- From: Sandro Hawke <sandro@hawke.org>
- Date: Sun, 15 Aug 2021 15:14:23 -0400
- To: public-credibility@w3.org
- Message-ID: <4df04d42-cadd-9af6-1f53-01910c8bc193@hawke.org>
A few points. *1. *Snopes COO Vinny Green is a member of this group who has contributed at several meetings over the years. The NY Times quotes Vinny as saying "As you can imagine, our staff are gutted and appalled by this". *2. *As an observer of new media (and news media), I've always disliked this practice of publishing news stories that are simply rewrites of other news stories. Ironically, Owen, you linked to the Fox News rewrite of the original BuzzFeed report. At least Fox linked to BuzzFeed. I found the BuzzFeed report vastly more informative and didn't notice anything added by Fox except that claim that Snopes is "liberal". As the BuzzFeed article explains: > "That was his big SEO/speed secret," said Binkowski, whom Snopes fired > without explanation > <https://www.poynter.org/fact-checking/2018/snopes-fired-its-managing-editor-%C2%97-and-she-doesnt-know-why/> > in 2018 (she currently manages the fact-checking site Truth or Fiction > <https://www.truthorfiction.com/about/>). “He would instruct us to > copy text from other sites, post them verbatim so that it looked like > we were fast and could scoop up traffic, and then change the story in > real time. I hated it and wouldn't tell any of the staff to do it, but > he did it all the time.” I know we've discussed a signal about whether a news report is original reporting or not, and the dangers of it looking like hundreds of outlets are investigating a matter, when they're really all just copying one article. *3.* Every signal is imperfect. Of course winning a Webby is no guarantee of perfect accuracy*.* But it's better than nothing, right? And winning a Webby (or two Webbys) 15 years ago (as in this case) is probably a weaker signal than winning it last year. So using the template in Reviewed Signals: Any Award <https://credweb.org/reviewed-signals-20200224/#any-award> we have: The website with main page URL [ https://www.snopes.com/ ] was honored as part of an awards process for the year [ 2006 ] for the prize with main page URL [ https://www.webbyawards.com/ ] and perhaps we could add: The website with main page URL [ https://www.snopes.com/ ] committed one or more major lapses in journalistic integrity in the year [ 2015 ]. (repeat for each year) That could be further refined. BuzzFeed argues Mikkelson's use of a pseudonym was perhaps a greater breach than the plagiarism. Those might be split out into separate signals. One could also link to the specific failure or the reporting about the failure. Owen, you or your software could weight this signal higher than the awards one. Of course, you could also give the awards signal zero weight, but I expect that would often leave you to rely on even lower quality signals. For the next version of Reviewed Signals, we could perhaps add some discussion that explains this more, although I think the text currently there is pretty good. The reason this is a good signal is that it's quite hard to game. You can't just make 100 news outlet websites that have won reputable awards. If you throw out this signal, what are you going to use in its place? *4*. Bob, yes, the W3C Annotation protocol is a serious contender for a way to share signal data, but as I understand standards work, it's essential to have the people who will be adopting the standard at the table when the standard is being set, or at least in active dialog with the folks setting the standard. It's rare for a standard to ever success without that. In the previous thread about trust.txt there was disagreement about whether to apply existing off-the-shelf standards or create something tailored to the community of potential adapters. This is a hard trade-off to make, but in the end it needs to be made in a way that gets the standard adopted if the work is to be useful. The same issue arises with general signal work. We could say, "use the W3C annotation protocol" and fill in the details about exactly how. But would anyone do it? Would it actually be fit for purpose? IMHO it's generally best (and is W3C practice) to wait until you have multiple business who all need the standard, *then *convene the meetings to make sure it's good enough for nearly all of them to agree to use it. That's what I thought we had when we started this group, but they didn't stick around. Part of the reason I don't think CredWeb is ready to move forward is because we don't have any vendors clamoring to move forward with a Credibility Signals system. For a full standards Working Group at W3C we'd need about 20 vendors (or other W3C members supporting the work). In a community group, we can do the work without first reaching that bar, but operating with zero strikes me as a plan unlikely to succeed. -- Sandro On 8/14/21 11:34 AM, Bob Wyman wrote: > On Fri, Aug 13, 2021 at 9:48 PM Owen Ambur <Owen.Ambur@verizon.net > <mailto:Owen.Ambur@verizon.net>> wrote: > > While I do recognize the relevance of awards to tribal vanity and > solidarity, this evidence reinforces my bias against considering > them to > be a credible indicator of credibility: > > > While plagiarism is a serious failing, I'm not sure that it is correct > to suggest that plagiarized content is any more or less credible than > original content. I assume that Snopes received its Webby award > because of a general perception that its content, however sourced, was > useful in determining the truthfulness of statements. If Snopes were > to post a plagiarised confirmation of its own plagiarism, that might > provide further evidence of their unacceptable behavior, but it would > also strengthen their position as a site that publishes truthful > evaluations of statements, memes, etc. Even if all of Snopes' content > was plagiarized, their credibility would depend on their skill in > choosing what to plagiarize. > > The important thing about credibility signals is to be aware not only > of what they indicate but what they do not indicate. Publishing > credible content does not imply that content is published either > honestly or legally. Credibility should be understood to be context > specific; limited to specific purposes and for particular periods of > time, etc. > > Nonetheless, users of Snopes might wish to know of Snopes' history of > poor content sourcing practices. (Those issuing awards for ethical > conduct might be particularly interested...) This confirms for me the > belief that we need a mechanism that allows one to associate > third-person, discoverable comments or annotations to a credibility > signal. It should be possible, on finding a signal of Snopes' > credibility, to create a new signal which says, in essence: "While > they may have once won an award for one thing, they are, or have been, > plagiarists." If credibility signals were provided as identifiable > elements, for instance via Verificable Credentials that record awards, > it should be possible to use the W3C Annotation protocol to associate > comments or qualifying statements with the identifiers of the > Verificable Credential. > > Snopes won the Webby. That fact can't be changed, however, it would be > useful if one could later make the statement "The winner of this award > has been found to have plagiarized content." Doing this would allow > others to better understand the meaning of, and the limitations of, > Snopes' Webby award. > > Is there any reason why the W3C Annotation protocol would not be a > reasonable mechanism for publishing signals about signals > (meta-signals)? Is there a better mechanism for publishing > discoverable, third-party statements about credibility signals? > > bob wyman
Received on Sunday, 15 August 2021 19:14:39 UTC