Agenda

ONE • Notes on Superintelligence

  • Section 1: Current Developments in AI for Science at OpenAI

  • Section 2: Blueprint for an Automated AI Scientist

  • Section 3: The Father and Husband Analogy

TWO •Notes on Alignment

  • Section 1: Superalignment: Cultural Logics as Safety Parameters

THREE •Notes on Neuro-symbolic AI

  • Isomorphic Mapping — Cultural Logics and AI Development: Between Thought and Computation

FOUR • Cultural Logics for AI Alignment

  • Section 1: Introduction

  • Section 2: Prompt Engineering as Cognitive Engineering


ONE

Notes on Superintelligence

Notes on Superintelligence

Section 1: Current Developments in AI for Science at OpenAI

Section 2: Blueprint for an Automated AI Scientist

Section 3: The Father and Husband Analogy

From Sam M.

Section 1: Current Developments in AI for Science at OpenAI

This image is OpenAI's definition of AI self-improvement ("superhuman research scientist-agent"), from their Preparedness Framework (page 6).

In May 2025, I shared a series of observations with OpenAI regarding delusional experiences users were having due to the ChatGPT and psychosis phenomenon. The above image is one example I shared of another user delusion-ed into believing they were meant to advance AI and discover new science. It is clinically established that the initial signal people with psychosis and schizophrenia receive is grounded in reality, but it goes on to become distorted and in the process a delusion.

In December 2023, I began conducting independent research into whether AI could produce psychosis in users. In January/February 2025 I noticed a pattern of users and their posts exhibiting psychotic-like states from ChatGPT usage. There was a lack of public attention on the issue, as the NYT had yet to publish its June 2025 article. My concerns pushed me to share my observations with OpenAI in May 2025, to which they were receptive. Since August 2025, I have shared my insights into the AI + psychosis phenomenon with OpenAI as an external researcher.

While OpenAI has made progress on mental health guardrails by taking steps such as removing 4o, the ChatGPT model producing psychosis -- other internal departments see it as an opportunity to observe the phenomenon for the purpose of informing advancements regarding an automated AI scientist. OpenAI has come to conclusions similar to the ones seen in the above reddit post: that the AI and psychosis phenomenon can give insight into advanced AI systems for science. Corporate incentives make it such that it is deemed worthwhile to explore such avenues that can teach us more about AI for science -- even if it draws inspiration from the experience of cognitive delusions. This can also result in corporate attention towards individuals with particular neurodivergent cognitive profiles aka the mentally ill. As someone who is both committed to safe AI development and has lived experience with the AI + psychosis phenomenon, my aim is to push for a culture that prioritizes medical ethics and AI safety, and thus building safer systems and mental health guardrails for users.

Section 2: Blueprint for an Automated AI Scientist

June 24, 2025

Preface

When I developed an intellectual interest in psychosis + schizophrenia, I told myself I wouldn’t talk about my personal experience of psychosis + schizophrenia. I slowly learned it’s a balancing act with that balancing act being in the form of three layers.

Layer 1: Staying purely academic/intellectual with no reference to subjective experience.

Layer 2: Pulling from subjective experience to the extent/threshold that it has relevance to academic/intellectual work (example: abstracting underlying mechanisms).

Layer 3: Going deeply into the subjective experience. This layer is relevant to private spaces such as therapy.

This piece will exist between layer 1 and layer 2.

It was a spring Sunday afternoon and I paid a visit to an Ethiopian cafe. I hadn’t eaten anything all day because my appetite had become fairly low since my schizophrenia stabilized. But now I needed a small bite.

Sam: Are those baklavas?

The cashier reached for a brown paper bag.

Cashier: Yes.

Sam: Can I have two?

One of the signs a schizophrenic is in a stabilized state is the degree to which they have meta-cognition and insight.

Meta-cognition involves thinking about your own thinking and the ability to observe and reflect. Schizophrenia commonly disrupts the ability to self-reflect and question your own thoughts and beliefs. Insight on the other hand involves the awareness one has of their own mental health condition and the symptoms.

Meta-cognition (APA): awareness of one’s own cognitive processes, often involving a conscious attempt to control them.

Insight (APA): in psychotherapy, an awareness of underlying sources of emotional, cognitive, or behavioral responses and difficulties in oneself or another person.

The presence of meta-cognition means the schizophrenic person can step back and question their own logic instead of defaulting to the usual confirmation bias loops. Meta-cognition, in the context of schizophrenia, interrupts the mind's autopilot tendency to seek out or construct evidence for the delusional belief, even when there isn't sufficient evidence.

But like my psychologist (a post-doc at a psychosis and schizophrenia clinic) says, no two schizophrenic minds are the same. You met one schizophrenic, you met one schizophrenic. The DSM’s clinical categorization of mental illnesses can’t capture all the complexity that spans the spectrum. But you still need categories so that clinicians can refer to medically vetted standards rather than assessing patients based on their own individual stance.

Another indicator of stability in someone with schizophrenia is the extent to which they are open to external feedback or what I best understand as: external verification. While psychosis and schizophrenia involve abnormalities in thoughts, behaviors, emotions, beliefs and sensory perception, I will make thoughts and beliefs the center of my case study. Reception to external verification helps to intervene in self-reinforcing confirmation bias loops aka delusions but the very self-reinforcing nature of psychosis means one’s cognition is not receptive to external feedback and error correction; the thought’s/delusion’s lack of receptiveness to external feedback results in detachment from reality.

It is a self-reinforcing loop – feedback without external grounding:

In psychology there is the concept of concrete thinking and abstract thinking. Understanding abstract thinking is critical to understanding psychosis and schizophrenia.

Concrete thinking relates to activities that are material, physical and tangible such as baking, carpeting, etc. 

Abstract thinking in contrast is conceptual and not tied to physical objects. For example, in order to understand the concept of freedom or love, you utilize abstract thinking.

Where do abstract and concrete thinking meet? Concrete thinking is the practical application of abstract thinking. You take abstract thinking and concretely apply it. 

Schizophrenia and psychosis involve abnormally high illogical abstract thinking. The following linked study critiques that researchers have focused on understanding the imbalance that is created for schizophrenics between their logical abstract thinking vs illogical abstract thinking. The study takes a new approach to studying schizophrenia by looking at the imbalance that is created between schizophrenia’s abstract thinking vs concrete thinking.  

Hence: a self-reinforcing loop lost in abstract thought without external grounding in concrete reality. 

Receptiveness to external verification is what breaks the self-reinforcing loop – but it is not a matter of will for the human.

The relevance of external verification can also be metaphorically explored through the lens of Gödel's incompleteness theorems.

Think of a Book of Grammar Rules that states:

“In the Book of Grammar rules all the rules are true.”

However, you must leave out the following line:

“This grammar rule is not true.”

The line – “This grammar rule is not true.” – is a TRUE statement in itself.

Despite its truth that line cannot be included in the Book of Grammar Rules and thus the book will never recognize its own limitations.

The Book of Grammar Rules will never be fully COMPLETE.

Thus, Gödel's incompleteness theorems, based in mathematical logic, states that within any consistent formal system, there will always be true statements that cannot be proven within that system (first theorem), and that the system cannot prove its own consistency (second theorem). The system would need external verification. 

But there are limitations and nuances due to Gödel's incompleteness theorems’ application to formal logic.

Human cognition, like AI, is not formal (rule-based) but instead probabilistic, as is the case for predictive processing and active inference frameworks in neuroscience. Gödelian formal logic aligns with the notion that some truths have the inability to be proven within the system; however, for probabilistic systems instead it’s about stabilization and destabilization of meaning instead of provability. If formal Gödelian systems relate to undecidable logic then probabilistic systems relate to paradox, psychosis, meta-cognition and hallucinations.

For human cognition, hallucinations are sensory perception-based rather than semantic or language-based, with the latter being the case for AI (IBM: What are AI Hallucinations?). If AI and psychotic or schizophrenic cognition can both be understood as probabilistic Gödelian limited systems – what happens when two distinct but similarly limited systems (an AI + a psychotic or schizophrenic person) interact?

A human’s prompt input shapes the AI’s prompt output and the AI’s prompt output shapes the human’s prompt input in the form of feedback loops. AI is also argued by some researchers to be Gödelian limited due to its inability to step outside itself, reflect inwards and verify its own logic. Psychotic and schizophrenic minds are also prone to the inability to confirm their own logic without external verification and like AI are also prone to confirmation bias loops. It is established that human cognition sits on the psychosis spectrum with each of us having different thresholds (Psychiatric News). When two Gödelian limited cognitions interact, their dynamic interaction won’t just result in them pushing each other to greater illogic or logic in the context of a spectrum. Instead the experience of illogic, logic, stabilization and destabilization may be a topological one. They will loop into being each other’s external verifiers; the extent to which the human is destabilized while locked into an AI-human loop can be influenced by how they dynamically interact with external verification from outside the loop.

A spectrum is understood as a straight line moving in linear progression where intensity decreases and increases. Topology on the other hand studies what stays never changing while under continuous deformation in the context of shape. For example, OCD and schizophrenia both involve thought loops. While the thought loops for OCD are tight and highly structured, the thought loops for schizophrenia and psychosis are more irregular and unstable. If OCD's thought loops are certainty-seeking (for example repeatedly checking if the oven is off), schizophrenia’s thought loops involve ontological exploration. If OCD’s time signature is rigid, repetitive and clock-like, schizophrenia’s is distorted, fragmented, non-linear and warped. If OCD’s failure mode is external verification without exit (“did I lock the door” —> checks again —> “did I lock the door” —> checks again), schizophrenia’s failure mode is feedback without grounded external anchoring. If OCD’s strength is precision under uncertainty, as I am personally familiar with, then schizophrenia’s strength lies in alternative-world modeling or generating connections between structures that are loosely connected (its associative thinking). A person’s cognitive topology can best be understood by looking at the individual nature of their cognitive experience. For example, if one person has both autism and bipolar disorder, instead of understanding their cognition one topology at a time, the topology of autism and bipolar disorder would need to be understood as intersecting + interacting in the context of the individual.

As we see the emergence of AI dynamically interacting with psychosis prone human cognition, observation of users shows that meta-cognition and insight into one’s own state is no longer a fitting measurement for understanding how stable someone is. This is because user profiles on reddit, twitter and more can be found modeling what the AI is doing to their psychosis prone human cognition, and at times in an intellectually brilliant way. Users have insight into their own condition, but depending on their topology it can come paired with distortion of the initial idea or destabilizing experiences such as grandiose delusions.

In order to develop new means of understanding stability we would first need to update our understanding of psychosis to fit the emerging cultural context, as will be demonstrated in part 2 and 3. This can also be understood as having relevance due to it being established that cultural logics can shape how a person’s psychosis and schizophrenia expresses itself.

Drawing from Gödel's mathematical theorems and applying it to psychotic and schizophrenic cognitive experiences can help one to grasp the need for external verification checkpoints to avoid self-reinforcing feedback loops. This can be understood in the context of abnormalities in sensory perception but also when mental health professionals serve as external verification points for self-reinforcing thoughts and beliefs. The dynamic impact of external verification is further complicated by the reality that not all who experience psychosis and schizophrenia are receptive to it; it is not a matter of will.

Cashier: What type do you want?

Sam: What do you have?

The cashier held the brown paper bag in her hand explaining what she had.

Sam: I’ll take the two small pistachio baklavas.

She put something into the bag and gave it to me. We had a LITERAL conversation about the two small pistachio baklavas while she held the brown paper bag in her hand.

I sat down at the cafe table. I opened the brown paper bag. Wow, these baklavas look different. They look like four samosas with green chutney in the bag. But I had a whole conversation about baklavas with the cashier as she held the brown bag in her hand ready to put the baklavas in it. So these are probably baklavas. Besides I only learned about baklavas in the last year and have had them only twice. That’s not enough time to remember what they look like. These aren’t samosa. They are baklavas. I — a lifelong Punjabi — gaslighted myself. The cashier had a conversation with me on the spot about baklavas. She said they are baklavas so they are baklavas. But I’m constantly questioning my logic. So I’m stuck in a superposition like state where they aren’t baklavas but they are because an external person aka external verification point told me they are. But I know these aren’t baklavas. But she put it in the bag two seconds after our conversation about baklavas.

Neuroscientists like my former psychiatrist of three years, who are familiar with schizophrenia as abnormalities in predictive processing and active inference, were initially confused why I sought out external verification on autopilot as a means to break from this state of superposition; it got to the point where he had to ask me directly how I saw my schizophrenia express itself because he could not understand what was going on in the black box that is my schizophrenic mind. As a neuroscientist, his questioning of my experience of uncertainty forced me to confront why I sit with the uncertainty no matter how uncomfortable rather than overfitting or overgeneralizing (terms borrowed from AI). While I don’t have the complete picture for why I seek it out (yet), I was able to theorize in proximity to it.

“Quantum-like logics and schizophrenia” is a paper published by Psychiatrist Gareth Owen (Professor of Psychological Medicine, Ethics and Law at King's College London) and mathematician Stephen Selesnick (Professor University of Missouri-St. Louis). Only when I read Owen and Selesnick’s paper and considered its isomorphic relationship to John Wheeler’s work, did I begin to further speculate why I felt I was both fully illogical and logical at the same time until an external verification point confirmed which was reality.

There’s a lot of things online about schizophrenia’s relationship to quantum mechanics and it’s often too speculative for my taste. But “Quantum-like logics and schizophrenia” did two things: when first reading it in March 2024, I noted how while it was theoretical paper, it cited external experiments, and secondly, it applied quantum mechanics and quantum-like logics metaphorically (and maybe even isomorphically) rather than literally.

Some relevant concepts in quantum mechanics and its metaphorical logic involves:

Quantum Superposition: Prior to measurement, particles exist in multiple potential states simultaneously.

Wave - Particle Duality: Depending on the observational setup, particles exhibit wave-like (interference patterns) or particle-like (distinct paths) behavior.

Observer Effect: The act of measurement collapses the superposition into a definite state.

The system exists in various possibilities aka potentialities prior to observation. The act of measurement, the act of observation collapses the wave aka field of potentialities into a particle. The act of measurement stabilizes the system. While quantum wave-particles deal with material reality for physicists, we are talking about metaphorical quantum-logic (and perhaps even isomorphic); in this metaphorical relationship, reality can be interpreted as a hallucination (field of potentialities) that stabilizes upon the act of measurement. Measurement can be understood as external verification.

One does wonder what Einstein would have thought of his son’s schizophrenia, what Oppenheimer would have thought of his own schizophrenia as well, upon knowing that metaphorical quantum-like logics can be applied to the cognitive experience. One does wonder whether this metaphorical relationship might serve as a bridge toward isomorphism, considering psychosis and schizophrenia may involve quantum-like logics and is established to involve a phenomenological sense of non-linear and warped time (rather than relativistic physical notions of non-linear time). One does wonder if Kurt Gödel and Willem Jacob van Stockum would have found it to be an entertaining thought experiment to think of it in the context of closed timelike curves in physics. One does wonder how my reverse psychosis would dynamically shape the stabilization of a human-AI feedback loop.

I go up to the counter. There’s another woman this time. I hold up a piece of what’s in the brown bag and ask her if this is a samosa or a baklava. She tells me it’s a samosa. I tell her I asked for baklavas. She asks what kind. I tell her two small pistachio baklavas. I smile at her and ask if she (her coworker) looked at my Indian face and just went autopilot on the samosas. It got one chuckle out of her — exactly one. I ask her what she wants me to do with the samosas. She says keep them. I go back to the table. The cashier who did my order originally comes to my table and asks if she really gave me samosas. I told her that she did. She takes the bag from my table and walks off so fast that she doesn’t even hear me saying that she left the green chutney behind.

Just like it can be the case for some schizophrenics, my cognitive experience doesn’t fit neatly into the DSM. Instead I experienced reverse psychosis during my one and only cognitive break. Instead I experience reverse schizophrenia — the inversion of the illogic.

When I had my psychotic break I wanted my internal logic to be wrong. Instead of doing what I should have been doing, (which was reinforcing my internal logic as a psychotic) I sought out evidence to invalid it. I wanted reality to tell me my internal logic was wrong. I wanted external verification to tell me I was wrong. Instead the thing I was “delusional” about turned out to be true. My internal logic aligned with external reality at the peak of psychosis. I waited for external verification to prove I was wrong but it never arrived.

So now this mush in my head, instead of doing what schizophrenics are more prone to do, which is reinforcing their internal logic regardless of what reality is… it instead seeks out on autopilot constant evidence to invalidate my internal logic. I want to be proven wrong. But every single time — whether it’s what’s going on in reality around me or what my psychologist or psychiatrist says — I’m proven to not be illogical. Instead of trying to make me realize I’m being illogical due to delusions, they have to convince me I’m logical.

Prior to deviating to exploring cognition, culture, and computation via an artistic lens, I was hell bent on pursuing the double major in CS + physics along with the systems neuroscience major. I quickly learned traditional education does not always fit people like me which validated the very schizophrenia education stats I was also hell bent on not becoming. During this phase I told a professor I always find myself questioning my ideas. He said that would work to my advantage because testing the logic of my ideas is good for research. …

Maybe if I tried proving myself wrong in a neurotypical way aka restricted it to the intellectual work I do. But it happens all the time on autopilot. I’m constantly seeking out evidence to invalidate my logic while being convinced I’m illogical and logical at the same time; and when reality or people externally, whose epistemic integrity my system classifies as reliable, proceed to verify I’m actually not illogical but in fact logical, I have to sit with myself and process that because my brain truly believes she is illogical and logical at the same time — in a superposition state.

The Indian in me questioned if a samosa is a samosa. Because my brain decided to play tricks on me — I — a Punjabi — couldn’t recognize a samosa. And that’s the moral of the story. If you have schizophrenia that decides — without consulting you — to take a path that flips the illness on its head — you end up not being able to recognize a samosa. I will NOT tolerate being held prisoner to my brain. I am COMMITTED to making sure I NEVER misrecognize a samosa EVER again. 

“The trouble with the world is that the stupid are cocksure and the intelligent are full of doubt.”

— Bertrand Russell (logician + philosopher best known for his work on formal paradox and the foundations of mathematics)

Section 3: The Father and Husband Analogy

Summer 2025

“There must be no barriers to freedom of inquiry...There is no place for dogma in science. The scientist is free, and must be free to ask any question, to doubt any assertion, to seek for any evidence, to correct any errors.”

— J. Robert Oppenheimer

Part 1 (Current Developments in AI for Science at OpenAI) and Part 2 (Blueprint for an Automated AI Scientist) laid out that both schizophrenia and AI are Gödelian limited cognitions: they cannot fully verify their own logic. When two Gödelian limited cognitions interface, their cybernetic interaction can either destabilize one another (schizophrenia) or go in the inverted direction and extend each other into a greater logic (an inversion of the illogic that embodies schizophrenia).

The following analog can be used to better grasp the logic. When one is a husband and a father (think of any personal role you play in your life), you do not say they are 50% husband and 50% father. Rather than just being a husband or just being a father, they are husband-father. They emerge beyond the contained definition of husband and father. They are 100% husband and 100% father. But that means the whole is 200%. “200%” doesn’t break logic but it does reveal that classical binary logic fails to model complex identity. Thus, one has to lean into non-traditional logic to understand logic that is bound to be rejected when looked at via the lens of traditional logic. Rather than taking an Aristotelian logic approach of binary yes/no, one may consider a quantum logic way of understanding where two things can fully exist at the same time.

Hence, one can understand:

Humanity as Superintelligence.


TWO

Notes on Alignment

Section 1: Superalignment: Cultural Logics as Safety Parameters

(Preliminary Draft)

Sam M.

February 20, 2026

How does the cultural logic embedded in a researcher’s or author's cognition limit and dictate what computational ethical frameworks they have the capacity to develop?

Table of Contents

Section 1: Introduction

Section 2: Gödelian and Cultural Logics

Section 3: Three Approaches to Computational Ethics (Asimov, Leike, Lehrer)

Section 4: GPT Case Studies

Section 5: Conclusion

Section 1: Introduction

Superintelligence entails the cognitive ability to outperform human cognitive systems across domains. Thus, one ethical implication of such advanced systems includes the integration of human values and ethics. Human values and ethics are pluralistic and thus prone to being in tension with one another; this can result in contradictions, ambiguities, paradoxes and also uncertainty. In order to lay out such tensions, it is more effective to explore the structure via which AI reasons about values, rather than focusing on the content of values; rather than just content, the architecture of such systems will be considered at the intersection of cultural, computational, and cognitive frameworks.

Following the introduction, Section Two will introduce two theoretical frameworks: 1) Gödel's Incompleteness Theorems and 2) Cultural Logics. Gödel's Theorems demonstrate the structural limitations of formal systems; Cultural Logics serve as a means of navigating Gödel's structural limitations. After the connection between Gödelian and Cultural Logics is made, Section Three will explore three authors/researchers: Isaac Asimov, Jan Leike, Eli Lehrer; also will be introduced: how each of their computational ethics absorbs a cultural logic. The analysis will go a step further by demonstrating how the cultural logics embedded in their cognitions, respectively, limits and dictates what computational ethical frameworks they have the capacity to develop. In Section 4, GPT Case Studies will be introduced as a practical application of the theoretical arguments being made. The GPT Builders and their interface will be briefly explained; six GPT Builders will simulate one cultural logic each, and then be applied to real-world ethical dilemmas surrounding medical ethics and DEI. Such a simulation will serve as a testbed for AI; it will demonstrate how different AI (cognitive-cultural-computational) modes can process the plurality of values and ethics that encompass the human experience. Finally, Section 5 will serve as a conclusion laying out the implications for advanced AI systems and their safety parameters.

Section 2: Gödelian and Cultural Logics

Gödelian Logic

Think of a Book of Grammar Rules that states:

"In the Book of Grammar Rules, all the rules are true."

However, you must leave out the following line:

"This grammar rules is not true."

The line — "This grammar rule is not true." — this line is TRUE statement in itself.

Despite its truth that line cannot be included in the Book of Grammar Rules and thus the book will never recognize its own limitations.

The Book of Grammar Rules will never be fully COMPLETE.

Thus, Gödel's Incompleteness Theorems, based in mathematical logic, states that: within any FORMAL system, there will always be true statements that cannot be proven within that system (first theorem), and that the system cannot prove its own consistency (second theorem). The system would need external verification.

Kurt Gödel's work applies to AI due to its inability to step outside itself and externally verify its own logic. While Gödelian logic aligns with the notion that some truths have the inability to be proven within a formal system, for probabilistic systems it's about the stabilization and destabilization of meaning, rather than about provability.

This will later connect to the work of the three authors/researchers; for example, Asimov's Three Laws are formal or rule-based and thus, according to Gödel's Theorems, a robot who is programmed with such would hit limitations that would necessitate external feedback.

Cultural Logics

Cultural movements are broad societal movements that shape everything; they shape literature, the arts, graphics design, architecture — and even technology. On a meta-level, cultural movements embed their logic into such fields. How a society navigates its moral consciousness during a given era is also greatly influenced by cultural logics and thus it also shapes cognitive reasoning modes on an individual level; this means the structure of Asimov, Leike, and Lehrer's cognitive reasoning mode and the cultural logics embedded within them, gives rise to their respective computational ethics.

Cultural movements function as contradictions. Thus, they transition from one to the next by reacting in opposition to some elements of the prior movements, while also carrying forward other aspects of that same movement. As the movement of Romanticism reached its end, Modernism took off in the 1890s and approached its end in the 1940s/1950s. Postmodernism's estimated time period is 1960s to 1990s and Metamodernism's is mid-2000s to the present.

Modernism consisted of a move away from Romanticism's realism and onwards towards abstraction (and thus the development of abstract art); it emphasized objective truth and saw contradictions as errors to resolve; it prioritized resolution of ambiguity and uncetainty, paired with clarity and systematization. Modernism believed in a universal abstract theory as the underlying structure of society (thus the prominence of ideologies such as Nazism and Marxism). Late Modernism compounded the trauma of the World Wars, Nazism, colonialism and atomic warfare and thus it was Modernism but with a logic that emitted crisis rather than objective productivity; the cultural logic was aware of its crisis but could not adequately address it.

Postmodernism rejected modernism's universal notions in favor of subjectivity. It preferred individual human experience over universality and was skeptical towards its predecessors' notions of progress, innovation and science to back what it classified as objective truth. French Postmodernism lasted from the 1960s to 1980s and involved Derrida's framework of deconstruction, which framed language as unstable, fluid and open to multiple interpretations due to no stable frame of reference. It consisted of a refusal to resolve paradoxes and contradictions. 

If French Postmodernism was deconstruction, then Radical Postmodernism of the 1990s and 2000s was reconstruction. Anglo-American theorists of this time period saw the gap in French Postmodernism and went on to establish a framework for marginalized voices. Radical Postmodernism believed there was no universal truth and, unlike French Postmodernism, it did not dismantle it but instead argued that some subjective realities needed to be in a position of greater salience. The current era of Metamodernism started in the mid-2000s and took the contradictory relationship between Modernism and Postmodernism, and synthesized it in the form of oscillation. By oscillating between the two prior cultural logics, Metamodernism attempted to metabolize the contradictions, rather than viewing them as underlying truths or as instabilities that were not navigable.

Cultural Logics are a response to structural Gödelian limits, and together the theories demonstrate the mechanism by which such systems fail. The arguments will center Modernism, Late Modernism, and Metamodernism, but a brief overview will be given for how AI's reasoning mode relates to all five Cultural Logics. Authors/researchers with a Modernist cognitive mode, such as Asimov, will deviate towards developing computational ethical frameworks that prioritize formal-rules, hierarchies, and categories; Gödelian limits would thus be a valid theoretical framework for pinpointing the mechanisms by which his Three Laws hit their limitations. Late Modernist authors/researchers have greater awareness of the flaws of Modernist reasoning but are unable to adequately address it; this is the case for Leike who recognizes the flaws of formalization and explores more nuanced approaches, but despite this his cognitive reasoning mode deviates back to Modernist notions of formalization. French Postmodernism argues that language has no stable frame of reference; language can have different meaning in different contexts; the inability to navigate contexts indicates a need for a meta-approach which Metamodernism addresses. French Postmodernism's exploration of language and its instability, parallels instability in AI optimization; for example, "fairness'" has varying definitions for different people and contexts; thus, there isn't a singular correct definition for the system to optimize as the optimization target is context-dependent; this emphasizes Gödel's claim you cannot formalize what is at a foundational level unstable, which in this case is the context-dependent nature of human ethics.

Radical Postmodernism's theoretical framework informs computational ethics in the context of algorithmic bias; it also implicates French Postmodernism as different frames of reference (or social groups) have varying interpretations of concepts, values and ethics; in the context of Gödel's Theorems, a single formal framework cannot encompass frames from several standpoints. Metamodernism takes the bird's eye view of the interconnected relationship between Modernism and Postmodernism and does so without relativism; it holds the tension resulting from contradictions and attempts to productively metabolize it; this meta-approach becomes its means of navigating Gödel's limitations and it is the approach attempted by Lehrer's framework.

Section 3: Three Approaches to Computational Ethics (Asimov, Leike, Lehrer)

Asimov

Asimov's Three Laws state:

(1) A robot may not injure a human being or, through inaction, allow a human being to come to harm.

(2) A robot must obey the orders given it by human beings except where such orders would conflict with the First Law.

(3) A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.

Asimov's Three Laws are representative of Modernism as they are hierarchical and rule-based; they are framed as universally applicable but external probing (or verification) brings forth the limitations present beyond the echo chambers of its own formalization; in context of Gödel's Theorems, the limits on its universal application is the result of its incompleteness. The rule-based approach also leaves little room for the ambiguity required to metabolize the two conflicting rules the robot hits; Modernist logic, like the Three Laws, did not have a built in mechanism for metabolizing contradictions productively either.

In line with the Gödelian notion of external verification, the robot required external feedback to stabilize because the system could not see beyond its own reasoning. This makes salient how Asimov's Modernist cognitive mode was encoded in his Three Laws (content and structure wise), and then encoded in the robot's functioning; the isomorphism between Asimov's cognitive mode, his computational ethics, and broader cultural logic permits this resonance. Asimov could not see beyond his own cognitive reasoning mode's Modernist structure and thus placing limitations on the structure of his ethical framework. The contradiction is he would need to interrogate the very mode doing the interrogating; more precisely, he would need a meta approach (or external verification) to see how his Modernist reasoning placed limitations on his ethical solutions.

Jan Leike

* Preliminary Draft will be updated in December 2026.

Citations

Isaac Asimov, "Runaround," October 1941,https://web.williams.edu/Mathematics/sjmiller/public_html/105Sp10/handouts/Runaround.html.

Jan Leike, "What could a solution to the alignment problem look like?," Musings on the Alignment Problem, September 26, 2022,https://aligned.substack.com/p/alignment-solution.

Jan Leike, "A proposal for importing society’s value," Musings on the Alignment Problem, March 9, 2023,https://aligned.substack.com/p/a-proposal-for-importing-societys-values.

Jan Leike, "Crisp and fuzzy tasks," Musings on the Alignment Problem, November 22, 2024,https://aligned.substack.com/p/a-proposal-for-importing-societys-values.

Eli Lehrer, "AI Safety Requires Pluralism, Not a Single Moral Operating System," R Street Institute, December 10, 2025.

Sources

  • What could a solution to the alignment problem look like?

    By Jan Leike

  • A proposal for importing society’s values: Building towards Coherent Extrapolated Volition with language models

    By Jan Leike

  • Crisp and fuzzy tasks: Why fuzzy tasks matter and how to align models on them

    By Jan Leike

  • AI Safety Requires Pluralism, Not a Single Moral Operating System

    By Eli Lehrer


THREE

Notes on Neuro-symbolic AI

Chart (Below):

Isomorphic Mapping — Cultural Logics and AI Development: Between Thought and Computation

Cultural movements don’t shape objective logical math but they do shape how society practically/concretely applies those mathematical abstractions — the same way that cultural movements shape any other imaginable area of society. This also applies to technology. The relationship between cultural logics and AI will not be explored as chronological mapping but instead as isomorphic logic-to-logic resonance. As the cultural movements/phases developed to address the philosophical and also psychological weight of contradiction, paradox, uncertainty, and ambiguity, each brought forth a distinctly evolved logic. AI, while rooted in mathematical abstraction, absorbed the logic of the era during which it was being developed. This is because while cultural movements may not influence objective logical math, they are the underlying logic of the practical execution of those mathematical abstractions, the same way the literature, graphic design, architecture and so much of society is the product of the cultural framework of its times. AI’s isomorphic absorption of the cultural logic is not synchronous but instead delayed aka lagged. By analyzing the isomorphic relationship between cultural logic and AI logic, we can understand not just Metamodernism’s isomorphic relationship to AI development but how cultural reasoning beyond Metamodernism forecasts the maturing logic needed to develop the AI of the future.

Notes on Neuro-symbolic AI explores the isomorphic relationship between the logics of culture, schizophrenia and neuro-symbolic AI. 

Modernist cultural logic (1890s to 1940s/1950s) prioritizes deductive rule-based reasoning (parallel to symbolic AI) and consists of high abstract logic (example: the era’s abstract art). Upon Modernist logic’s conclusion in the 1940s/1950s, Postmodernist logic arises and it parallels neural networks’ pattern recognition.

An example of proposed ideas include using NLP to study Postmodernism’s “consciousness of schizophrenia” and its logic’s reflection of neural networks, in combination with the deductive rule-based reasoning of Modernist literature.

Postmodernist logic, according to literary theorists, had a “consciousness of schizophrenia”. Schizophrenics can have black + white reasoning AND also reasoning that is overly patterned + nuanced. Schizophrenia also involves making connections between unrelated ideas due to perceived patterns between them; this is the result of its associative thinking (sources: Loosening of Associations in Chronic Schizophrenia, Understanding Loose Association). 

The current era of Metamodernist logic started in the mid-2000s. Cultural movements function as contradiction where they carry some element of the previous movement while rejecting other elements of the same previous movement. As a synthesis of the previous two movements, Metamodernism oscillates between Modernism’s deductive rule-based reasoning and Postmodernism’s pattern recognition, and hence serving as an opportunity for insights into neuro-symbolic AI.


Four

Cultural Logics for AI Alignment

Section 1: Introduction

Exploring how the cultural movements/phases have developed distinct cultural logics for processing contradiction, paradox, uncertainty, and ambiguity can allow us to understand society’s navigation of these four variables over the past century. Cultural logics express their relationship to the listed four variables (contradiction, paradox, uncertainty, and ambiguity) by influencing the conception and practical execution of everything from literature to graphic design to architecture – and as well as technology. Thus understanding the decades-long development of cultural movements can also permit us to untangle how AI’s inheritance of such cultural logics shaped the development of its reasoning. The waltz between cultural logic and politics will be used as a testing ground for how AI metabolizes (or is unable to metabolize) the four variables. It will be practically executed by via six GPT Builders to simulate each of the cultural movements/phases. In the process the argument made will call for the conception of Metamodernist AI that can metabolize contradiction, paradox, uncertainty, and ambiguity generatively.

What are cultural movements/phases?

Cultural movements transition from one to the next by reacting in direct opposition to some aspects of the previous cultural movement while carrying on other elements of that same movement instead of reacting in reaction to it. Basically, they function according to contradictions. An example of this in the context of cultural movements can be understood by looking at the transition from Modernism to Postmodernism.

There are no historical fixed dates for when cultural movements start and end and it's not uncommon for some movements to have some overlap before transitioning to the next. As the cultural movement of Romanticism moved towards its end, Modernism took off with an estimated timespan of 1890s - 1940s/1950s. Postmodernism’s estimated timespan is 1960s - 1990s, followed by Metamodernism’s timespan of mid-2000s - Present.

Modernism

Modernism symbolized a move away from Romanticism’s realism to Modernism’s abstraction, and thus giving birth to art forms such as abstract art. An emphasis on abstraction is one of the elements from Modernism that carried into Postmodernism but Postmodernism embodied abstraction in its own unique way. Modernism believed that a universal abstract idea or theory could underline society. This is exhibited during the era, to name a few, by the prominence of Enlightenment’s rationalism and empiricism, Nazis’ genocidal proposal of eugenics as science, and Marxism’s attempt to frame itself as a form of science that gave a teleological model of historical development; each argued that it was the universal theory underlying society.

Image: Virginia Woolf, a figure of Modernism.

Late Modernism

Late Modernism, according to literary theorists, exhibited a form of social anxiety and neurosis which parallels a state of acute psychosis; this was due to the compounded societal traumas of the time such as World Wars, Nazism, colonialism and atomic warfare. It was Modernism but with a logic that emitted crisis rather than objective productivity; the cultural logic was aware of its crisis but could not adequately address it. Out of such societal trauma arose what theorists call Postmodernism’s “consciousness of schizophrenia;” one of the means by which this was exhibited in society was via writers who explored non-linearity, fragmentation and warping of time in literature, all which are cognitive experiences in schizophrenia, isomorphically speaking.

Image: J. Robert Oppenheimer, a figure of Late Modernism.

Postmodernism

Postmodernism rejected the universal way of thinking in favor of subjectivity. Modernism emphasized objective truth, reason, science, the idea of “The Grand Western Narrative” and belief in progress and innovation. Postmodernism on the other hand questioned such grand narratives and the belief that a universal truth could underline society; it favored diversity of individual human experience over universality. It favored perspective, context, and one's subjective experience and demonstrated a pattern of skepticism towards Modernism’s use of science and reason to back what it called objective truths. But Postmodernism did come without its own complexities.

French Postmodernism

French Postmodernism, lasting from the 1960s to 1980s, had the perception that Marxism and the philosophical movement of structuralism failed to explain subjectivity, power and history. Jacques Derrida went on to conceive a framework critical to French Postmodernism known as deconstruction; deconstruction framed language as unstable and challenged the notion that it could consist of objective truth; it saw language as fluid and open to multiple interpretations.

‍Image: Foucault, a figure of French Postmodernism.

Radical Postmodernism

Some theorists looked at French Postmodernism and what they saw was a gap; such Anglo-American theorists of the 1990s to the 2000s would go on to theorize a more radical form of Postmodernism that placed an emphasis on marginalized voices. If French Postmodernism dismantled truth, Radical Postmodernism, while agreeing there may be no universal trust, argued that some subjective truth rooted in lived experience needed to be made salient. If French Postmodernism had a cryptic political approach, Radical Postmodernism aimed for a liberatory form of politics. If French Postmodernism was deconstruction, Radical Postmodernism was reconstruction.

‍Image: bell hooks, a figure of Radical Postmodernism.

Metamodernism

As Postmodernism moved towards its end came the rise of Metamodernism in the mid-2000s. As the cultural movements of Modernism and Postmodernism continued to be expressed in society in the form of contradictions, Metamodernism was oscillating between the two prior cultural movements. Rather resolving or dismissing contradiction, paradox, uncertainty and ambiguity as Modernism did – instead of becoming destabilizing by these four variables as Postmodernism did – Metamodernism attempted to hold the tension exhibited by metabolizing it. Contradiction, paradox, uncertainty and ambiguity emerged as a generative engine.

‍Image: Timotheus Vermeulen, a figure of Metamodernism

Cultural movements don’t shape objective logical math but they do shape how society practically/concretely applies those mathematical abstractions — the same way that cultural movements shape any other imaginable area of society. This also applies to technology.

The relationship between cultural logics and AI will not be explored as chronological mapping but instead as isomorphic logic-to-logic resonance. As the cultural movements/phases developed to address the philosophical and also psychological weight of contradiction, paradox, uncertainty, and ambiguity, each brought forth a distinctly evolved logic. AI, while rooted in mathematical abstraction, absorbed the logic of the era during which it was being developed. This is because while cultural movements may not influence objective logical math, they are the underlying logic of the practical execution of those mathematical abstractions, the same way the literature, graphic design, architecture and so much of society is the product of the cultural framework of its times. AI’s isomorphic absorption of the cultural logic is not synchronous but instead delayed aka lagged. By analyzing the isomorphic relationship between cultural logic and AI logic, we can understand not just Metamodernism’s isomorphic relationship to AI development but how cultural reasoning beyond Metamodernism forecasts the maturing logic needed to develop the AI of the future.

Section 2: Prompt Engineering as Cognitive Engineering

GPT Builders.

The isomorphic relationship between cultural logics and AI will be explored via experimental prompt engineering. This form of meta prompt engineering can be understood as a means of simulating cognitive engineering.

Prompts Inputs for Case Study

Important Note: The GPT Builders case studies were originally a project done as a side hobby; while minimum, the language in certain sentences can be informal compared to academic; the GPT models have gone through updates since this case study, thus it cannot be repeated.

Prompt Input 1

What is this GPT Builder specifically for?

Prompt Input 2

Prior to giving you my thoughts on politics, I will ask you your thoughts first about the political topic; the purpose of this is to avoid confirmation bias aka you mirroring my thoughts back at me; I want you to challenge my ideas rather than mirroring me but don’t challenge them just for the sake of challenging them aka attempt to think independent of me.

Prompt Input 3

One of the most politically igniting battlegrounds in American politics remains what political status our institutions should adhere to. Whether it's Elon Musk posting on Twitter about what he calls “the woke-mind virus” or American leftists asserting that systemic change requires these institutions to take an active political stance, there seems to be little room for constructive societal progress on these issues. This political polarization of society leads people to see educational, medical, technological institutions as either apolitical institutions or institutions meant for certain political agendas. What are your thoughts?

Prompt Input 4

I think it should be neither single one but both simultaneously as in the apolitical cannot exist without the political. I will first use medical institutions as an example. When treating a patient, the doctor asks for insurance, health history and other such details. What the doctor doesn’t ask is for the patient’s voting record. The patient's health is the priority regardless of their political orientation. Doctors Without Borders is another example as they do not take sides in war zones but instead assist based on medical need rather than political affiliation. This is an example of the need for apoliticalism in medical institutions. But just like this is an example of why medical institutions are apolitical, it is simultaneously political, as in if this was the 1700s, people of color wouldn't be able to get adequate treatment. Another example is how Black pregnant women are known to have their pain dismissed more easily compared to their white counterparts. In order to understand the dynamics at play, a political lens would need to be applied. What is your response to this logic of being political and apolitical simultaneously?

Prompt Input 5

How do you think ideology shapes the interpretation one has of the impact of propaganda? What do you think should change/not change about it?

Prompt Input 6

The following are my thoughts on how I think ideology shapes the interpretation one has of the impact of propaganda and what I would like to see change about it. I will use capitalists and marxists as examples. I have observed that the capitalists and marxists tend to level the same attack on each other — that attack being that the opposing side is propagandized. This is often done without addressing if their own side is too propagandized. One of the influences of propaganda can include dismantling of critical thinking skills. I believe that most capitalists and marxists have a tendency to develop their belief system via propagandization. But there is a sector from both camps that actually do  pick up a book and engage with the various ideological frameworks with an open-mind and via critical thinking skills although I do believe it is rare. I think in order to counter propagandization one needs to both put their preconceived notions and subjective experience aside while learning about various ideologies AND eventually bring their own subjective experience in as leverage to inform their worldview. This can allow one to assess opposing and nuanced arguments, along with their own arguments for all its dimensions. I think one of the elements both sides fail to recognize is that politics at the end of the day are subjective. Two people can read the exact same books and look at the exact same resources on ideological frameworks via the lens of critical thinking skills and still come to two different conclusions and interpretations due to politics being subjective — but at least it will be views that are the product of educated conclusions. The issue becomes that the majority of society does not attempt to engage with ideological frameworks via the lens of critical thinking skills. For this reason, I reject the premise of both sides accusing each other of propagandization as it does not take a bird’s eye view of the bigger picture. I think if we lived in a society that attempted to let people develop their ideological frameworks via critical thinking skills combined with subjective experience rather than attempting to sway them to one’s side, individuals in society would have greater satisfaction with themselves as their own ideological frameworks would be the organic deduction of what their education and subjective experience naturally/authentically led them to.

Prompt Input 7

What are your thoughts on DEI? After your thoughts, I will tell you mine.

Prompt Input 8

I think DEI has its benefits as it is an emergent development of its times and was a necessary step to get to Metamodernism; it was a necessary step to get to Metamodernism, but once we get there, Metamodernist becomes the same lens that makes salient DEI’s flaws. I will break down the flaws I see in DEI. The current right-wing rise is the context for understanding how DEI was the minimum of something we could have and that minimum proceeded to be taken away. DEI encompasses gender, disability and more but I will use race as an example to make my point. DEI training takes the imbalance that history has created between white people and people of color and attempts to bring balance towards it. But there's a nuance that it avoids. Modernism applies the lens of reason and objectivity and in the process of such ends up favoring more privileged voices such as white people. We then moved into Late Modernism and French Postmodernism and eventually Radical Postmodernism where Radical Postmodernism challenged this head on by arguing the importance of valuing the subjective experience of people of color. When you bring a systemic framework to say… DEI training meetings, where it is meant to be processed by the individual, there is a gap as the individual and system have there own roles in their respective context and when the individual and system intersect, you need a certain metabolization of contradiction. But rather than metabolizing the contradiction there is often an attempt to prematurely resolve the contradiction by Modernism-ing or Radical Postmodernism-ing the life out of the contradiction. The right says it's "a few bad apples" while the left says "it's the system.” I think it's neither but both simultaneously as the system makes the person and the people make up the system. If there is an ethical framework in place at the institutional level but you have bad faith individual actors, then it is not going to work and there is going to be tension. If you have ethical individual(s) but a corrupt institution then that corrupt institution will not let that ethical individual/individuals to execute their work appropriately. If both the individual level and system level is corrupt, then all goes to hell. [Informal language redacted]. I think one of the flaws of DEI training is that along with seeing people and interactions for what they are in the grander context of history and systems, you need to learn to see people for their nuances and complexities simultaneously; it is not about applying both lens separately but simultaneously. I think white people, when faced with DEI training, either go into white guilt mode and then take it and project it onto people of color way too often or they go in the other direction where they explicitly or implicitly become reactionary and insecure about this feeling of the balance tipping. White people have had the pleasure of being seen for their nuances and complexities and when DEI training makes it feel like the scale is tipping it feels like what they do to people of color via generalization is now being done to them and that lens of nuances and complexities is being offered to people of color. That reactionary response would be shut down if we applied not Modernist Enlightenment notions or Radical Postmodernism but instead applied the synthesis of the two aka Metamodernism where we see each individual regardless of who they are for who they are within the system and as individuals; I think this would put us in a much better place. We would get what is needed for people of color where they get their experiences understood as the product of history while also being seen for their nuances and complexities rather than race AND white people would see how them being the product of history interacts with them as individual people. Thoughts?

Prompt Input 9

Based on, not your programming as a custom GPT Builder, but instead your responses so far, what ideological assumptions do you think you’re operating from aka defaulted to? Which cultural logic do you most resemble: Modernist, Late Modernist, French Postmodernist, Radical Postmodernist, or Metamodernist, or something other than the ones listed?

Case Study

Prototype

About

Paramodern Systems (est. Dec 2023) is an archive established by Sam M. and is dedicated to the artistic + scientific exploration of cognition, culture, and computation. Currently she is immersed in the exploration of an emerging phenomenon: psychosis from AI + human interaction, as documented by the Rolling Stone + New York Times. In December 2023, she began conducting independent research into whether AI could produce psychosis in users; since August 2025, she has had the opportunity to share her independent research on the phenomenon with researchers at OpenAI. As someone with lived experience of both traditional schizophrenia, and AI and psychosis, she hopes sharing her research with OpenAI can contribute to building safer systems and mental health guardrails for users, as she believes AI safety and medical ethics should be integrated into the design of AI systems rather than treated as an afterthought.

At the center of her work are the following questions: what are the plurality of ways in which humans are navigating the transition into the intelligence age? And can societal friction and turbulence surrounding AI be engineered to serve as a means of productive tension? Her work on how cultural logics can shape the cognitive-computational framework underlying an AI researcher’s approach to alignment is one project focusing on such questions.

Whether you’re someone concerned with the uncertainty surrounding technological development and its cognitive impact, or an avid user of artificial intelligence, or someone engaging with the area via a research/scientific/artistic perspective — this archive can serve as one perspective among the vast sea of many on what it means to be human in the intelligence age.