How Not to Deploy a Twitter Feature
Twitter announces a new feature is rolling out for iOS, the ability to record audio tweets. It demonstrates this with an uncaptioned video of its Twitter avatar making bird noises:
Tweets with audio are rolling out on iOS and we only have one thing to say about it pic.twitter.com/CZvQC1fo1W
Twitter’s support account acknowledges it is inaccessible and seemingly alludes to a lack of plan to address it:
This is an early version of this feature. Making these types of Tweets accessible to everyone is important and we’re exploring ways to make that happen.
Twitter’s dedicated accessibility account has not tweeted anything about it. In fact, its last tweet was on Global Accessibility Awareness Day, three weeks prior to this new feature being launched.
Adding descriptions to images is a great way to include everyone in your conversation. These descriptions, aka alt-text, enable folks who use screen readers to interpret images in Tweets. Starting today, you no longer need a setting to add alt text and it's available on 📱 & 💻. pic.twitter.com/wRDJZwSihL
Twitter Able? Nothing since June 11.
The project lead is proud and, based on the team involvement, rightly so for getting solid representation.
Voice Tweets made my timeline fun today. This is a great example of allowing a team to chase after a customer problem and use their gut. Designers, PM, Eng, and Research worked in lock step and mgmt got out of the fucking way. Congrats to the team! 👏🏾👏🏾👏🏾
Users rightly excoriate Twitter for failing to include any accessibility affordances with this new feature.
So basically @Twitter created a new feature, did not consult with @TwitterA11y or @TwitterAble, and their excuse for it being inaccessible is that it's new. And that they are working out ways to implement accessible features *after* the fact, which- after is already too late.
No. This is not how we do things. We do not release entire new features on a MASSIVE platform (whether we're "testing" them or not) without even a workaround to make them accessible.
[Quote tweeted audio: Testing. 1, 2, 1, 2. Mic check. Is this thing on? Tweet with your voice.] twitter.com/TwitterSupport/status/12733…
If you are not prepared to use your tweet as the transcript of your audio clip, then do not do it. twitter.com/Twitter/status/127331…
And shame on @TwitterA11y for not getting ahead of this and insisting there be some form of captioning. pic.twitter.com/PgsCmgm0wm
You've proven that failure to prioritise accessibility (and use resources like @arhayward) has serious consequences, and learned that you've given people a tool to exclude Deaf people from your platform.
You HAVE captioning, but it's locked away in studio behind ads accounts.
Sometimes I wish designers and developers would get out of the f@*king way of accessibility. Diversity really does matter. #a11y
I don't know who needs to read this but I have never wanted to hear a voice tweet and will never listen to them. Also, how long until the twitter app backdoors autoplay on these and then buries option to control in "accessibility" options? Also also, this idea #SUX.
Really excited for Twitter to become even less accessible to deaf and hard of hearing people with this voice notes feature. Love that for us.
Hello, new tech feature, let's do a speculative harm analysis to consider ways that terrible people might use you (and maybe what could be done about it?): a thread. twitter.com/Twitter/status/12733…
As a deaf Twitter user: I vote no against voice tweets.
Plus, you know that voices have unique signatures (certainly of interest as we wear masks that make traditional modes of surveillance harder)? You’re handing that data over when you use voice tweets.
As a Disabled software engineer directly affected by this blatantly inaccessible voice tweet feature, it is incredibly soul crushing to see a massive company with endless resources release shit like this and say "We'll figure it out later."
This Time, Twitter has excluded those who are D/deaf, have auditory processing differences, or — like me — prefer text because it's faster and less distracting. Design that excludes any of us harms us all by making the platform less usable, therefore less diverse and valuable.
Twitter's core problem isn't accessibility.
The reason they ship inaccessible products as "tests," with no roadmap for making them accessible, is bad product management and design.
The reason they've gone this long without a formal accessibility organization is bad leadership.
After many people called Twitter out for deploying this feature, the project lead repeatedly argues that this is an experiment. An experiment they felt must be shipped immediately instead of in a year to build the accessibility. And yes, the first and second tweet of this set have been deleted since I wrote this post.
We formed a design systems team recently that is responsible for accessibility across Twitter. This is a test. Hypothetically a test can go away in a few weeks after we launch. The work you’re describing is important. However that would have meant shipping in 2021/22.
@dj_diabeatic This is fair feedback. We weren’t even sure how people would use the feature once in the world. Adding captioning of UGC isn’t trivial for us given our tech stack. We wanted to prove the concept and learn before investing into what is essentially a test.
Feedback heard. This is an incomplete product in many ways because it’s an experiment. We want to see if people enjoy using it, then we’ll learn, and iterate. If we included all the things asked of us, we would not ship for another year. 2 wks after we may learn no one uses it.
Here’s where all of that falls down. Nobody appears to have requested this feature. In fact, you can do it today by recording a video on your iOS device and keeping your thumb over the camera lens. Arguably, the only new part of this experiment is an animated avatar instead of a black screen.
This was built as if it was an MVP from a start-up. Twitter is not a start-up. Its user base, net worth, geo-political influence, all tell us this is a mature business, if only by age and mass. Yet it continues to behave like a start-up.
In a mature business, you go back and fix broken code, enhance existing features. You deploy new features based on requests, market research, and competitive analysis.
Twitter could have deployed this same feature by adding a “disable camera” or “audio only” button to its video recording feature in the iOS app. That is an MVP. It likely requires far less code and can be rolled out quietly, without all the fanfare of a new feature and its accompanying scrutiny. I suspect adding the avatar and the surrounding animation would have been trivial.
Then that team could have moved on to working with the speech API built into iOS or Android, making it possible to auto-generate captions from voice recognition software already on the phone, giving the user the ability to correct it.
That, in my opinion, would have been a far better use of their time. It would have given them a quick-hit project to test the waters, while going full-steam toward supporting a significant feature that currently leaves video on Twitter limited to those who are able to listen.
None of this speaks to the potential abuse, of course. Twitter’s algorithms for identifying hate speech lean on parsing text. There is likely some image recognition in there. But have they stood up any process to parse audio?
If I report an audio tweet because I feel someone is at risk of self-harm, or there is hate speech, or targeted harassment, does Twitter have the human capacity to listen to all of them?
I’ll close with Chancey’s tweet, which came in just before I hit the button to publish.
Tech industry culture overwhelmingly values fast deployment above inclusive, accountable development. I know it's hard to be the person in the meeting who demands to name & stop encoded bias. But if you're listening — we need you to be that voice in the room when we aren't there.
If you are generally interested in making your tweets more accessible, I gathered some tips under an obvious post name: Improving Your Tweet Accessibility
Update: 19 June 2020
Quite a lot has happened since I wrote this Wednesday night and published it Thursday morning. By Thursday afternoon the lead on the project (who is also Twitter’s Head of Design and Research) posted a mea culpa:
I appreciate the feedback and direct conversation about #a11y from our passionate community. It's clear we have a lot of work ahead to make Twitter more inclusive for people with disabilities. I will advocate for a11y to be part of our design from the beginning of all projects.
One can probably make a convincing case that his title is not user-centered design and user research, so arguably designing new features and throwing them out into the wild to research how popular they are is on-brand for his current title.
Not everybody was on board with the statement.
I mean, cool, thanks for somewhat acknowledging the issue but at this point it goes beyond accessibility. Your responses and those from official Twitter accounts were garbage and it took thousands of tweets for y'all to see it. All should apologize and then take action. twitter.com/dantley/status/127366…
Concrete timelines, especially for audio tweets, or this statement is performative.
Talk is cheap.
While an open secret for a while, this is the first time most users became aware that Twitter does not have a dedicated accessibility team (nor process). Instead, employees kick in time where they can and often never know when new features are coming.
Just to clarify, given that this seems to have gained some traction… we are volunteers in so much as the work we do is notionally on top of our regular roles, rather than being full time.
We are all otherwise paid employees – Twitter is not outsourcing unpaid labour!
Twitter also had its own crow-eating, but only from its support account:
We're sorry about testing voice Tweets without support for people who are visually impaired, deaf, or hard of hearing. It was a miss to introduce this experiment without this support.
Accessibility should not be an afterthought. (1/3) twitter.com/dantley/status/12736…
We’ve fixed several issues related to vision accessibility, including making voice Tweets identifiable on the timeline and making accessibility improvements to the voice Tweet experience. These updates will be available in a forthcoming iOS release. (2/3)
We’re already exploring ideas for how we could support manual and auto transcriptions.
We’re also looking at how we can build a dedicated group to focus on accessibility, tooling, and advocacy across all products, in partnership with the @TwitterA11y and @TwitterAble teams. (3)
The takeaway here should be that Twitter is not pulling the feature, it will release accessibility fixes for iOS only, it is not building a dedicated accessibility team, and therefore Twitter is going to continue to treat accessibility as an afterthought.
Probably the best, most honest, and most hopeful response from Twitter came from one of the designers on the team as a nine-part thread (actually a two-part Voice Tweet with nine tweets acting as its transcript):
My personal response to the Deaf/HOH Community:
Uh point blank period, we fucked up, I fucked up. Uh we launched a test and we should have included accessibility features in that test. That was a huge error, and one I personally have made myself sit [cont] pic.twitter.com/avAImLl9t7
It is Friday afternoon, and this inaccessible feature has gotten some press:
- Twitter Just Rolled Out a Feature That’s Inaccessible to Disabled Users at Slate;
- Somehow, Twitter does not have a team dedicated to accessibility at TechCrunch;
- Twitter’s audio tweets revealed an accessibility miss, and now the company wants to fix it at The Verge.
I hope this cock-up on Twitter’s part helps drive home the value of building in accessibility from the start. Some of its team seems to understand it, but Twitter has a long history of not getting it as a company (five years from supporting images to allowing alt text, four more to turning it on by default, for example). The best thing the rest of us can do is not make Twitter’s mistakes.
Update: 20 June 2020
I did the same and in under two minutes had a response with a video that contained nearly accurate captions from my own audio tweet test.
👋 we updated the voice tweet to include *automatic* captions… pic.twitter.com/u69t9We1Yk
There are other tools that provide automated captions as well as an API for third parties to access them. While automated captions are not a good overall solution, Twitter could at least lean on these in the short term, maybe even letting users generate and correct them before posting.
Obviously this will not address audio clips that are not voice, nor voice under non-ideal conditions, but it is dramatically better than what Twitter offered at launch or promises it will deliver in the near term.
Update: 2 July 2020
Twitter’s Voice Tweets Upset Users Again Over Accessibility at Wall Street Journal.
Update: 16 July 2021
It took just over a year, but Twitter has enabled captions for Voice Tweets. Which are automated captions. And cannot be edited. And are only for iOS. And have to be enabled in accessibility settings. And which are preceded by the text
(Transcriptions auto-generated by Microsoft). And which are not transcripts.
But yay, progress! After only a year!
We have news! Starting today, when you create a Voice Tweet (only available on iOS for now), auto captions will be generated in supported languages.
Access captions from the CC button on Web & mobile device Accessibility settings.
Try it out. How’d it go? pic.twitter.com/U2g3V1oeUA
Thanks for taking the time. We appreciate it.
Thank you for this comprehensive summary of the issue. Twitter introduced the audio only live feature in September 2018, which allows what you suggest – an audio only feed of the Periscope (video) live. So, they have been dabbling with this and testing it for some time. Which makes it even more unbelievable that they have hit go on the audio tweet feature, even if only to limited users, without consideration of accessibility….