Continue with Hints already attached to someone else
edited September 28, 2020 in Suggest an Idea
Jean Steele said: I recently viewed a tutorial from the Product Managers where they stated future enhancements of the familysearch software. Specifically, they noted they were going to stop showing Hints that were already attached to people. I would ask you to reconsider this as I have found many duplicates, partial families already made and incorrectly attached sources to mixed up families or erroneously attached to a sibling. This is a great help to me in finding additional partially made families.
Jordi Kloosterboer said: Possibly have an extra tab for hints already attached to other people. or have an option to show those hints that people have to opt into explicitly.0
Carolyn Wheeler said: I so agree with Jean Steele. Please continue with hints that are already attached to someone else. At this point there are not enough of them to be a major complaint (at least for me anyway), and they are often the only way where I can find duplicates, incorrectly attached sources, and/or incorrectly formed families.0
John Huff said: Thanks for your feedback on this issue. The goal is to make it so that the "identification of duplicates, partial families already made and incorrectly attached sources to mixed up families or erroneously attached to a sibling" becomes easier by presenting the tree-profile-already-attached-to-a-historical-record and the tree-profile-with-a-hint-to-that-same-historical-record as Possible Duplicates of each other instead of presenting the unresolved hint.0
Juli said: That'll only work if the existing attachment makes sense. Sometimes, it just absolutely doesn't. For example, what if someone clicked on totally the wrong person (because someone jogged his elbow at just the wrong moment), and then couldn't figure out how to fix it? Would the algorithm now compound the problem by suggesting John Quigley as a possible duplicate for Ignatia Mendoza?0
Tom Huber said: Interesting conundrum, Juli. I would presume that pulling up the "possible duplicate" would result in the two being displayed side-by-side as they are now in the new merge user interface.0
Tom Huber said: At that point, one can open either profile and edit the sources. However, it would be better if any system that can do that would allow us to see the other person(s) attached, rather than just the merge.
I would expect the new system to be very astute in the way it handled attached records. Primary people in the source is understandable, but not the names of those in attendance for a funeral, such as unrelated pallbearers.0
Tom Huber said: Another problem just struck me and that is people in the record with the same name. Obituaries are a good example where this can happen. If the system under consideration found a case where two people with the same name were both recorded (and indexed), how would the system know that these were not the same person or were not (possibly) related.0
Tom Huber said: We certainly run into enough users who automatically merge "possible duplicates" just because the system suggested that two records were possibly duplicates of each other. I already saw (and fixed) that situation after the new merge system was implemented.0
Juli said: Obviously, the correct fix for my hypothetical jog-to-the-elbow is for the user to learn how to use Source Linker, but that's just one of thousands of ways that a source can end up attached to entirely and completely the wrong part of the tree, and why Possible Duplicates absolutely CANNOT replace correct hinting, and why hinting must NOT take existing attachments into consideration. At all.0
Adrian Bruce said: I really couldn't agree with the idea of continuing to show hints, that have already been attached to a profile, against another profile.The hinting mechanism can already come up with some dubious suggestions (particularly if the profile is only part complete) and if we assume that the first attachment of the hint was correct, then a dubious hint against a second profile is (a) redundant and (b) adds bad data to the second profile that makes it look more and more like the first.
Juli has a sound point about what happens when the first attachment of the hint is incorrect, but I can't help feeling that fixing an incorrect attachment through carrying on hinting, is the wrong way to go about things.
Let me follow up what happens when there is no hint because the hint has already been attached... If the user is an adept genealogist, then they will probably realise that the correct attachment is missing - e.g. the 1901 census isn't attached and nor is it being hinted. Then they'll do the old-fashioned thing and go searching for it, find it, realise that it's already attached and either remove the first attachment because it was plain wrong or realise that the profile in fact is actually a duplicate and resolve that duplication.
So, to me, using continued hinting of attached sources (a) goes against the grain of what many of us have been told (accept or reject, don't ignore - provided you understand it) and (b) is a misuse of that function if it's intended to resolve duplicates and bad prior attachments. One bit of sofware (hinting) should have one function.
I shall now, having apparently disagreed with Juli, take cover!0
Tom Huber said: I'm not sure that Adrian and Juli are at odds in their thoughts.
I think that FamilySearch is looking at ways to reduce bad hints that impact certain parts of the world (as noted in many discussions about the hinting mechanism).
No matter how I look at this, a bad hint here and there is always going to be a problem. I don't see any way around it. Whether the bad hint is the result of a death of a child, whose name is used for the next **** child born to that family, or a place problem in which a hint is generated for a person when they lived elsewhere at the time -- these are the same problems that can badly impact an index when it is set up incorrectly, and in reality, may have the same root cause.
First, I really appreciate John Huff joining this discussion. He has alerted us regulars to a possible change with respect to hints and the operation of the source linker.
The possible duplicates option is fraught with some very real dangers, more related to inexperienced users than for anything else.
Right now, the way the source linker is set up, we get a very obvious flag when a source has been attached to another profile that "hints" (bad pun) at the other already attached profile as being a possible duplicate for the source we are working.
Unfortunately, inexperience goes away only with experience in researching possible sources dealing with my relative. For me, with experience came the understanding that I need to do some serious research with both profiles -- both the one for which the source is already attached, and the one that I am currently working. The problem was made worse by not being able to correct the current index, which may have a problem with a gross indexing error (usually - but not always - created by a badly organized indexing project).
"Pit" errors are always going to plague the single-tree, open-edit approach. But the solution also exists with that same approach. It is as much a cognitive issue as anything else. Some users will never be able to understand that there are no quick and easy ways to approach research. I am still fixing badly attached sources that were done by my close relative, whose approach was, "I've got some time, I'll work the Recommended Tasks for five minutes..."
That approach has always been a problem because the time restraint doesn't allow the user to take the time to truly "vet" or determine if the source is for that relative or not. In addition to the time restraint, there is the problem that the family lived in the same area for generations (currently over 300 years for one location and over 200 years for two others), and reused the same Biblical given names over and over. That factors into bad hints, but it also factors into mistakes in otherwise very conscientiously written and subsequently published family histories.
In my opinion, the current source linker "flag" is a usable solution, at least for me. Others may disagree, and that's fine. I think the potential solution now being considered by the developers described by John Huff, has too many dangers, although FamilySearch's current approach is the same used by Ancestry -- a hinting mechanism that mostly works on a limited basis, and a stronger search routine that can bring up all the potentially applicable sources for the person -- while not ideal for those inexperienced users looking for the low-hanging fruit -- does work well for me, especially when I combine Ancestry's search results on a tree I've developed in parallel to the tree in FamilySearch with the hinting system in FamilySearch.
Of course, I take a lot of time trying to make any given profile as complete as possible. With additional records becoming available all the time, going back through my family is required from time to time and that slows down expanded research.0
Paul said: I'm also completely in favour of retaining this. It can be a help in finding true duplicates - which may have some very useful detail attached compared to the version you are looking at / have created. Alternatively, these hints are commonly attached to the wrong individual, so need to be detached and put on the ID of the individual to which they really do refer.
Showing hints that have already been attached to another ID is a great way of both finding errors and genuine duplicates. I don't think I would have found many of these otherwise.0
gasmodels said: I guess I lean in favor. In particular those hints for those IGI records when may lead to incorrect merges that need to be resolved. I have also found some duplicate records that helped me to clean up the tree. But hints have lead to records where a record for my relative has been incorrectly merged into another. I was able to restore the incorrectly merged record and then merge it correctly. This is a significant issue which only affects members trying to resolve missing ordinance issues.0
Paul said: Yes, I have also been very vociferous in my comments concerning the current algorithms that lead to both crazy "possible duplicate" and "record hint" suggestions. So I appreciate there is definitely another side to the argument here.
However, that ("other side") might include the idea that there should be no possible duplicate, or "unattached" record hints suggestions AT ALL - because we know many users will just go ahead and do everything "FamilySearch" suggests. Disappointingly, I quickly found even the new merge procedure has not stopped that.
I don't think the specific factor being discussed here is likely to cause any more harm than either of the loose FamilySearch algorithms, which are currently responsible for both very good and extremely bad suggestions appearing on the person pages.0
John Huff said: Thanks to this group for a thoughtful review and discussion. This is a change planned for the future and we will do further analysis before any implementation.0