UPDATES: In January 2020 we published a more detailed analysis of the NAD v. Harvard consent decree, and in February MIT announced a similar settlement.
Harvard and the National Association of the Deaf (NAD) have finally reached a proposed settlement in their four-plus year legal wrangle over whether or not Harvard should be required to provide accurate captions on their video content. We will have a more detailed article describing the specific provisions of the settlement later this week, but in the meantime here are three key takeaways based on our close reading of the 15 page proposed consent decree.
Captions are Definitively Required for All Public Video Content
For the last few years colleges and universities have relied on a variety of arguments to avoid captioning video content, ranging from “we have too much video to caption all of it,” to “it’s difficult or impossible for us to identify which videos need captioning,” to “we’re waiting for clarification of the standards from the U.S. Federal Government.” This settlement demonstrates that even if you have one of the best university legal teams in the country, those arguments are not defensible.
In the consent decree Harvard agrees to provide captions for all videos posted on any public-facing Harvard website or web application. This includes “department sponsored student organization content” and MOOC content. Harvard and their university-sponsored student organizations produce a huge amount of video. The argument that you produce too much video to keep up with captioning all of it, or that some videos are not important enough to warrant captions, clearly does not hold up in court. If a video is important enough to post on one of your websites, you must make it accessible with captions.
Accuracy Matters: “Captions” Means Real, High-Quality Captions
Another question that frequently comes up when organizations grapple with large quantities of video is, “does the quality of the captions matter?” Many video platforms generate “auto captions” using automatic speech recognition and machine learning algorithms. Do these machine-generated captions meet the legal standards for captions? The Harvard consent decree adds to the large body of legal precedent indicating that the answer is clearly “no, machine-generated captions are not sufficient.”
Complicating matters further, in the years since the Harvard case was filed many new captioning service vendors have sprung up, hoping to capitalize on what they undoubtedly viewed as a “captioning gold-rush.” These new vendors typically use crowd-sourced labor to attempt to clean-up machine-generated captions. The result is usually better than machine-generated captions, but it is typically not good enough to reach the standard of equally effective communication, as required by law.
The NAD felt that the issue of captioning quality was so important that they incorporated the concept of accuracy in the definition of captions included in the consent decree that Harvard’s lawyers have signed. The definition states, “… at an accuracy rate equal to that offered by a vendor captioning service such as….” They go on mention a specific vendor that is comparable to AST for accuracy. While I think it was inappropriate to name a single vendor in the proposed consent decree, the intent of the definition is clear: less experienced, second-tier vendors like those that use crowd-sourcing do not provide the accuracy needed to meet this definition of “captioning.”
Timeliness and Responsiveness Also Matter
The NAD v. Harvard case adds one more constraint that colleges and universities need to keep in mind. High-quality captions should be provided at the time the video is posted, and if you have legacy content that was not captioned (or was poorly captioned) you should put in place a process that allows potential viewers of the video to request proper captions, and receive them promptly. Harvard agreed to a turnaround time of five business days for captioning legacy content when captions are requested through a public request process.
If you’re putting in place a video accessibility request process for your own organization, the process described in the Harvard consent decree should serve as a good model. The request form should be easy to find and easy to use (and of course it should meet WCAG 2 Level AA accessibility guidelines). A five business day turnaround time leaves very little cushion for delays on the part of staff who may be monitoring the requests, so you should have your captioning service vendor accounts set up in advance, and you should utilize workflow integrations that streamline the process of requesting and receiving captions.
This is another reason why working with a top-tier vendor, and one that specializes in working with educational organizations like AST, is critical. If you want to respond to video accessibility requests from members of your community promptly, you need to have a vendor that responds promptly to your needs, and one that does so without taking shortcuts that cause accuracy to suffer.