Cognitive Sovereignty

paul@paulwalk.net (Paul Walk) — Thu, 02 Apr 2026 00:00:00 +0000

By the third stage, the syndrome is fully manifested. The person has abandoned their own cognitive sovereignty: the capacity for independent judgment and the ability to develop and defend ideas without outsourcing core thinking. They can no longer distinguish between what they think and what the machine has generated for them.[...]They have reorganized their mental life around the capabilities of the machine.

Marvin Starominski-Uehara: https://er.educause.edu/articles/2026/3/genai-and-creeping-cognitive-displacement-syndrome

This is the first time I have come across the phrase "cognitive sovereignty". The idea that people may be suffering from a self-inflicted Creeping Cognitive Displacement Syndrome is a worrying one. The evidence provided here is anecdotal - but it is offered by a practicing teacher who describes what he has observed in students:

The most disturbing cases of this cognitive hollowing that I have encountered involve students who are aware of their dependence on AI tools yet feel unable to stop relying on them. They describe the experience of trying to write without AI assistance as physically uncomfortable, cognitively exhausting, and emotionally intolerable. One student shared that they feel "stupid" when they try to draft text themselves, as though their own thinking is inadequate compared to the fluency of the machine.

In my own working life, I have several times heard professional acquaintances say something like "I only use AI to create a quick draft, so I don't have to start from scratch. Then I take it from there, rewriting and editing. It's my own work.". When I hear this, it sounds a little self-delusional to me. It's a pretence that the process of editing is where the real "value" is added. Editing is important - sure - but if it were primary, then we would recognise and celebrate editors more than authors.

For what it's worth, I'm actively trying to protect my cognitive sovereignty, by practicing Generative AI Vegetarianism.

Experimenting with `human.json`

paul@paulwalk.net (Paul Walk) — Mon, 23 Mar 2026 00:00:00 +0000

Inspired by Alan Levine (AKA Cogdog) I added a human.json to this website:

{  
  "version": "0.1.1",  
  "url": "https://www.paulwalk.net",  
  "vouches": [  
    {  
      "url": "https://cogdogblog.com/",  
      "vouched-at": "2026-03-23"  
    }  
  ]  
}

This is a machine-readable file designed to indicate to the world that the content here (all of it!) is written by a human - me - and not generated by a so-called "AI" system.

Beto Dealmeida, the creator of this new protocol, says:

One of the problems with the internet today is that a lot of the content is AI generated. There's no way to know for sure if a site is maintained by a real human or if it's just slop. The only way to know for sure is by getting to know the authors, which usually takes time and requires developing a relationship with them through other channels, like email or social media. But what if we could expand that trust by building a web of vouches between sites?

Beto Dealmeida: The human.json Protocol

This idea feels very "web-like" to me - in the sense that it works as the web was designed to work in the beginning. Indeed, it has something in common with older initiatives such as as blog-rolls and web-rings.

Since I learned about this via Alan's blog, it seemed only polite to make his blog the first that I "vouch" for in my human.json file.

This protocol is very new, and the author is inviting feedback. In fact, I already found a small issue with the protocol: it requires me to assert the canonical identifier (URL) for my website in the form of a prefix - allowing anything with that prefix to be covered by this assertion. So, the prefix I would like to use for this site is https://paulwalk.net, which is what I would consider to be the canonical URL for this website. However, the URL https://www.paulwalk.net is the one most likely to be used by others, and my website generator software (Hugo) is configured to use this - so this is the one I have used in my human.json file.

I notes that someone else had raised this issue, and that the next version of the protocol will support adding multiple canonical URLs to solve this problem.

Ai on Paul Walk's Website

Cognitive Sovereignty

Experimenting with `human.json`