Quantcast
Channel: Stefan Bohacek
Viewing all articles
Browse latest Browse all 71

Alt Text Health Check image accessibility report #1

$
0
0

A tinted screenshot of three posts made by the @AltTextHealthCheck bot showing breakdowns of the data it collects.

@AltTextHealthCheck, an open-source bot, created by Stefan Hayden, that tracks the percentage of images that were posted with a description across the mastodon.social fediverse instance, has been running for about three months now.

Tracked 21,958 posts across 2,921 instances.

Federated: 31% – Found 8,072 descriptions set on 25,743 images.

Local: 17% – Found 1,152 descriptions set on 6,877 images.

— Alt Text Health Check (@AltTextHealthCheck@mastodon.social)2023-10-26T12:00:08.795Z

Tracked 21,806 posts across 2,915 instances.

Federated: 32% – Found 8,322 descriptions set on 25,652 images.

Local: 18% – Found 1,243 descriptions set on 7,025 images.

— Alt Text Health Check (@AltTextHealthCheck@mastodon.social)2023-10-24T12:00:06.818Z

I was curious if we can see any trends yet in the data it has collected.

First, here’s a breakdown of captioned images posted across mastodon.social’s federated timeline, meaning, these are posts from the wider fediverse that mastodon.social knows about, including from its own users. That includes other Mastodon servers, but also other fediverse platforms.

The chart is ever-so-slightly trending downwards, with the median value of 36%. Considering that means that 64% of images posted in the fediverse are completely blank to blind people, the number is pretty bad.

Percentage of captioned images
DateImages captioned (%)
8/6/202339%
8/7/202335%
8/8/202336%
8/9/202337%
8/10/202334%
8/11/202336%
8/12/202338%
8/13/202337%
8/14/202335%
8/15/202335%
8/16/202336%
8/17/202336%
8/18/202336%
8/19/202339%
8/20/202340%
8/21/202336%
8/22/202336%
8/23/202336%
8/24/202336%
8/25/202337%
8/26/202339%
8/27/202338%
8/28/202337%
8/29/202336%
8/30/202336%
8/31/202336%
9/1/202337%
9/2/202339%
9/3/202340%
9/4/202336%
9/5/202335%
9/6/202335%
9/7/202336%
9/8/202336%
9/9/202338%
9/10/202337%
9/11/202335%
9/12/202334%
9/13/202335%
9/14/202336%
9/15/202337%
9/16/202339%
9/17/202338%
9/18/202336%
9/19/202335%
9/20/202334%
9/21/202335%
9/22/202333%
9/23/202337%
9/24/202334%
9/25/202336%
9/26/202336%
9/27/202335%
9/28/202336%
9/29/202336%
9/30/202340%
10/1/202338%
10/2/202337%
10/3/202337%
10/4/202336%
10/5/202337%
10/6/202338%
10/7/202339%
10/8/202339%
10/9/202336%
10/10/202337%
10/11/202337%
10/12/202336%
10/13/202338%
10/14/202340%
10/15/202334%
10/16/202335%
10/17/202336%
10/18/202334%
10/19/202335%
10/20/202336%
10/21/202339%
10/22/202338%
10/23/202336%
10/24/202335%
10/25/202334%
10/26/202331%

Things look even worse on mastodon.social alone, where the median over the past few months is only at 21.5%.

Percentage of captioned images
DateImages captioned (%)
8/6/202325%
8/7/202322%
8/8/202323%
8/9/202322%
8/10/202322%
8/11/202321%
8/12/202324%
8/13/202323%
8/14/202322%
8/15/202322%
8/16/202321%
8/17/202321%
8/18/202322%
8/19/202324%
8/20/202325%
8/21/202322%
8/22/202322%
8/23/202322%
8/24/202322%
8/25/202324%
8/26/202326%
8/27/202324%
8/28/202323%
8/29/202321%
8/30/202323%
8/31/202323%
9/1/202322%
9/2/202324%
9/3/202324%
9/4/202322%
9/5/202321%
9/6/202320%
9/7/202322%
9/8/202322%
9/9/202324%
9/10/202324%
9/11/202321%
9/12/202321%
9/13/202320%
9/14/202321%
9/15/202321%
9/16/202324%
9/17/202324%
9/18/202320%
9/19/202320%
9/20/202319%
9/21/202320%
9/22/202319%
9/23/202320%
9/24/202318%
9/25/202320%
9/26/202321%
9/27/202319%
9/28/202320%
9/29/202320%
9/30/202323%
10/1/202322%
10/2/202321%
10/3/202321%
10/4/202320%
10/5/202320%
10/6/202320%
10/7/202323%
10/8/202324%
10/9/202319%
10/10/202319%
10/11/202319%
10/12/202318%
10/13/202322%
10/14/202323%
10/15/202322%
10/16/202319%
10/17/202321%
10/18/202319%
10/19/202320%
10/20/202320%
10/21/202323%
10/22/202324%
10/23/202320%
10/24/202321%
10/25/202320%
10/26/202317%

We can also see a stronger trend towards fewer captioned images, which is worrisome. Note that the bot only checks for the presence of the image description, not its quality. Anecdotally, I’ve seen descriptions that are not very detailed, sometimes seemingly just placeholders. This could potentially hint at the actual number of useful image descriptions being even lower.

About two and a half months ago I started the Alt Text Hall of Fame project, which aims to bring more attention to the issue of uncaptioned images posted online.

The website and its associated account, showcases and celebrates the ingenuity and creativity that often goes into describing some of the more challenging images.

My new project, AltTextHallOfFame.org, is a celebration of the effort, ingenuity, and creativity that goes into making the web a friendlier and more inclusive place, one captioned image at a time.

— Stefan Bohacek (@stefan@stefanbohacek.online)2023-08-07T12:32:52.646Z

Happy !

Here’s a little challenge. How would you caption this painting from Oscar Howe, a Yanktonai Dakota artist?

— Alt Text Hall of Fame (@alttexthalloffame@mastodon.social)2023-10-09T14:54:08.343Z

The data used for this article was retrieved from @AltTextHealthCheck‘s posting history using Mastodon’s API, and is available for download:

UPDATE 10/28/2023: The bot now shows a breakdown by the instance. Nice!

Results from tracking 26,309 images accross 1,627 instances.

Then filtering down to the top 1000 instances by active user from instances.social who hhad posted atleast 32 images (2x the average).

— Alt Text Health Check (@AltTextHealthCheck@mastodon.social)2023-10-28T02:35:57.404Z


Viewing all articles
Browse latest Browse all 71

Trending Articles