Dragon Naturally Speaking and broken wrists
a no-handed view of the further enshitification of software
I heard a woman on Ted X this summer say that if any one of us lives to be old enough we will be disabled. Facing and dealing with your own disabilities therefore is a factor of age, and happenstance, and birth. In October 2025, I fell about 4 m (12 feet) from the top of a building I’m building. I landed on my feet and dropped into a proper tuck and roll. Regrettably at 61 a little fatter than I was when younger, I was a little late. I shattered both wrists and cracked my head nicely on the concrete. The result is I am an author who has wrists that are immobile for as much as five months. By the end of this adventure, I will have gone through for medical procedures and tons of rehab to get back to being a touch typist. Until then, I am a disabled American.
While meant as a joke, I say I can’t even pick my own nose. But let that tiny frustration stand-in for all things I cannot do during morning and evening ablutions. I can’t drive. I can’t fasten my own seatbelt. I can’t open most doors. And yet, I still sometimes get the stink eye from people watching us park in handicap spots.
This brings me to the topic of this post: dictation software. This is the third time that I have used Dragon natural speaking during the last 30 years. I’ve had other injuries, and I have striven to avoid carpal tunnel syndrome. The efforts never last very long because my fingers type as fast as my brain writes. Both of those activities happen faster than I can dictate while also saying aloud the various punctuation and commands statements that format a manuscript such as this.
Dictation software, like a cane, a crutch, glasses, hearing aids, mobility scooter, prosthetic limbs, or a wheelchair is simply an augmentation that allows us humans to take care of ourselves and the world around us. It appears that modern dictation software has failed to recognize its role in supporting people with disabilities.
Dictation software is distinct from voice-to-text utilities found in so many apps and devices. Dictation software is intended to produce professional level, well-formatted text with appropriate grammar. The largest commercial markets for dictation software is in the legal and medical space where notes are often transcribed. That said, folks with limited agility in their hands may be the third strongest market. The point is the software needs to be “ADA” compliant. ADA is the Americans With Disabilities Act. While ADA does not govern software, there are published national and international standards called WCAG that bridges this narrow gap. WCAG provides specifications for font size, color contrasts, and other accessibility issues.
What struck me when installing Dragon natural speaking again, now for the third time, I was frustrated by the user interface. There is a floating doobly-do as the primary user interface. When you start dictating, the interface disappears.
Folks with two good working hands would click outside of their word processor to get the doobly-do visible again. For those of us with no working hands, were awkwardly working hands controlling the dictation software required more dexterity than I would expect (or could do). The knobs and twiddily bits for Dragon natural speaking do allow for some modification of the software it seems like they removed features, detuned the software, and completely forgot their space as a tool that supports disabled folks trying to work out a modern desk.
The first time I installed Dragon speaking a few decades ago, I had to spend hours training the software on my speech patterns, word choices, & spelling preferences. I suppose a lot of that was the software training me to speak more clearly, unnaturally slow, and say the words for the various punctuation as I “write”.
One might think without spending hours of training the software and yourself, the software would be better, simpler, & faster. It is not. As long as I speak American English with no discernible accent, and distinct boundaries to each word spoken, Dragon understands most of what I’m saying. The time training made the experience better in the old days. my own dictation voice is more stilted than the worst text to voice utility I use on my mobile phone: flat intonation, no cadence, robotic.
Dictation software is now competing, or incorporating, generative AI features into the tool set. As an author with two novels out, 2 in the publishing pipeline, and more to write, I feel the impinging threat of generative AI. I wish not to use AI because all generative AI is theft. All generative AI is theft because in order to predict the next word in a phrase the AI engines perform better if they have ingested all written works of all time (substitute “images” or “photographs” as needed for visual generative AI).
Most most modern speech to text tools promote their use generative AI as a benefit. This makes the assumption that the author dictating a novel of her own creation needs generative AI to fix words, finish a sentence, and punctuate a manuscript.
The day is here where I am not given the choice to train dictation software, because possibly the investment is made on having AI do the hard work instead of the author. I can easily picture the day when a medical transcriptionist is typing the literal words of the medical practitioner into a report well AI is “helping” with the selection of words and phrases. Or a legal practitioner reading the transcription of a deposition and seeing words in print that were never said at the deposition table. This seems like a space where AI should not have a role. Medical notes depositions and court transcripts only have legal value if they accurately represent the words spoken.
It seems intuitive that speech-to-text and dictation software should be the same, but this author would argue it is not. I wish to state at all times that my writing is free of generative AI. I wish that to be a guarantee, and yet as dictation software and word processors incorporate more AI-like features this promise may be hard to hold because I don’t know (I may not know).
I am a snob with mild puritanical hints in my belief structure. By this I mean, I believe in work. You do your own work and reap your own rewards. All generative AI requires large language models absorbing the largest possible data set. With writing, this means all public, private, and copyrighted works. Let me restate that: I am a temporarily disabled snob who writes novels and short stories. While I miss the use of my hands, I should be able to write by dictating using modern tools. It was easier a few years ago, before generative AI became commonplace.
The mental draft of this post was funny and recounted recent weirdness while using dictation software. I guess that’s my next post, huh?