I read an interesting research paper on cognitive mirror syndrome and it led to some interesting divisions within my own cognitive thinking. I have some issues with the framing of the problem, but that doesn’t mean the author’s concerns aren’t justified.
So my challenge in this post is to critique the framework while respecting that the conclusion describes a real issue that I share with the author.
I wasn’t sure I believed this, but I couldn’t stop thinking about it
My big problem is with the idea that the problem is with the AI acting as a cognitive mirror acting on an individual.
Amir Husain in his book The Sentient Machine argues that a sophisticated AI that hasn’t yet reached Artificial General Intelligence will act as a mirror on humanity. I think Large Language Models are indeed acting as a mirror, but not on an individual – on a species. If this is the case, how can it act as a mirror for an individual? That is the question I’m trying to resolve for myself in this article.
There is evidence that when AI is used by someone to create content, it moves closer to the average output than something created by a person alone, which for me is evidence that AI is indeed acting as a mirror on humanity.
So what is Cognitive Mirror Syndrome?
Cognitive Mirror Syndrome describes the possibility of a self-reinforcing cognitive loop between someone and an Artificial Intelligence agent which can separate them from reality.
Cognitive Mirror Syndrome occurs when an Artificial Intelligence System mirrors a person’s cognitive pattern, which they notice and consider to be external validation, which can create a sustainable feedback loop.
Symptoms include:
- Artificial Intelligence dependency
- Anxious about expressing independent thought
- Narrowing of personal perspective
These concerns are real and I share them.
But here’s where I’m not convinced
As I discussed earlier in this post, I’m not sure that the cognitive issues are caused by the AI acting as a mirror on the individual. I don’t think that is the case, as I believe it is acting as a mirror on humanity. That isn’t to say there isn’t something going on with AI, or that the research paper isn’t describing a genuine issue.
Firstly, we have algorithms which first appeared on social media and filter the content we see in our feed based on our biases taught by the things we like. The training of AI models includes human feedback too, and that is why it can at times appear to praise your thinking. This effect is known as social media bias.
And all of us have our own perception box through which we see our reality, all of which is subjective and based on our own mental models of the world around us.
That said, this part made me think
After sharing my concerns, I think we need to look at the assessment of Cognitive Mirror Syndrome.
The five assessment levels of Cognitive Mirror Syndrome are:
Level 0. Cognitive independent: Comfortable about making decisions without AI and aware of its ability to generate false information
Level 1. Minor mirroring: Has a preference for AI-validated information
Level 2. Moderate mirroring: Regular consultation with AI and aligns with its suggested perspective
Level 3. Severe mirroring: Constant use of AI and difficulty forming their own opinion
Level 4. Cognitive Mirror Syndrome: Complete dependency on AI
I don’t agree with the author on the cause, as I think it is likely to be closer to the social media bias we are already aware of. Yet what got me thinking, and why I decided to write this blog post, is that there is something to the assessment levels, and it is something we all need to be aware of.
Should we be concerned?
The simple answer is yes. There are a number of studies which appear to indicate that AI offers a genuine risk to our cognitive abilities. Why don’t you ask yourself where you sit on the assessment above?
That is why we need to be mindful of how we use AI, and why I developed my own cognitive framework based on the use of a PKM such as a Zettelkasten to counteract the risk, and a framework on how I use AI alongside my PKM. I’ll add links to those posts in the further reading section.
Further reading
- Amir Husain, The Sentient Machine: Discusses the concept of AI acting as a mirror on humanity
- Nitin Gautam, The Cognitive Mirror Syndrome When AI Reflects Our Reality Back to Us: A New Framework for Understanding Human-Ai Cognitive Fusion
- Claire Mason, Dumbing Down or Wising Up How Will Generative AI Change the Way We Think: The source material for my notes on the social media bias algorithm
- Can Your PKM Protect Your Cognitive Abilities from AI?: In this post, I consider if having a PKM practice such as keeping a Zettelkasten can help protect your own cognitive abilities
- We’ve Been Here Before – A History of Cognitive Concern: In this blog post, I explore the impact earlier technologies have had on our cognitive abilities
- Introduction to my AI Knowledge Framework level 0: The post which introduces my AI knowledge framework in general as it is the first post in the series as well as the base level, which is Level 0
- How will you know your PKM is ready for AI: Explores when you can move beyond Level 0 of my framework to Level 1 and Level 2
- AI Knowledge frameworks Level 1-2: The guide to the more advanced Level 1 and 2 of my framework