Last March, when a judge ordered the University of California–Berkeley to make 20,000 videos and podcasts accessible to people with disabilities, the university balked. The videos and audio files contained lectures by Berkeley professors that the university wanted to make available to anyone, as an act of public outreach, but disability rights groups had sued on the grounds that the materials lacked captions, were often incompatible with the screen readers that blind people use to access the Internet, and other related issues. So the judge ordered the university to make the materials accessible. Instead, Berkeley shut the program down, locking the formerly public materials behind a firewall. The university said it was just too expensive to retrofit accessibility into their public program.
The Berkeley story shows what happens when you don’t build accessibility into a system from the beginning. More recently, the World Wide Web Consortium (W3C), a standards body that pretty much defines what the Internet is and how it works, has been engaged in a debate about whether the next generation of online videos will be accessible. If we don’t want to be forced to choose between excluding disabled users or taking tens of thousands of videos away from the public, the W3C had better get this one right.
Some background: The Internet works because computers all over the world can talk to each other, and that only happens because corporations, universities, governments, and others agree on basic standards for how devices connect. Since the mid-1990s, the W3C has served as the organization overseeing that process. Every time you open up a browser, click on a link, read an alt-tag on an image, edit HTML, or do anything else online, you’re probably interacting with a system based on standards emerging from W3C. Generally speaking, it’s a body that tries to use debate to locate consensus among its members. When it comes to the next set of standards for online video, however, that consensus has been harder to find.
This summer, the W3C released a new set of guidelines around “Encrypted Media Extensions,” a standard that allows streaming video in HTML5 to contain Digital Rights Management software. DRM technologies are ways that copyright owners try to limit users’ ability to copy, modify, or share their work. As you might imagine, companies like Netflix and Amazon are very invested in making sure that their copyright is protected. Anti-DRM proponents, on the other hand, argue that DRM requires making browsers more vulnerable to hackers and stifles competition (if you can’t read a book on the e-reader of your choice, the cause is likely DRM). It’s a big debate that, in various ways, has been happening for over 15 years, starting with physical CDs and now moving to the Internet. In this latest chapter of the debate, modifying online video for the purposes of accessibility has emerged as a central divisive issue.
In this latest chapter of the debate, modifying online video for the purposes of accessibility has emerged as a central divisive issue.
To understand these issues better, I spoke to Cory Doctorow, the well-known author and journalist who is working with the Electronic Frontier Foundation against the proliferation of DRM. Doctorow argues that these new standards from W3C will create criminal and civil risk for people who want to adapt video for accessibility reasons. I also talked to Judy Brewer, director of the Web Accessibility Initiative at W3C. Brewer is deeply committed to accessibility of the Web, but believes that some current concerns reflect a misunderstanding of how the technology works.
Accessible Web video is important. Deaf people need captions. Blind people need audio descriptions. People with photosensitivity need strobe warnings. There are tools that can modify video so that colorblind people can distinguish shades (especially important for interactive websites where, for example, everything on sale is marked in red). Controls for videos need to work for people using diverse interfaces. Once online video embraces these new DRM standards, will it still be possible (without breaking the law) for people to modify videos to put in these kinds of accessibility features? In a statement for the Electronic Frontier Foundation, Doctorow argues that the new rule will make it dangerous for groups or individuals to modify inaccessible video to accept captions, strobe warnings, or audio descriptions, especially if they want to automate the process.
Automation is a big part of the story here. Remember those 20,000 Berkeley videos? The university’s argument was that it would be too expensive to hire people to fix the accessibility issues. They’re right. Today, moreover, the automated tools for turning speech into text are pretty lousy (the Deaf community calls automated captions “craptions“), but the technology is improving rapidly. In the near future, Doctorow says, machine learning is going to open up new horizons for making videos accessible—that is, unless DRM gets in the way.
Doctorow and I talked at length over the phone about DRM, the W3C, and accessibility. He acknowledges that, once you have the information—captions, audio descriptions, the location of strobes, etc.—the new standards allow you to input that information, assuming you hold the copyright. “If you have an army of humans and unlimited budget to make transcriptions, you’re good,” he says. But what if, he continues, “you want to remove DRM to run video through an automatic subtitling [program] or to add a description track?” Right now, of course, such automated programs aren’t good enough, but the technology is improving rapidly. When it’s ready, he says, “the W3C standards] don’t have a way to take the video, unencrypted, and send it to a machine learning algorithm to add those subtitles” without hazarding criminal charges for breaking DRM.
Brewer largely disagrees. For the last 20 years, Brewer has ensured that a W3C working group of accessibility experts reviews all specifications to make sure they support accessibility. The group has been involved, she tells me, throughout the W3C’s development of the new EME specification. Over the phone, she says that she’s seen enormous growth in understanding of disability and access within the industry. When she started, people were surprised disabled individuals were on the Web in the first place. Now, technology is much more sophisticated and the conversations about how to make the Web accessible are more specific.
Doctorow says machine learning is going to open up new horizons for making videos accessible—that is, unless DRM gets in the way.
In response to the EFF complaint, Brewer suggests that we tease apart the specific concerns rather than talking about accessibility too generally. In our phone call (and in a written statement about EME), Brewer says that some of EFF’s concerns aren’t really on the video-production side at all, and don’t have anything to do with the presence or absence of encryption. Risk of photosensitive seizure reactions (including seizures related to strobe effects), for example, could be mitigated through software on the users’ machines, rather than by embedding warnings in video files. As we went through, case by case, it became clear that, to a great extent, Brewer’s argument is that the onus for accessibility rests with the content creator (or rights holder), and that humans will continue to play a key role, even as auto-generated captioning (for example) improves. “Production of captions can be arranged by exchanging permissions with captioners,” she wrote in a follow-up email, “as is currently done in cases where captions are not generated in-house.” The key, she told me, is that the standard does not block access to accessibility information (captions, transcripts, audio description, etc.).
When I push Brewer on the automation issue, she admits that “it could be possible to construct very particular use cases” where DRM and automation conflict, “but I feel like that’s beyond what I’d ordinarily comment on for the standard [the new rules on DRM in Web video].” There’s nothing in the new standard, she made clear, that would stop an automated system from inputting accessibility information. The burden rests with content creators or rights owners to ensure that their product is accessible. For W3C, if a media company doesn’t add captions and DRM makes it hard for outsiders to add captions, the media company should be held accountable and pushed to do better.
Brewer has not allayed all my concerns, in part because the law on copyright (as opposed to laws on disability) suggests that legal obstacles to accessible video remain. It does seem that, if third parties want to make media accessible, the new standards will enable rights holders to punish those third parties under the Digital Millennium Copyright Act. We’ve seen companies use copyright to limit accessibility before. In 2014, Wired ran a story on the ways that publishers were limiting text-to-speech devices for e-books. Audiobooks are big business, after all, and a cheaper screen reader might have seemed threatening to publishers. Publishers deployed DRM to keep books inaccessible. We were having this debate back in 2003 as well, when early DRM was deployed to stop “Fansubbing,” the practice, especially among anime users, of inserting their own subtitles into Japanese-language videos. Video has only grown more essential to the Internet since then, both in terms of access and profits.
“Different forms of digital rights management already exist,” Brewer says. “What W3C can do in regards to EME is to confirm that accessibility support throughout the W3C standards ecosystem is leveraged in any potential use of DRM.” I have the sense that she does not want to get too deep into the core issue of whether there should be DRM in any particular system, but to make sure that she’s supporting accessibility in all contexts.
Doctorow, though, isn’t ready to give up the fight. We may have been having this debate for years, but “it’s taking on new urgency,” he says, “because of the domains in which DRM operates—tractors, heart monitors, and defibrillators. We’re one system away from a toaster that won’t toast third-party bread.”
“W3C is on the wrong side of history here,” Doctorow says.
