Wikidata

Wikidata:Requests for comment/Findagrave removed as a source for information

< Wikidata:Requests for comment
An editor has requested the community to provide input on "Findagrave removed as a source for information" via the Requests for comment (RFC) process. This is the discussion page regarding the issue.

If you have an opinion regarding this issue, feel free to comment below. Thank you!


THIS RFC IS CLOSED. Please do NOT vote nor add comments.

The following discussion is closed. Please do not modify it. Subsequent comments should be made in a new section. A summary of the conclusions reached follows.

consensus to keep Findagrave as source --Pasleim (talk) 17:31, 29 September 2018 (UTC)[reply]

Should data items here at Wikidata that are sourced to Findagrave be removed? Wikipedia considers Findagrave to be unreliable because it is community sourced. The argument for/against Findagrave has been going on for at least 5 years at Wikipedia. The discussion has begun at User talk:Nikkimaria but should be open for all to discuss here.

Discussion edit

  •  Keep the source and the data We should be using the best available information at the time. When better information comes along, we edit. When we have contradicting reliable information we move the more reliable to the top position in the data field. No information, when some information is available, does not serve anyone. Freebase, the progenitor of Wikidata, imported the majority of its data from Wikipedia, which is also user generated and contains errors, as we all know. --Richard Arthur Norton (1958- ) (talk) 16:46, 18 June 2017 (UTC)[reply]
Then Wikidata need to define what is better information and define some quality scale. I hope we all agree that a source that is researched by professional researcher like d:Property:P3217 from Dictionary of Swedish National Biography has an internal quality process and has much better quality compared with a community-sourced. Wikidata need to have a process to make it obvious for the reader this quality/trust difference....
FindAGrave is community-sourced and has a design problem
  1. FindAGrave never add primary sources for its information ==> major problem you can nearly never check what they present
    1. adding pictures of the grave is encourage but not adding sources like birth certificates, death certificates etc...
  2. on 99% of the FindAGrave profiles there are no external links proving what is stated ==> you can never check what is stated and its a big risc that what is written is a copy/paste from Wikipedia
  3. a test comparing two "community-sourced" genealogy sites WikiTree d:Property:P2949 and FindAGrave d:Property:P535 found a mismatch of more than 200 000 profiles link
    1. another problem I have seen with community driven sites like FindAGrave is that you don't have a mature change process.
      1. profiles are protected
      2. profile managers are not active
      3. missing something like WikiTree/WIkidata talk pages makes the change process less transparent
      4. in FindAGrave you also has no support for version management ==> you can never track changes or see an earlier version
      5. the location model in FindAgrave is that you select from a predefined list box prefilled with places that maybe is ok in the USA but not in a place like Serbia ==> you cant sometimes add the correct place
I suggest keep FindAGrave but make it visible that this source doesnt have the quality process you find in sources produced by professional researchers d:Property:P3217, d:Property:P2180... - Salgo60 (talk) 09:54, 20 June 2017 (UTC)[reply]
Regarding #3: I don't think we can draw any useful conclusions about Find A Grave from the first table on that page, if that's what you're referring to. Two of the things in the list are finding cases where WikiTree is probably linking to the wrong Find A Grave page (which of course has nothing to do with Find A Grave). Two are listing cases where the two sites have different information, but that doesn't tell us which one has the right information (assuming WikiTree has the correct link). The rest are finding cases where Find A Grave has more detailed data than WikiTree. - Nikki (talk) 15:33, 20 June 2017 (UTC)[reply]
The conclusion is that as FindAGrave don't source its information you can't check it. More detailed doesnt help much if its not sourced....
If you look at a WikiTree profile the intention is as with Wikipedia/Wikidata that you add sources connected to the facts, which makes it a better member in this echo system. As said earlier FindAGrave is not interested in using sources so they have very very weak evidences for the facts they state. For famous people I see often that both WIkitree and FindAGrave just copy/paste from Wikipedia...
As said earlier we need a quality scale of sources. The debate over at WikiTree is that a lot of people gets upset because the Error reporting system tells its a difference between WikiTree and FindAGrave as FindAGrave has no sources its some kind of a dead end...... - Salgo60 (talk) 16:16, 20 June 2017 (UTC)[reply]
The New York Times does not list their sources either: "Bill Dana was born William Szathmary in Quincy, Mass., on Oct. 5, 1924, the youngest of six children. His father, Joseph, a real estate developer, was an immigrant from Hungary; his mother worked in a millinery shop." My estimate is that 1 in 20 New York Times front page stories has a correction appended to it. Pre Internet stories may or may not have had a correction added to the printed "corrections" column several days later. --Richard Arthur Norton (1958- ) (talk) 20:11, 20 June 2017 (UTC)[reply]
@Richard Arthur Norton (1958- ): please read Wikipedia:en:WP:V so we are on the same page. If we should speak about if a source is good enough or not and think it doesnt matter if we can verify it I feel we are moving into the Fake news direction..... You have some good presentations by User:Dtaraborelli on this subject as its the future direction of Wikimedia see also WikiCite - Salgo60 (talk) 21:05, 20 June 2017 (UTC)[reply]
I would say the biggest problem with FindAGrave it is not verifiable as they have no sources - Salgo60 (talk) 21:08, 20 June 2017 (UTC)[reply]
from the examples Nikkimaria provided, I think that is actually false - FindAGrave often links to images of gravestones, which can be viewed by anybody to verify the transcribed information on dates, names, etc. So verifiability for FindAGrave can be quite a bit better than for many other sources we trust. ArthurPSmith (talk) 13:52, 27 June 2017 (UTC)[reply]
But the information provided is not a simple transcription of the gravestone - in one of the examples I've listed there is no image of the gravestone provided at all, and in many other cases the writeup provides details not supported by the gravestone. Nikkimaria (talk) 00:02, 28 June 2017 (UTC)[reply]
"in one of the examples I've listed there is no image of the gravestone..." And is that the case for every item where you've removed a FaG citation, or are you also removing them in cases where there is an image? Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 10:39, 28 June 2017 (UTC)[reply]
The gravestones doesn't always give correct dates or year though, cause they messed up at the order for the dates on them... See for instance this memorial stone for this guy (speed skater) Ole Olsen (Q11993678). The year is discussed here in norwegian on his norwegian wikipedia-discussionpage with several external links for possible dates and year. Regards Migrant (talk) 19:17, 28 June 2017 (UTC)[reply]
As Wikidata has
its easy to do a search and compare them see blogpost Benchmark wikitree findagrave wikipedia
I am not impressed of the quality - Salgo60 (talk) 23:49, 29 June 2017 (UTC)[reply]
  • Remove, not a valid source. Poor sourcing leads to poor data quality; data that we can't trust is absolutely worse than no data at all. Reliable sources disagreeing on something is a red herring since this isn't a reliable source. Nikkimaria (talk) 20:51, 18 June 2017 (UTC)[reply]
  • Automatically, importing FindAGrave data seems to be a bad idea. However, automatically removing it is also a bad idea. The point of quoting sources isn't just reliability, it's being open of the provenance of a statement. Removing data based on sources we don't like encourages people to provide no information about where the information they add is coming from.
    A person who doesn't trust FindAGrave can choose to ignore a value when we tell him that's where the information is coming from. Sometimes FindAGrave has the image of the tombstone which is very useful for having a good idea of the birth and death date of a person. ChristianKl (talk) 22:49, 18 June 2017 (UTC)[reply]
  • If it's desired from the Pedia's to be able to import data from Wikidata that's community-sourced, I think we could find a solution whereby a bot adds type of reference (P3865) "community-sourced website" to claims that link FindAGrave and similar websites and the "only sourced"-flag could filter out those references when importing data. ChristianKl (talk) 08:03, 19 June 2017 (UTC)[reply]
  • It is very difficult to form an opinion based on the scarce input which was given in this RfC until now. I do not see a problem with “community sourced” databases in general, so do we have any statistical information about data quality in FindAGrave that implies serious concerns? Which Wikipedias consider this source unreliable, and where do I find relevant discussions? —MisterSynergy (talk) 08:30, 19 June 2017 (UTC)[reply]
    • The initial Wikidata data set was from Freebase via Google and was compiled mostly from Wikipedia, which is community sourced. Some of the information came from IMDB and similar data sets. This is for biographical data, I do not know about geographical data. --Richard Arthur Norton (1958- ) (talk) 15:06, 19 June 2017 (UTC)[reply]
  • I don't think we should have restrictions on which sources are allowed. Wikidata provides data for many Wikimedia projects, not just the English Wikipedia. We can't apply restrictions on sources from one project without forcing those restrictions on other projects (possibly against their wishes). It is better for Wikidata to remain as neutral as possible and for individual projects to filter out sources they don't want when using the data. The most important thing (in my opinion) is that the data has references. If the data has references, people using the data can choose which ones they want to trust. Removing or disallowing Find A Grave references will not prevent people from using it as a source, we just won't know that they have. There is no limit on how many references a statement can have. If someone wants the data to be referenced to somewhere other than Find A Grave, they are welcome to add additional references using a source they think is more reliable. - Nikki (talk) 09:41, 19 June 2017 (UTC)[reply]
  • The site seems to provide pictures of gravestones and dates of birth/death. I think we should include dates of birth/death from gravestones even if they are known to be incorrect. If they are incorrect, they should simply be marked as such.
    --- Jura 10:18, 19 June 2017 (UTC)[reply]
  • I'd like to have some specific examples of problems here; as a general rule if some information was added to wikidata I'd like to have whatever source for that information recorded alongside that piece of information. We have a huge number of statements in wikidata that are referenced to "imported from" a specific wikipedia - which is obviously not a reliable source, but it is important to have that sourcing recorded so we know something about where the information came from. Without further information it seems to me that "Findagrave" would be no different from the wikipedia case. ArthurPSmith (talk) 14:47, 19 June 2017 (UTC)[reply]
    • The Wikipedia case is specifically addressed here; that would be an appropriate way to address use of this site as well. Nikkimaria (talk) 01:21, 20 June 2017 (UTC)[reply]
      • Odd that we have a rule that you cannot list Wikipedia as the source of a fact when the Freebase data set was mostly derived from Wikipedia ... and then we imported disambiguation pages and other odd miscellany from Wikipedia. Does that mean we delete the tens of thousands of source tags for facts we later imported from Wikipedia to add data to new fields? I see them all the time. --Richard Arthur Norton (1958- ) (talk) 02:09, 20 June 2017 (UTC)[reply]
      • Nikkimaria the Help page suggests to *replace* wikipedia links with more verifiable source links, not to delete them without adding a source. We've had lengthy discussions on this in other RFC's and elsewhere. In any case, you didn't respond to my request for a specific example of the sort of problem you see with this data, can you provide something that actually shows what problems might be caused here? ArthurPSmith (talk) 14:39, 23 June 2017 (UTC)[reply]
        • Yep, I'm suggesting that the approach of not treating the links as real sources and replacing is a viable option - though I'd be interested in seeing the RfCs you refer to. As pointed out above, this is a user-generated site that usually does not include reliable sources for the information presented. Nikkimaria (talk) 00:06, 24 June 2017 (UTC)[reply]
          • @Nikkimaria: well the recent inconclusive Verifiability and living persons RFC also went into the reliable source question - also inconclusively. A couple of years ago we had this RFC on improving bot imports that specifically discussed sourcing to wikipedia. And there's been much discussion on Project Chat through the years too. There is definitely no consensus to remove "imported from xxwiki" statements - or references to any other "unreliable" sources at this point, with the wikidata community. And you still provide no specific example of a problem - what I'm expecting is for you to provide preferably several cases along the lines of: "wikidata item Qxxxx has this statement Pyyyy vvvv sourced to FindAGrave, but it is incorrect according to this more reliable source zzzz". If there's not a track record of incorrect information from this source then I don't see why we are even discussing it here. ArthurPSmith (talk) 15:00, 26 June 2017 (UTC)[reply]
            • A couple of quick examples: Conchita Montenegro (Q2991961) cites it for a death date that contradicts multiple reliable sources (eg [1] or [2]), and Henry Mandeville (Q24169858) cites it for a birth date that contradicts multiple reliable sources (eg. [3] and [4]). I see no reason why we should not replace it with more reliable sourcing. Nikkimaria (talk) 02:50, 27 June 2017 (UTC)[reply]
              • thank you for the examples. However, I note that neither BFI nor the newspaper you point to as sources for Conchita Montenegro (Q2991961) death date (April 26, 2007) provide references, and the FindAGrave date (April 22) is supported also by IMDB. None of these meets a standard of "verifiability" beyond linking to the referenced source and seeing what it says. So, in a case like this, we should add the alternate stated date and its sources and leave it there as a disputed piece of information - if you suppress the April 22 entry you are hiding the fact that a disagreement exists. In the case of Henry Mandeville (Q24169858) I note that FindAGrave has a more precise death date than either of your references (which state only the year) so I'm wondering why you trust them more on the birth date? In the FindAGrave case, both birth and death dates can be further verified by examining the dates engraved on the gravestone, in the image provided there. So the FindAGrave values are MORE verifiable than the values from your other sources. Maybe whoever arranged the gravestone got it wrong by 2 days for some reason; again both dates should be added to indicate there is some disagreement on the topic. I don't find in either case here that FindAGrave is providing anything but a useful reference. ArthurPSmith (talk) 13:48, 27 June 2017 (UTC)[reply]
                  • The point of considering a source "reliable" is that it is trustworthy in itself; something like BFI meets that, something like IMDB does not. Cite the gravestone directly for what the gravestone says if you must; don't cite a user-generated writeup that says what the gravestone doesn't. Nikkimaria (talk) 00:02, 28 June 2017 (UTC)[reply]
I am not against scoring data sources for their accuracy if it can be done objectively. Comparing two data sets for, say, birth year can only show that they disagree, not which one is correct. Also "community-sourced" is not a synonym for "inaccurate", as a data set I would say that IMDB and Findagrave are more accurate than any other data set containing similar data, because there is no other data set with that data. --Richard Arthur Norton (1958- ) (talk) 04:32, 21 June 2017 (UTC)[reply]
I think step one is to define some common terms of quality. Saying something is more accurate because it is a bigger dataset is odd.
My advice measure
WikiTree with 14 000 000 million community sourced profiles thought it had about 2000 profiles marked unsourced or maybe 10 000. After finding a bug in the software WikiTree now have +1 000 000 profiles marked unsourced. The same is with the above mentioned error reporting tool. It has found millions of errors.
Yes you can say something is incorrect just by checking the dates... lesson learned is with todays copy/paste and lack of sources and requirements of Verifibility people add things that makes no sense like mother born after child.
Quality is not something you just get you need a process.....
  • We need to measure quality
  • Compare different sources
  • make it visible that this data is possible to verify with a source that has a proven record of quality ...
- Salgo60 (talk) 14:29, 21 June 2017 (UTC)[reply]
  • I have also politely asked User:Nikkimaria to migrate the Findagrave links she is fervently deleting at Wikipedia to Wikidata, before the information is lost. So far, she has not migrated any data, but continues to delete the information at Wikipedia. I think it had the opposite effect and caused her to start to delete the information here at Wikidata also. Can someone convince her of the importance of not losing the data? Is there anyway to see how much information she lost? --Richard Arthur Norton (1958- ) (talk) 19:53, 23 June 2017 (UTC)[reply]
    Whether data get's deleted from Wikipedia is a Wikipedia problem. The data is also not fully lost. It would be possible to go through the history of an item and find all the deleted links. ChristianKl (talk) 20:36, 23 June 2017 (UTC)[reply]
  •  Keep A source is a source, and each source will always have a credibility measure. It would be true that Findagrave has community sourcing and that should be noted, it does not have its summary removal, it should simply allows its rating against other sources, or pointing to other potential source. If it is the only available source, it is better than nothing. For the vast bulk people of the world they will not be in published sources, especially the more notable sources, so these sources of less quality must be retained.  — billinghurst sDrewth 22:08, 23 June 2017 (UTC)[reply]
  • That's not exactly true. Of the three points of Wikidata:Notability, two are related to sourcing: #2 requires that an item have serious published sources about it, while #1 requires that an entry on another project, which generally speaking requires that sources are available to meet their notability policies. Nikkimaria (talk) 13:51, 24 June 2017 (UTC)[reply]
  • Seriousness is another standard than reliability. Wikidata also references information which we know to be wrong (and therefore not reliable) and marks those claims as deprecated.
As far as I understand our current consensus we don't see user-generated content automatically as not-serious. Wikipedia is a serious website. On the other hand, we also don't automatically import user-generated data from places other than Wikipedia but only add it when a Wikidatian explicitly considers the data to be valuable in a specific use-case.
Apart from that, the standard asks whether the item can be described with serious and public sources and not that the claims on its statements are supported by those sources. ChristianKl (talk) 14:14, 24 June 2017 (UTC)[reply]
What is the standard of "seriousness", if not reliability? Nikkimaria (talk) 01:22, 25 June 2017 (UTC)[reply]
It's an intention to transfer reliable information. I think FindAGrave is a website that has that intention. ChristianKl (talk) 08:45, 25 June 2017 (UTC)[reply]
Not the implementation. Is that standard recorded somewhere? Nikkimaria (talk) 02:50, 27 June 2017 (UTC)[reply]
I don't think there's a single point where the standard get's defined. It's rather, that there are frequently discussions in RfC's and the project chat where we agree about whether items are notable according to our criteria or aren't. ChristianKl (talk) 09:12, 27 June 2017 (UTC)[reply]
  •  Comment Looks like we can close this. As point of procedure, do reverts and deletions count as "non automated main namespace edits"?
    --- Jura 10:30, 27 June 2017 (UTC)[reply]
    • I think normal reverts and deletions are "non automated main namespace edits". If you on the other hand delete claims in bulk via a bot or QuickStatements that's an automated edit. ChristianKl (talk) 10:56, 27 June 2017 (UTC)[reply]
  •  Keep FindAGrave is clearly a useful source for wikidata since it (often) provides images of gravestones that can be directly examined to verify the information on birth and death dates, etc. ArthurPSmith (talk) 13:53, 27 June 2017 (UTC)[reply]
  • So cite the gravestone, as a primary source. And in cases where the gravestone isn't included, or doesn't provide the information, the Findagrave writeup is not a reliable substitute or supplement. Nikkimaria (talk) 00:02, 28 June 2017 (UTC)[reply]
You used the same argument before at the IMDB discussion, so I will use the my same counter argument. You wrote: "That's not how this works" This isn't about what I think, or what you think. It is based on community consensus, if you want to change that consensus, you will have to persuade the community with statistics, not with emotions and anecdotes. --Richard Arthur Norton (1958- ) (talk) 03:19, 5 July 2017 (UTC)[reply]
  •  not different from other sources Statements should be sourced and when statements are doubtful it make sense to deprecate or remove statements or to keep remarks about "unreliable sources". It is unhelpful to remove just references. d1g (talk) 17:06, 7 September 2017 (UTC)[reply]
Not source, but each statement: Help:Deprecation d1g (talk) 17:53, 7 September 2017 (UTC)[reply]
Which doesn't really solve the problem, unfortunately. Nikkimaria (talk) 17:57, 7 September 2017 (UTC)[reply]
@Nikkimaria: what is a problem and how? d1g (talk) 18:01, 7 September 2017 (UTC)[reply]
You stated above that we should "keep remarks about 'unreliable sources'"; we cannot do that if we cannot mark a source as unreliable. Nikkimaria (talk) 18:03, 7 September 2017 (UTC)[reply]
@Nikkimaria: "source as unreliable" - nonsense.
Statements can be incorrect, but not sources. d1g (talk) 18:08, 7 September 2017 (UTC)[reply]
Nonsense. Nikkimaria (talk) 18:10, 7 September 2017 (UTC)[reply]
  • My conclusion based on Wikidata policy and the consensus above. For the purposes of Wikidata, the website Findagrave is considered a "serious" source of data, and we recognize that it may contain typos and other errors, as do all data sets. At Wikidata when two sources differ, for example for the date of birth, the correct date is raised in rank, and a widely reported incorrect date is deprecated in rank. If it is inconclusive as to which is correct, the two dates are given the same rank. All Wikidata statements should be sourced. Findagrave typos and other obvious errors should be reported to Findagrave using their edit feature. For people born/died in the United States there are several databases containing birth and death dates that are based on official government documents such as birth certificate, marriage certificates, death certificates, and self reported documentation such as draft registrations, and passport applications. Even these dates may differ from each other. There are dozens of examples of people self reporting their birth dates in the WWI and WWII draft registration and giving different years for each. The same for people filling in passport applications. Generally the rule of thumb is that the document closest to the event in time, is more likely to be correct. For example the birth date given on a birth certificate is more reliable than the birth date given 20 years later in a draft registration or a passport application. --Richard Arthur Norton (1958- ) (talk) 05:04, 30 October 2017 (UTC)[reply]
  •  Keep Findagrave is community sourced is not a good reason for not trusting it. Wikidata is also community sourced. I will usually trust a gravestone more than other references for dates of birth and death and if the image can only be found on Findagrave than it is a great resource. --Jarekt (talk) 16:23, 22 November 2017 (UTC)[reply]

Do I correctly read this as a strong, but not unanimous consensus to keep? And, unless someone thinks I'm reading that wrong, can we close on that basis? - Jmabel (talk) 23:54, 28 July 2018 (UTC)[reply]

The above discussion is preserved as an archive. Please do not modify it. Subsequent comments should be made in a new section.
Retrieved from "https://www.wikidata.org/w/index.php?title=Wikidata:Requests_for_comment/Findagrave_removed_as_a_source_for_information&oldid=754784426"