Impossibility Of Content Moderation: Scientist Debunking Vaccine Myths Gets A YouTube Strike For Medical Misinfo

from the it's-tougher-than-you-think dept

Here's another one in our never ending series on the impossibility of doing content moderation at scale, and how (all too frequently) people trying to expose bad behavior are punished as if they're promoting bad behavior.

This involves a scientist who streams on YouTube as Scientist Mel, and tries to educate people about science, including debunking bad science takes. This included a recent two hour episode debunking anti-vaxxers and anti-maskers. The video does look at a bunch of ridiculous conspiracy theories and scientific claptrap and nonsense... and then debunks it. But, YouTube dinged her channel for misinformation:

And... look, you can see how this happens, right? You have reviewers with limited time and a two hour video is a lot to take in. And in skimming through the video, you will come across conspiracy theory nonsense regarding COVID and COVID treatments. But it's there so that Mel can debunk it. So... Mel appealed the strikes. And YouTube upheld the appeal, claiming "we reviewed your content carefully, and have confirmed that it violates our medical misinformation policy." They were condescendingly nice about it in their form email: "We know this is probably disappointing news, but it's our job to make sure that YouTube is a safe place for all." Yeah, that's not what this is doing, of course.

Of course, as this started to get a bit more attention... on Twitter... suddenly YouTube realized it might have made a mistake, and over the weekend admitted it was a mistake and the strike was removed (and the video returned).

So, in this case, the end result worked out -- but like so many cases, that only happened once the initial "bad" result started to get some attention elsewhere (Twitter being the world's best "consumer complaint" forum).

Of course, I'm sympathetic to both ends of this argument. I can see why this clearly erroneous strike/takedown sucks for Mel. But I can also see exactly how it happens. When tons of people are screaming at every social media site out there to make sure they take down any medical misinformation, and videos often involve (as in this case) multiple hours of content, it's literally impossible to review them completely and understand the context in which the conspiracy theory nonsense was shown. That doesn't excuse YouTube messing up the appeal, of course, but the appeals process apparently isn't particularly robust, and likely suffers from the same limitations as the initial review.

And this kind of issue is only going to get worse and worse as various laws around the world demand that "bad" information be removed quickly or websites will face significant liability.

Filed Under: content moderation, debunking, myths, science, scientist mel
Companies: youtube

Seven Years Ago, CERN Gave Open Access A Huge Boost; Now It's Doing The Same For Open Data

from the tim-berners-lee-would-be-proud dept

Techdirt readers will be very familiar with CERN, the European Council for Nuclear Research (the acronym comes from the French version: Conseil Européen pour la Recherche Nucléaire). It's best known for two things: being the birthplace of the World Wide Web, and home to the Large Hadron Collider (LHC), the world's largest and most powerful particle accelerator. Over 12,000 scientists of 110 nationalities, from institutes in more than 70 countries, work at CERN. Between them, they produce a huge quantity of scientific papers. That made CERN's decision in 2013 to release nearly all of its published articles as open access one of the most important milestones in the field of academic publishing. Since 2014, CERN has published 40,000 open access articles. But as Techdirt has noted, open access is just the start. As well as the final reports on academic work, what is also needed is the underlying data. Making that data freely available allows others to check the analysis, and to use it for further investigation -- for example, by combining it with data from elsewhere. The push for open data has been underway for a while, and has just received a big boost from CERN:

The four main LHC collaborations (ALICE, ATLAS, CMS and LHCb) have unanimously endorsed a new open data policy for scientific experiments at the Large Hadron Collider (LHC), which was presented to the CERN Council today. The policy commits to publicly releasing so-called level 3 scientific data, the type required to make scientific studies, collected by the LHC experiments. Data will start to be released approximately five years after collection, and the aim is for the full dataset to be publicly available by the close of the experiment concerned. The policy addresses the growing movement of open science, which aims to make scientific research more reproducible, accessible, and collaborative.

The level 3 data released can contribute to scientific research in particle physics, as well as research in the field of scientific computing, for example to improve reconstruction or analysis methods based on machine learning techniques, an approach that requires rich data sets for training and validation.

CERN's open data portal already contains 2 petabytes of data -- a figure that is likely to rise rapidly, since LHR experiments typically generate massive quantities of data. However, the raw data will not in general be released. The open data policy document (pdf) explains why:

This is due to the complexity of the data, metadata and software, the required knowledge of the detector itself and the methods of reconstruction, the extensive computing resources necessary and the access issues for the enormous volume of data stored in archival media. It should be noted that, for these reasons, general direct access to the raw data is not even available to individuals within the collaboration, and that instead the production of reconstructed data (i.e. Level-3 data) is performed centrally. Access to representative subsets of raw data -- useful for example for studies in the machine learning domain and beyond -- can be released together with Level-3 formats, at the discretion of each experiment.

There will also be Level 2 data, "provided in simplified, portable and self-contained formats suitable for educational and public understanding purposes". CERN says that it may create "lightweight" environments to allow such data to be explored more easily. Virtual computing environments for the Level 3 data will be made available to aid the re-use of this primary research material. Although the data is being released using a Creative Commons CC0 waiver, acknowledgements of the data's origin are required, and any new publications that result must be clearly distinguishable from those written by the original CERN teams.

As with the move to open access in 2013, the new open data policy is unlikely to have much of a direct impact for people outside the high energy physics community. But it does represent an extremely strong and important signal that CERN believes open data must and will become the norm.

Follow me @glynmoody on Twitter, Diaspora, or Mastodon.

Filed Under: experiments, knowledge, open access, open data, science, sharing
Companies: cern

Elsevier Says It's Infringing To Link To Sci-Hub; Hypocrite Elsevier Links To Sci-Hub All The Time

from the because-of-course dept

Academic publishing giant Elsevier really, really, really hates Sci-Hub, the site that offers up access to lots of academic research. Elsevier has sued the site directly and tried many times to get it blocked (which, to date, seems to have only helped it get more attention). Last week, Elsevier got all legal-threaty against Citationsy, a site that helps scholars create citations. Elsevier claimed that Citationsy was infringing its copyright by linking to Sci-Hub.

But, here's the issue: as Martin Paul Eve pointed out, Elsevier, itself, points to Sci-Hub pretty damn often.

That said, the more entertaining fact is that scholarly publishers frequently end up linking to Sci-Hub. Here’s one I found on Elsevier’s own ScienceDirect site:

One might also consider this article on ScienceDirect, which in earlier versions contained a now-removed Sci Hub link:

The list goes on. Consider this Elsevier article:

Heck, I found too many examples to list them all, but this accepted manuscript (published online by Elsevier on their site YESTERDAY) contains a Sci-Hub link:

As Cory Doctorow at BoingBoing points out, it shouldn't be at all surprising that tons of academic research links to Sci-Hub:

It's also not hard to understand this. You see, the researchers who write the papers that Elsevier publishes are scientists, not private-equity-backed looter/profiteers, so they are more interested in science and scholarship than ensuring that Elsevier continues to rake in billions. And since Elsevier doesn't pay for any of the work it publishes, it's hard for them to exert pressure to end this practice.

Now, in theory, the referees, peer reviewers, editorial boards and advisors for each journal could lay down the law on this stuff and ban links -- but they're also all volunteers who are not paid a dime by Elsevier.

The reality is that scholarly publishing corporations contribute virtually nothing to scholarship (that's a peer-reviewed, empirical finding -- not a statement of opinion). Indeed, they are so lax and inattentive, with so much of their efforts focused on rent-seeking, that they don't even notice when they themselves violate their own signature policies.

Of course, expecting Elsevier to recognize how hypocritical it's being is a fool's errand as well. Elsevier is not in the business of actually understanding how academics work. It's in the business of squeezing as much cash as possible from universities.

Filed Under: copyright, research, science
Companies: elsevier, sci-hub

Why Carl Malamud's Latest Brilliant Project, To Mine The World's Research Papers, Is Based In India

from the sci-hub-to-the-rescue-again dept

Carl Malamud is one of Techdirt's heroes. We've been writing about his campaign to liberate US government documents and information for over ten years now. The journal Nature has a report on a new project of his, which is in quite a different field: academic knowledge. The idea will be familiar to readers of this site: to carry out text and data mining (TDM) on millions of academic articles, in order to discover new knowledge. It's a proven technique with huge potential to produce important discoveries. That raises the obvious question: if large-scale TDM of academic papers is so powerful, why hasn't it been done before? The answer, as is so often the case, is that copyright gets in the way. Academic publishers use it to control and impede how researchers can help humanity:

[Malamud's] unprecedented project is generating much excitement because it could, for the first time, open up vast swathes of the paywalled literature for easy computerized analysis. Dozens of research groups already mine papers to build databases of genes and chemicals, map associations between proteins and diseases, and generate useful scientific hypotheses. But publishers control -- and often limit -- the speed and scope of such projects, which typically confine themselves to abstracts, not full text.

Malamud's project gets around the limitations imposed by copyright and publishers thanks to two unique features. First, Malamud "had come into possession (he won't say how) of eight hard drives containing millions of journal articles from Sci-Hub". Drawing on Sci-Hub's huge holdings means his project doesn't need to go begging to publishers in order to obtain full texts to be mined. Secondly, Malamud is basing his project in India:

Over the past year, Malamud has -- without asking publishers -- teamed up with Indian researchers to build a gigantic store of text and images extracted from 73 million journal articles dating from 1847 up to the present day. The cache, which is still being created, will be kept on a 576-terabyte storage facility at Jawaharlal Nehru University (JNU) in New Delhi.

India was chosen because of an important court battle that concluded two years ago. As Techdirt reported then, it is legal in India to make photocopies of copyright material in an educational context. Malamud's contention is that this allows him to mine academic material in India without the permission of publishers. But he also believes that his TDM project would be legal in the US:

The data mining, he says, is non-consumptive: a technical term meaning that researchers don't read or display large portions of the works they are analysing. "You cannot punch in a DOI [article identifier] and pull out the article," he says. Malamud argues that it is legally permissible to do such mining on copyrighted content in countries such as the United States. In 2015, for instance, a US court cleared Google Books of copyright infringement charges after it did something similar to the JNU depot: scanning thousands of copyrighted books without buying the rights to do so, and displaying snippets from these books as part of its search service, but not allowing them to be downloaded or read in their entirety by a human.

The fact that TDM is "non-consumptive" means that the unhelpful attitude of academic publishers is even more unjustified than usual. They lose nothing from the analytical process, which is merely extracting knowledge. But from a sense of entitlement publishers still demand to be paid for unrestricted computer access to texts that have already been licensed by academic institutions anyway. That selfish and obstructive attitude to TDM may be about to backfire spectacularly. The Nature article notes:

No one will be allowed to read or download work from the repository, because that would breach publishers' copyright. Instead, Malamud envisages, researchers could crawl over its text and data with computer software, scanning through the world's scientific literature to pull out insights without actually reading the text.

The thing is, if anyone were by any chance interested in reading the full text, there's an obvious place to turn to. After all, the mining is carried out using papers held by Sci-Hub, so…

Follow me @glynmoody on Twitter, Diaspora, or Mastodon.

Filed Under: academic papers, carl malamud, copyright, india, journals, research, science, tdm, text and data mining
Companies: sci-hub

PLOS ONE Topic Pages: Peer-Reviewed Articles That Are Also Wikipedia Entries: What's Not To Like?

from the good-for-the-academic-career,-too dept

It is hard to imagine life without Wikipedia. That's especially the case if you have school-age children, who now turn to it by default for information when working on homework. Less well-known is the importance of Wikipedia for scientists, who often use its pages for reliable explanations of basic concepts:

Physicists -- researchers, professors, and students -- use Wikipedia daily. When I need the transition temperature for a Bose-Einstein condensate (prefactor and all), or when I want to learn about the details of an unfamiliar quantum algorithm, Wikipedia is my first stop. When a graduate student sends me research notes that rely on unfamiliar algebraic structures, they reference Wikipedia.

That's from a blog post on the open access publisher Public Library of Science (PLOS) Web site. It's an announcement of an interesting new initiative to bolster the number of physicists contributing to Wikipedia by writing not just new articles for the online encyclopedia, but peer-reviewed ones. The additional element aims to ensure that the information provided is of the highest quality -- not always the case for Wikipedia articles, whatever their other merits. As the PLOS post explains, the new pages have two aspects:

A peer-reviewed 'article' in [the flagship online publication] PLOS ONE, which is fixed, peer-reviewed openly via the PLOS Wiki and citable, giving information about that particular topic.

That finalized article is then submitted to Wikipedia, which becomes a living version of the document that the community can refine, build on, and keep up to date.

The two-pronged approach of these "Topic Pages" has a number of benefits. It means that Wikipedia gains high-quality, peer-reviewed articles, written by experts; scientists just starting out gain an important new resource with accessible explanations of often highly-technical topics; and the scientists writing Topic Pages can add them to their list of citable publications -- an important consideration for their careers, and an added incentive to produce them.

Other PLOS titles such as PLOS Computational Biology and PLOS Genetics have produced a few Topic Pages previously, but the latest move represents a major extension of the idea. As the blog post notes, PLOS ONE is initially welcoming articles on topics in quantum physics, but over time it plans to expand to all of physics. Let's hope it's an idea that catches on and spreads across all academic disciplines, since everyone gains from the approach -- not least students researching their homework.

Follow me @glynmoody on Twitter or identi.ca, and +glynmoody on Google+

Filed Under: peer review, science, wikipedia
Companies: plos

Leading Biomedical Funders Call For Open Peer Review Of Academic Research

from the nothing-to-hide dept

Techdirt has written many posts about open access -- the movement to make digital versions of academic research freely available to everyone. Open access is about how research is disseminated once it has been selected for publication. So far, there has been less emphasis on changing how academic work is selected in the first place, which is based on the time-honored approach of peer review. That is, papers submitted to journals are sent out to experts in the same or similar field, who are invited to comment on ways of improving the work, and on whether the research should be published. Traditionally, the process is shrouded in secrecy. The reviewers are generally anonymous, and the reports they make on the submissions are not made public. Now, however, the idea of making peer review more transparent as part of the general process of becoming more open is gaining increasing impetus.

A couple of weeks ago, representatives of two leading biomedical funders -- the UK Wellcome Trust and the Howard Hughes Medical Institute -- together with ASAPbio, a non-profit organization that encourages innovation in life-sciences publishing, wrote a commentary in Nature. In it, they called for "open review", which, they point out, encompasses two distinct forms of transparency:

'Open identities' means disclosing reviewers' names; 'open reports' (also called transparent reviews or published peer review) means publishing the content of reviews. Journals might offer one or the other, neither or both.

In a 2016 survey, 59% of 3,062 respondents were in favour of open reports. Only 31% favoured open identities, which they feared could cause reviewers to weaken their criticisms or could lead to retaliation from authors. Here, we advocate for open reports as the default and for open identities to be optional, not mandatory.

The authors of the commentary believe that there are a number of advantages to open reports:

The scientific community would learn from reviewers' and editors’ insights. Social scientists could collect data (for example, on biases among reviewers or the efficiency of error identification by reviewers) that might improve the process. Early-career researchers could learn by example. And the public would not be asked to place its faith in hidden assessments.

There are, of course risks. One concern mentioned is that published reviews might be used unfairly in subsequent evaluation of the authors for grants, jobs, awards or promotions. Another possibility is the 'weaponization' of reviewer reports:

Opponents of certain types of research (for example, on genetically modified organisms, climate change and vaccines) could take critical remarks in peer reviews out of context or mischaracterize disagreements to undermine public trust in the paper, the field or science as a whole.

Despite these and other concerns mentioned in the Nature commentary, an open letter published on the ASAPbio site lists dozens of major titles that have already instituted open reports, or promise to do so next year. As well as that indication that open reports are passing from concept to reality, it's worth bearing in mind that the UK Wellcome Trust and the Howard Hughes Medical Institute are major funders of biomedical research. It would be a relatively straightforward step for them to make the adoption of open reports a condition of receiving their grants -- something that would doubtless encourage uptake of the idea.

Follow me @glynmoody on Twitter or identi.ca, and +glynmoody on Google+

Filed Under: biomedical, open access, open peer review, research, science
Companies: asapbio, howard hughes medical institute, wellcome trust

Study Shows That Wartime Program To Abolish Copyright On German Science Books Brought Significant Benefits To US

from the perhaps-we-should-do-it-more-often dept

As Techdirt readers know, there is a ratchet effect that means copyright always gets longer and stronger. As well as being inherently unfair -- why must the public always lose out when copyright law is changed? -- there's another unfortunate consequence. If the term or breadth of copyright were reduced from time to time, we would be able to test the effects of doing so on things like creativity. For example, if it turned out that shortening copyright increased the number of works being produced, then there would be a strong argument for reducing it further in the hope that the effect would be strengthened. The fact that we have been unable test this hypothesis is rather convenient for copyright maximalists. It means they can continue to call for the term of copyright to be increased without having to address the argument that this will cause less creativity, or reduce access to older works.

Even though it is not possible to test the effects of reduced copyright directly, two US academics, Barbara Biasi and Petra Moser, have spotted a clever way of investigating the idea indirectly, in the field of science publishing. As they write in a post on CEPR's policy portal, in 1942 the US Book Republication Program (BRP) allowed US publishers to reprint exact copies of German-owned science books, effectively abolishing copyright for that class of works. They have looked at what impact this dramatic change had on the use of those reprinted works by scientists. Comparing citation rates before and after the BRP was introduced is not enough on its own: citation rates fluctuate, so it is necessary to compare the BRP citation rate with something else. The researchers' solution is to look at the citation rate of Swiss books from the same time:

This approach addresses the issue that English-language citations may have increased mechanically after 1942, if English-language scientists published more after the war. Like German scientists, Swiss scientists were leaders in chemistry and mathematics and wrote primarily in German, but due to Switzerland's neutrality, Swiss-owned copyrights were not accessible to the BRP. [Office of Library Services] estimates of a matched sample of BRP and Swiss books (in similar fields and with similar levels of pre-BRP non-English citations) confirm the significant increase in citations in response to the BRP.

Specifically, there was a 67% increase in citations of BRP books compared to similar Swiss books. The research suggests this was driven largely by the 25% drop in average prices seen after the BRP scheme was introduced. The reduction in price seems to have allowed a wider range of US libraries to purchase the more affordable BRP texts, whereas Swiss books remained concentrated in the holdings of two wealthy research libraries (Yale and Chicago). Better access was correlated with more citations: the data shows that the latter increased most near the locations of BRP libraries. The researchers conclude:

In the context of contemporary debates, our findings imply that policies which strengthen copyrights, such as extensions in copyright length, can create enormous welfare costs by discouraging follow-on science, especially among less affluent institutions and scientists.

Critics might point out that this is just one study of one rather specific area. But that's an argument for reducing copyright terms, perhaps on a trial basis, to see whether the results of this research are confirmed. However, the copyright ratchet will never allow that, not least because the companies involved probably know it would confirm that constantly strengthening copyright is bad for everyone except themselves.

Follow me @glynmoody on Twitter or identi.ca, and +glynmoody on Google+

Filed Under: benefits, copyright, germany, public domain, science, us

Genome Of A Man Born In 1784 Recreated From The DNA Of His Descendants

from the I-am-your-(great-great-great-grand)-father dept

The privacy implications of collecting DNA are wide-ranging, not least because they don't relate solely to the person from whom the sample is taken. Our genome is a direct product of our parents' genetic material, so the DNA strings of siblings from the same mother and father are closely related. Even that of more distant relations has many elements in common, since they derive from common ancestors. Thus a DNA sample contains information not just about the donor, but about many others on the relevant family tree as well. A new paper published in Nature Genetics (behind a paywall, unfortunately) shows how that fact enables the genomes of long-dead ancestors to be reconstructed, using just the DNA of their descendants.

As an article in Futurism explains, the unique circumstances of the individual chosen for the reconstruction, the Icelander Hans Jonatan, aided the research team as they sought to piece together his genome nearly two centuries after his death in 1827. The scientists mainly came from the Icelandic company deCODE Genetics, one of the pioneers in the world of genomics, and highly-familiar with Iceland's unique genetic resources. The following factors were key:

For one, Jonatan was the first Icelandic inhabitant with African heritage. Iceland also boasts an extensive and highly detailed collection of genealogical records. The combination of Jonatan's unique heritage and the country's record-keeping for inhabitants' family trees made this remarkable recreation possible.

For cultural and historical reasons, Iceland has one of the most complete genealogical records of any nation. This allowed the research team to establish with high probability 788 of Jonatan's descendants. Samples were taken from 182 of those individuals and then genotyped -- a kind of DNA screening. The deCODE group picked out those genomes most likely to provide the longest DNA sequences that had been passed down through the generations from Jonatan's mother, by looking for fragments of African-pattern chromosomes amidst the otherwise European genetic material. The full genomes of 20 of those 182 were sequenced, and then the parts derived from Jonatan's African ancestry pieced together to recreate 38% of his mother's DNA. From this, the researchers were able to establish that Jonatan's mother was probably from the African region spanned by Benin, Nigeria and Cameroon.

This kind of large-scale reconstruction in the absence of physical samples has never been achieved before, and is certainly a major triumph of biological and computational technology. An important question is whether this is a one-off, made possible by the unique circumstances of Jonatan's life, or whether it could be applied more widely. According to the Futurism article:

Theoretically, a technique like this could help researchers create "virtual ancient DNA," which would allow scientists to recreate the DNA of historical figures. Agnar Helgason of deCODE stated that "Any historic figure born after 1500 who has known descendants could be reconstructed."

While it's exciting, there are still major hurdles to overcome in terms of the potential future applications. The quantity, scale, and detail of the DNA from living ancestors required to recreate a person's DNA make it impractical for use within most families. Additionally, with each new generation identifiable DNA fragments get smaller and more difficult to work with.

As DNA sequencing becomes cheaper and more accurate, it will be possible to carry out DNA profiling and collection faster and more economically. Similarly, as computational power increases, chromosome fragments can be analyzed and stitched together more easily. In due course, these kinds of genomic reconstructions will probably become more common. Already, deCODE's research confirms how DNA can establish the connections not just between present-day members of a family, but also with those long dead. When unexpected patterns of maternity or paternity are revealed, they will bring with them who knows what social consequences for their descendants.

Follow me @glynmoody on Twitter or identi.ca, and +glynmoody on Google+

Filed Under: dna, genetics, genomes, science