Topic on Talk:Reading/Web/Projects/Related pages/Flow

RelatedArticles and displaying non-free content

27
Summary last edited by Jkatz (WMF) 18:42, 14 March 2016 8 years ago

The fix for this has gone live!

https://phabricator.wikimedia.org/T124225

Finnusertop (talkcontribs)

According to Wikimedia Foundation's resolution on licensing policy (https://wikimediafoundation.org/wiki/Resolution:Licensing_policy), all content on Wikimedia projects should be free, with the exception of those non-free files used under a fair-use doctrine (or similar).  According to the resolution, such uses are conditioned by the projects' Exemption Doctrine Policies (EDP). The resolution goes on to say that exceptions granted by the EDPs should be minimal. The English Wikipeda's EDP is "Wikipedia:Non-free content" (https://en.wikipedia.org/wiki/Wikipedia:Non-free_content). It stipulates that a non-free file may only be used in those articles that a specific, relevant and valid non-free use rationale is written for.

The RelatedArticles beta feature dynamically links the article the user is viewing to other related articles. Along with the titles of these related articles, an image from each target article is displayed to the user. Many times, the displayed image is non-free (eg. the image displayed for an article about a book is the cover used in the infobox of the target article). Since the user is displayed this information on the page they are viewing now, and not the target page, the use is not covered by the image's non-free use rationale. Because the links between related articles are created dynamically, there can be no purposeful evaluation of whether the use constitutes a valid fair-use claim.

Does the Reading team think that the RelatedArticles feature conforms with the foundation's resolution on licensing policy? Does the team think the feature is in conformity with English Wikipedia's EDP on non-free content? ~~~~

Slaporte (WMF) (talkcontribs)

Hi Finnusertop, thank you for your question.

I appreciate your thoughts on this topic, and I wanted to share some research. We believe the use of thumbnail images to help suggest related articles to users complies with our internal policies and is fair use under U.S. copyright law.

As you may know (and can read in detail about on wikilegal), the fair use analysis involves an evaluation of: (1) the purpose and character of the use; (2) the nature of the copyrighted work; (3) the substantiality of the portion of copyrighted work used; and (4) the effect on the market value of the copyrighted work. The purpose of this evaluation is to determine whether the creative work is transformative—i.e., adding something new—or whether it merely supplants the original work.  

In general, thumbnail images accompanying online search engine results are considered to be transformative under U.S. copyright law. This is the case even when the images are an exact copy of the original image. The key is determining whether or not the purpose of the image has been changed. In a similar case, Perfect 10 v. Amazon, the United States Court of Appeals for the Ninth Circuit evaluated Google’s use of thumbnail versions of images owned by Perfect 10, Inc. as part of Google’s search engine results. Ruling in favor of the search engine, the court concluded that the use of the Perfect 10’s images alongside a hyperlink was used to aid in research and access, and not in a manner that directly competed with Perfect 10’s business. Put another way, Google’s use of Perfect 10’s images created something new. For this reason, the court found that Google’s use of Perfect 10’s images was highly transformative and valid under the doctrine of fair use.

Akin to Google’s use in Perfect 10, Inc. v. Amazon.com, Inc., the purpose of the RelatedArticle beta feature is to disseminate and provide access to information. Moreover, the tool uses reduced resolution copies of the images and could not be considered market replacements for their original. Taken together, the use of the thumbnail images as part of the RelatedArticle tool would be considered fair use under U.S. copyright law.

The use of the thumbnail images are also valid under Wikimedia Foundation's resolution on licensing policy the project's Exemption Doctrine Policies (EDPs). Specifically, these policies are structured to ensure that users avoid publishing infringing content online. Here, all the images used are hosted under a free license, in the public domain, or a project's EDP rationale, and the images are always displayed in connection with the article where they appear. As a fair use, these images also fall within the rationale of their related article. It is also worth noting that the RelatedArticles beta feature is a software change in how media is displayed, not how it is hosted. Since it does not upload new media or alter how images are hosted on the projects, the feature itself falls outside the scope of the licensing policy.

Again, thanks for your question, and happy new year!

Seraphimblade (talkcontribs)

The question wasn't about legality or WMF . It was about overriding project policy. Could you please respond directly to that concern?

Slaporte (WMF) (talkcontribs)

The RelatedArticle feature meets our licensing policy by only displaying images next to an article only if the image is in included in the article under a valid Exemption Doctrine Policy.

If a project removes an image from an article because it doesn't meet the EDP, it will also not appear in the RelatedArticle feature.

Nunabas (talkcontribs)

And yet another failure to understand enwiki's NFCC policy. Usage of a file in one context may be allowable but not under others. enwiki specifically bans this type of usage, see point 6 of https://en.wikipedia.org/wiki/Wikipedia:Non-free_content#UUI Keep in mind that enwiki uses the term non-free media because their policy on usage of non-free media is stricter than the law, in line with what the WMF dictated and the spirit of the project. How are you going to implement this extension in a manor that doesn't violate existing policy?

Slaporte (WMF) (talkcontribs)

I'll take a look at point 6 as well, thanks. I'm looking to understand how to apply the local guidelines on fair use in a manner that is based on the principles behind the guidelines, and not just an overly strict interpretation.

Nunabas (talkcontribs)

First thing you want to do then is stop treating non-free media like fair use. If you try to use WP:NOT as a reason for ignoring policy you will have serious negative backlash. enwiki's policy on non-free media is designed to be stricter than needed, keep in mind its policy not a guideline. Given how clearly the usage of non-free media is dictated on enwiki there is zero chance that your tool will be allowed to use non-free media. Yes it may be easier and prettier to do so, it is not acceptable under policy, regardless of how "overly strict" you think the local EDP is.

Nunabas (talkcontribs)

Wow.. The response that Slaporte (WMF) gave is shockingly horrifying. This has nothing to do with the legal allowances. Hell if the communities wanted it we could have a lot more usage than current policy dictates. I know for a fact that de.wiki has zero non-free media not because of any legal issues but for principle issues. en wiki takes a more balanced view which allows some usage under limited criteria. This tool by no means meets those. Thus usage of non-free media should be prohibited unless the local community allows it for this tool.

Slaporte (WMF) (talkcontribs)

Hi Nunabas, sorry, I didn't mean to horrify you with legal analysis!

The RelatedArticle feature should not display images if they are not included under a valid EDP, respecting de.wiki's choice to not display fair use media. If you see any inconsistencies on this point, let me know. Thanks!

Nunabas (talkcontribs)

Actually its not the legal speak that is shocking, but rather your failure to grasp the local policy and usage of non-free media. Just because its legal doesn't mean that its allowed. As is referenced in other places each usage of non-free media on enwiki is treated and reviewed separately. Just because a file is uploaded and used correctly under the EDP for a given article (enwiki's policy is on a per page/usage acceptability) doesnt mean that the media can be used elsewhere under the EDP automatically.

Masem (talkcontribs)

Speaking as one that is heavily involved with NFC at en.wiki, I have to agree this is a troubling position, given how much work we have made to uphold the Foundation's Resolution on non-free media at en.wiki. The legality and fair use aspects, I fully agree with, we're not talking about the legal stance here, but in this case it is the effort to minimize non-free for en.wiki, and I can see Nunabas' point on a wiki like di.wiki where no non-free is allowed. (And what happens with Commons? )

I would suggest that this feature rely on pulling either the first free image as tagged in the article, or have a magic keyword that an alternate, free image can be used if there is no non-free. For example, most images for video games are non-free, so I would think that if a video game related article came up we could use one of the Wikiproject Video Game icons as a default replacement, and making sure we have the ability to put that in place. Thus, a per wiki policy can be established for how to handle Related Content.

Slaporte (WMF) (talkcontribs)

I appreciate your work on NFC on en.wiki. As I mentioned above, the RelatedArticle feature only pulls images that validly appear in an article, so it should respect the local policy's choices on including fair use media.

Can you explain more about why it would be better to show a free image for an article, when a non-free image is allowed in the article under the local policy and U.S. law? I can talk with the engineers about your suggestions, and it's helpful if I understand the position that you are coming form.

Dragons flight (talkcontribs)

Local en.wiki policy allows the use of non-free images in _articles_ when they convey irreplaceable information; however, local policy also prohibits non-free images from being used for purely aesthetic or navigational purposes. For example, non-free images generally aren't allowed in navigational portals, even if the same image would be allowed in the article itself. In other words, it is not just a question of whether a non-free image is allowed in Wikipedia, but also a question of what pages it is allowed to appear on and in what contexts it is allowed to be used.

Present policy would not allow a Wikipedia editor to use non-free images for a navigational purpose, like a see also list. This is not a legal issue. It is purely a matter of the choices the local community has made in an effort to promote free content whenever possible (and by extension reduce non-free content to a minimum).

Slaporte (WMF) (talkcontribs)

Dragons flight, that's clear and helpful, thank you. I've found the section you are referring to in English WP:NFC. Do you know when/where that standard was discussed so I can read about it further? I'm trying to understand the reason for setting such a stringent local policy beyond what's required in the Licensing policy.

Ruud Koot (talkcontribs)

This standard was not discussed in a single place at a single point in time: it was developed by the community during the past 15 years. The substantial WP:NFC and the many pages it links to should provide sufficient background information. I'm honestly shocked that a WMF employee can be so ignorant about this issue.

Free content is one of the core pillars of Wikipedia. As such, non-free content is only used in the body of encyclopedic articles, and then only if irreplaceable. We do not even display non-free content on the main page if related to the featured article of the day. Slapping non-free images on every article as clickbait is not going to be acceptable.

Slaporte (WMF) (talkcontribs)

Hi Ruud, I appreciate that you are taking the time to walk me through it. I'm familiar with WP:NFC, but I'm more familiar with the underlying law than the Wikipedia-specific policy -- I did not realize that fair use media was not used on the main page, for example! I'm still interested in understanding the underlying policy considerations for such a strict rule around navigation in particular, if you have any pointers.

Ruud Koot (talkcontribs)

I do not have any links at the ready, but you should be able to find discussion in the archives on at least the following issues (try searching the archives of the village pump or various noticeboards):

  1. Whether we should use non-free images related to today's featured article on the main page? (Outcome: no, on ideological grounds, we want to minimize use of non-free contents, especially on important pages like the main page.)
  2. Whether we should use non-free logos for decorative or navigational purposes in for example navigation boxes? (Outcome: no, because they serve no essential purpose and we want to minimize use of non-free content.).
  3. Whether we should use images of DVD covers or screen caps of television programs in lists of television episodes? (Outcome: no, because in lists these images are only useful for navigational purposes and not for criticism as they would be in articles on individual episodes.)

Especially the latter would seem relevant here.

Slaporte (WMF) (talkcontribs)

Much appreciated.

On the third point, I'm curious if the decision was made on purely ideological grounds (eg, content should be free), in the interest of the mission (eg, ensuring content is maximally accessible by being maximally free), some perceived legal risk, or some combination thereof. I'll see what I can find.

Anomie (talkcontribs)

(FYI, I'm writing here in my volunteer capacity)

I believe ideological and mission grounds are the main reasons: if the non-free content isn't specifically increasing understanding of the topic, we don't need it and so we shouldn't include it to better serve the mission and ideology. There's widespread acknowledgment that the policies around NFC on enwiki are much stricter than what we could get away with under fair use for the purpose of Wikipedia. The only potential legal issue that I can think of informing the debate is that setting our requirements so far within what fair use allows means we usually don't have to concern ourselves with the extremely vague boundaries of what fair use actually covers (we get enough of that trying to figure out threshold of originality in borderline cases, and probably err on the side of caution there too).

Besides enwiki-specific policy, wmf:Resolution:Licensing policy#Resolution in item #3 also requires stricter limits than what fair use would allow. IMO, that too could easily be read as prohibiting use for "decorative" purposes such as navigation.

One further note: wmf:Resolution:Licensing policy requires that non-free content be "identified in a machine-readable format", which theoretically makes it easy for the engineers behind this feature to identify which images are problematic, but there's no specification as to what form that identification must take which makes it more difficult. enwiki identifies them by placing them in en:Category:All non-free media, but other wikis might take other approaches.

Jdlrobson (talkcontribs)

The main issue this stems from is that Commons has no structured data which is a big investment technically.

There are some short term solutions that would be more easily achieved:

  • Allow edits to override the default generated page image (see https://www.mediawiki.org/wiki/Topic:Sws91tugwkprn0dj )
  • If there is some way at identifying pages which contain non-free images e.g. a category on that page we could disable the PageImage auto generation.
  • There is a workaround that allows licenses via screen scraping in Extension:CommonsMetadata that might be possible to be used here to check images selected and ensure they are free but that needs some investigation.
Ruud Koot (talkcontribs)

Commons having no structured data is a red herring: images uploaded to commons.wikimedia.org are free, nearly all images uploaded to en.wikipedia.org are non-free (or, as noted above, check if they're in Category:All non-free media if you want to be really sure you're not missing any free files).

Kusma (talkcontribs)

Gather has similar problems with non-free content, see Talk:Gather#Gather_violates_English_Wikipedia_policies_in_an_undetectable_manner. Automatically displaying images without checking their free-/non-free status should just not be done.

Jdlrobson (talkcontribs)
Slaporte (WMF) (talkcontribs)

Thanks, Jon!

Gestumblindi (talkcontribs)

@Slaporte (WMF) @Jdlrobson I think the issue for German-language Wikipedia is slightly different: Indeed, there should be no issue with non-free media, as German Wikipedia shouldn't contain any non-free images anyway (local policy based on the absence of "fair use" in the law of German-speaking countries). But the freely licensed images still require attribution in most cases. And there is no direct link from the "related pages" thumbnails to the image description / license. You have first to navigate to the article and then click on the image to get attribution / license information. The thumbnails are still images protected by copyright, though freely licensed, and some will certainly say that this indirect attribution isn't sufficient - especially in countries where there is no "fair use", so you can't say that using the thumbnails without direct attribution amounts to "fair use".

Kusma (talkcontribs)

In general, I think we never should rely on a fair use defense when we use Free Content. While it may be legal in some countries to do so, Wikimedia as the biggest player in the Free Content movement should take special care to honor the license restrictions that most free files have. If Wikimedia doesn't go the extra mile to clearly get image attributions right, how can we expect anyone else to do so? This seems like a central part of our Free Content mission to me, but perhaps I just do not understand what the mission is.

Most Wikipedias (all that I am reading) are careful to always link non-PD files to their image description page for attribution. I can't see a compelling reason why auto-generated content should be different here.

Jkatz (WMF) (talkcontribs)

The fix for this has gone live!

https://phabricator.wikimedia.org/T124225

I think in the future this should be that the API let's the end-digester know what the status is, so that local wiki's can determine whether or not they want the image, but this is a good fix for now.