and lower, don't result in PQS points.)
0 -
Not sure if you saw this or not…
"Specifically, can we kindly be told what this is all about? See . "
On this page, if you click "SHOW MORE DETAIL" it provides a link to "Census Tree" :
1 -
Tiffany's link in usable form. The community platform still does not play well with URLs.
4 -
Thanks for the correction. You don't know what you don't know- until you know it! LOL
0 -
@Tiffany Farnsworth Nash any time you put the URL on a line by itself, the platform software replaces colons with %3A. It's a quick fix once you've done it few hundred times.
3 -
Curious. Until now, I've tried to avoid going into the Census Tree project because the end results in FS Family Tree are so depressing. For no particular reason, I decided to have a look at https://www.censustree.org/ today and was slightly surprised. What follows may either be wrong or a statement of the obvious to some of you but just in case…
On the Census Tree site, I could see no direct mention of any updates to FS FamilyTree. It seems all about producing separate stuff to enable:
academic researchers to follow men and women throughout their lives, and to connect people to their family members and others. These links are essential for [academic] research …
However, there is a PDF on the site entitled Breakthroughs in Historical Record Linking Using Genealogy Data: The Census Tree Project. Page 9 of that PDF refers to FS hints suggesting Historial Census Records that can be attached to Profiles, and a second sort that directly links two or more Historial Census Records together. Both are in the context of developing the person/person linkages in the Census Tree Project. It then says:
We have developed several tools that allow volunteers to validate both types of hints by attaching records to profiles on the Family Tree.
(My emphasis). This is the only reference I could see in that paper, or on the site, to making updates to FS FamilyTree. It suggests the project identifies samples of census-related hints, and updates the corresponding profiles in FS FamilyTree - not by applying FS FT hints but by some more direct updates - then does some sort of validation assessment on the profiles(?) (no idea how).
This might explain why I've never seen any of the US Census Tree updates on my American relatives' profiles - they simply haven't come up in the samples as a suitable (or unsuitable) test. But those of you who have been affected, appear to have just cause to complain that this so-called validation is making the situation worse. Even assuming that the validation is, ahem, valid.
This might also explain the inaction by FS Management - they believe that any updates are small in number, solely for hint validation purposes, whereas the volume is anything but small if you are affected.
Curious - as I say, this might be obvious or plain wrong…
4 -
FamilySearch Genealogies is an area of FamilySearch that I was previously unfamiliar with. So thanks to those who have pointed me in that direction. Of the thousands of profiles that I follow, the only ones that were changed by the FamilySearch contributor are those of distant cousins in Argentina. (I wonder how patrons from Argentina are feeling about this).
I just wish that this seemingly automated process followed the "Hint" pattern such that patrons who are familiar with these families can curate the potential changes.
FYI, the FamilySearch contributor changed 154 profiles that I follow. Those changes break down in the following way:
28% - simply added the source: "Computer Generated Trees," entry for <name of person>
14% - added one or more alternate names
7% - added or updated one or more events (i.e. birth)
51% - added one or more new profiles (usually children, usually just one but sometimes as many as 6 or 8)
I haven't had time yet to look into all of these changes, but of the two that I've researched, one provided some valuable new information 🙂, and one was totally wrong and will need some effort to clean up 😒.4 -
If you scroll down the right side bar under the person's name you can see the sources attached to the trees.
Click on view sources and it leads to a page that will show you what sources were used to create that tree.
0 -
Thank you for your comments, which have at last made sense of the confusion I was having in accepting what I had recorded in earlier posts:
"… we are given to believe that a couple named Joseph and Dorothy WRIGHTSON had children with surnames of WHITE, KINSLEY, and BARRIS or BOWERS."
Viewing the detail behind the chart I can now see that the females with these different names ((recorded from US Census returns) were indeed children of Joseph & Dorothy Wrightson - but all showing in their married names (as is mother Dorothy, of course).
This further illustrates my point that these projects are fine as "stand-alone" efforts, but once added to Family Tree (either directly or via Computer Generated Trees) they break one of the main pieces advice that is given to users in entering the names of females: never to use a married name, but always the maiden one.
As already mentioned, these trees are going to cause many more duplicates being added to Family Tree, as well as leaving extra work to users in having to confirm and change all those married names being imported to the tree to the correct maiden ones.
I now understand the point being made by @Áine Ní Donnghaile in her comment:
"So we have spent hours, days, and weeks cleaning up the duplicates created by the USCensusProject only to have their "contributions" return."
1 -
Many thanks for your advice.
The OP's link does indeed lead, if you click on the individual entries listed on the Computer Generated Trees source detail page you mention, to proper FS Record sources in the 'Argentina, Mendoza, Catholic Church Records, 1665-1975' collection, e.g. /ark:/61903/1:1:QVN9-CHJC.
So, why do the Sources that get added to FT by the auto-population activity not mention these genuine Record sources? They are surely far more useful than a Genealogy marked 'MEDIUM DATA ACCURACY', and finding them is in my view pretty opaque to those of us who are not that familiar with the Genealogies UI.
Also, the Records that I have reached in this way all seem to be restricted (the dreaded padlock icon) and do not therefore appear in searches of the relevant collection - why not?
I still feel users could do with a lot more clarity here.
3
