6/1/26 - May Comments

Scroll down for comments on May's shared articles. There are a lot of them!

In June and July, annual paid memberships are discounted at $40. It's a great time to support Disability Thinking Weekday. Just click the "Subscribe or Upgrade" button below to get started.

May 5, 2026 - Bad ideas and bad representation

About a disability activist seeing her image used in a public advocacy campaign without her consultation or consent, Mark Johnson says: "Amen on permission" ... but also adds: "There’s a big battle regarding pair of transit and they’re gonna name the biggest transit station in Denver after Wade Blank." Wade Blank was one of the early leaders of the disability activist movement in Denver, Colorado, that became ADAPT, fighting first for accessible public transportation, and then for home care and getting disabled people out of nursing homes and other institutions.

May 7, 2026 - Banning shock, protecting voting rights, and exploring disability fashion

Commenting my personal take on the article by Tanzila Khan on disability fashion in his Disability Debrief newsletter, Peter Torres Fremlin says that it was "An interesting piece for me to edit as my experience is much more like yours than Tanzila's!"

May 13, 2026 - Monthly Essay: About AI

As I hoped, the Monthly Essay on Artificial Intelligence drew some in-depth responses, including:

Greg Green, with some interesting cautionary comments:

Any time that I have used AI to "assist" in enhancing or generating artwork it consistently tries to remove physical disabilities from the artwork that I sample with.  Even with very specific prompts to identify a hemiparesis (even using plain language), or use of an orthotic for a missing leg, AI will revert the art to a character that does not have the disability described in the visual or text prompts. Essentially - AI erases the physical presence of disability without significant modeling and training on custom LLMs.  

When I use ChatGPT and Google Gemini for an inquiry, I have noticed a tendency for the AI to respond confidently with answers that are factually incorrect. When called out, it will state that it knows it was incorrect but provided the answer anyway.  

When I use ChatGPT and Google Gemini for a command, it will tell me that it has executed the command even though it has not.  The example might be that I want it to delete a previous query from the chat record, or I want it to cancel an appointment from my calendar.  Where it lacks authority to execute the command it will not tell you it is unable to do so, and will instead tell you that it has successfully completed the command. 

The problem is that AI is an unreliable source of data. It readily lies about information by fabricating data, quoting unreliable sources, or hallucinating. There is no vetting of sources, and you have nothing to substantiate that the response AI is giving you is not truthful unless you query further by seeking sources, fact-checking from reliable sources, or querying multiple times using different question sets designed to test for different results based on system bias.

In the US, there are no ethical guardrails for AI. There is no federal regulation of how AI is used.  We are already seeing the impact of bias in algorithms used to train AI models that are being used in decision making processes (an easy reference would be the federal purge of services led by the DOGE team).  Workday is being sued for age discrimination biases in its AI-based Applicant Tracking System (ATS) used by hundreds of companies.

Regarding Mental Health and AI, there was an interview with an AI developer who was talking about the difficulty of getting AI to be signed off as a therapist, but how they could immediately release software as a "friend".  Same software, but marketed as a virtual friend. Funding models for AI can be subscription based, time-in-application based, or based on data collection.  The last two are more prevalent, and should be incredibly concerning.

We are in a conundrum here.  Pandora's box is open, and will not go back quietly.

I really appreciate the specific examples of wayward AI Greg provides here. It's sobering. And I can't help thinking that a pro-AI person's response to them would probably be something along the lines of, "Well, you must be doing it wrong." That's just my gut instinct. I haven't heard anyone say it. But, it's often tech people's first response when ordinary users say that tech is screwing up.

Annette Bourbonniere, who shared several insights as well:

Loved how thoughtful your comments are about AI.  AI has been part of our lives for a long time.  Speech recognition software, Grammarly, Zotero and other citation managers, and even summaries made by Zoom.  All of these are helpful to writers, especially writers and businesspeople with disabilities, as well as to researchers and college students.

That said, you are wise to use caution.  AI in employment or healthcare decisions is downright scary.  When used during the recruitment of employees, it almost always filters out the potential of disabled employees. In healthcare, it can surreptitiously employ QALY’s (Quality-Adjusted Life Years) and DALY’s (Disability-Adjusted Life Years) to deny or delay care to those of us with disabilities.  Both of those dangerous metrics are banned in many countries, but the US uses them far more than we know.

The talent and originality of writers, artists, and others should not be replaced by AI.  We need the humanity and diversity that they all bring to us.  ChatGPT has also replaced companions for vulnerable teens, leading to terrifying consequences.  

Even the accuracy of AI can and should be questioned.  Even Grammarly will err when it doesn’t know the content.

We are able to use our own research capabilities online to look at ethical concerns.  Like you, I have chosen to use Claude because of those ethical concerns.  

I write and edit my own content and record in my own voice.  But for some business purposes, Claude has more information than I have.  For example, when looking at business grants, Claude can provide potential funders that fit my business needs.  It can also use content I have already provided to optimize the match between my skills and needs and what the grantors are looking for.  Claude can reply to requests by asking me to further check a source to ensure I interpreted it correctly.  I’ve only used it for a short time and in a limited way, but I’m far more comfortable with Claude than with ChatGPT, Co-Pilot, and Gemini.

On the other hand ... These positive experiences don't entirely answer the main objections, nor do they claim to. These things can be useful, particularly for disabled people. But we have to stick with being careful users and avoid becoming enthusiasts.

And lastly, these thoughts from Jen Rohrig:

I still refuse to use AI at all as I don’t believe it’s at all reliable. I will never use it at work where accuracy is important. It’s disturbing that it’s being forced into so many workplaces (especially places like health care as a method for note taking where errors can be deadly). There are also many news stories about layoffs related to AI being used in place of actual workers. I also feel like anyone who does use it needs to be really strict about fact checking – which just makes me wonder why use it at all.

I do think AI could potentially be useful I some cases but not with the way it’s been implemented – with the push to get it in anything and everything. Apparently new hearing aids would also have AI and from what I understand it’s mostly in the use of filtering out background noise. While it may be useful and actually do the job correctly I’m not sure I have enough faith in it to trust it. What if I were to get new hearing aids and they kept filtering out important sounds? Like a fire alarm?

While it may help people with disabilities do things I think there has to be an awareness of what it can and cannot do – and that no matter what you do you have to check the output it provides every time. Using myself as an example (to not to speak for anyone else and their use of it): if I were to use it to transcribe audio I may or may not be able to accurately check it for errors if I can’t hear the same dialogue.  How would I know for sure it was accurate unless someone else checked it for me? But these transcripts are already in Zoom calls and other meeting software. I will always have more faith in a human who is able to review and fact check everything.

Granted we do have issues of people not wanting to do the work, but that is more about the pay they receive for doing it. Instead of being encouraged to provide better pay and workplace conditions companies are replacing humans with AI in an attempt to cut costs. What happens when everyone is out of a job? This doesn’t even get into the environmental impacts of AI use – which has already been reported.

Have you read “Disabling Intelligences: Legacies of Eugenics and How We are Wrong about AI” by Rua M. Williams? I believe that book does a good job of outlining all the issues with the use of AI and the ways it plays into the issues of eugenics.

Jen brings up another interesting point ... about the risk of poorly-designed or misapplied adaptive tools being promoted as cheap, second-rate alternatives to more expensive but reliable human help. It's a dilemma a lot of disabled people face, and not just with AI.

Huge thanks to everyone who read and responded to the first proper Monthly Essay. I'm working on the next one right now, to be posted on Wednesday, June 10, 2026.

May 15 - Friday Video Share: "Was This Random Lady Rude To Shane? Opinions Needed!"

Mark Johnson says of Shane and Hannah:

"They're a cute couple who've created a HUGH platform, please reach out to them."

I agree. And I intend to!

Thank you everyone for reading and sharing your thoughts and reactions!

Disability Thinking Weekday is a Monday-Friday newsletter with links and commentary on disability-related articles and other content. You can help promote the newsletter by forwarding it by email or posting on your social media. You can also comment by sending me an email at: apulrang@icloud.com. Collected comments are shared around the first of each month. A free subscription sends a newsletter to your email each weekday. Benefits of paid subscription include:

• A monthly recap with links to all of the previous month's shared articles, organized by topic.
• Listing as a supporter, and a link to your website if you have one.
• You can recommend one disability-related article for me to share per month in a weekday post.

In June and July, annual paid memberships are discounted at $40. It's a great time to support Disability Thinking Weekday. Just click the "Subscribe or Upgrade" button below to get started.

I am so grateful for your help and engagement, in whichever forms you choose!

Free subscribers: 1,281
Paid subscribers: 47