Member-only story
2021 accessibility predictions
Yeah, machine learning and personalization made the list, but I also have a couple of other thoughts.
Authors note: Because of Medium’s refusal to address its accessibility issues for both authors and readers, I’ve moved my last three years of blogs to Substack. Please sign up there for notices of all new articles. Also, I will be updating older articles (like this one) and the updates will only be published on Substack. Thank you for your continued readership and support.
It’s time for the annual “Year-End” accessibility predictions.
Machine Learning
Machine learning is going to continue to be exceptionally important to the advancement of accessibility. The challenges of scaling accessibility in a world where only 30 % of the tests are automated are well documented in this note from the W3C titled Challenges with Accessibility Guidelines Conformance and Testing, and Approaches for Mitigating Them.
TL;DR — Software *used* to be monolithic. One major release every 12–18 months, and a few minor releases focused on specific areas. That is no longer the case.
SaaS can release hundreds of times per week.
Native apps releases happen weekly-ish (more often if needed)
Improved accessibility test automation is essential to avoid accessibility regressions
With the world going to SaaS and native app in droves, accessibility can't live successfully in a world where 70 % of the tests are manual. It’s not enough to GET software accessible; you have to KEEP software accessible. Machine learning will allow us to crush the 30/70 split between automated and manual testing. Image processing, natural language processing, pattern recognition, and supervised and unsupervised learning will be essential to shifting that percentage towards the automated side. That way, the tests can be run with every, single check-in preventing good accessibility from being replaced with broken accessibility.
My accessibility dream is a world that any check-in that breaks an automatic test is rejected. Something like this:
