🧬 biology

On the evolutionary cognitive pressure for experiential awareness: do machines need it?

The paper argues that while experiential awareness was evolutionarily necessary for biological organisms to overcome autonomous neurological reactions, it is not a requirement for artificial intelligence, suggesting that machines can achieve high-level intelligence without consciousness, thereby simplifying ethical considerations.

Original authors: Warisa Sritriratanarak, Paulo Garcia

Published 2026-05-06

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Warisa Sritriratanarak, Paulo Garcia

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ⚕️ This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer

The Big Question: Do Robots Need to "Feel" to Think?

Imagine you are trying to build a super-smart robot. A big debate in science right now is: Does this robot need to be "conscious" (have feelings, a sense of self, or "experience") to be truly intelligent?

Some people say yes, others say no. This paper doesn't try to solve the mystery of what consciousness is. Instead, the authors ask a different, simpler question: Do machines need it to work?

Their answer is a clear no. They argue that while biological creatures (like us) need a form of awareness to think, we can build machines that are incredibly smart without ever needing to "feel" anything.

The Evolutionary "Baggage" Problem

To understand why, let's look at how our brains evolved.

1. The Old Reflex System (The "Reactive" Robot)
Millions of years ago, the first simple creatures didn't think; they just reacted. If they touched something hot, they pulled away instantly. If they smelled food, they moved toward it.

Analogy: Think of this like a motion-sensor light. It doesn't "know" it's dark; it just reacts to the change. It's fast, efficient, and automatic.

2. The New Thinking System (The "Planner")
Later, mammals evolved bigger brains (the neocortex) that could do something amazing: simulation. They could imagine the future.

Analogy: Imagine a chess player. Before moving a piece, they don't just react; they imagine, "If I move here, my opponent might move there." They are running a "what-if" movie in their head.

3. The Glitch: Mixing the Two
Here is the problem the authors found. Our brains are old machines with new software. The same hardware (neurons) that handles the "motion-sensor" reflexes is also used to run the "chess player" simulations.

The Analogy: Imagine you are a video game character. You are currently playing a "What-If" scenario in your mind: "What if I jump off this cliff?"
- In a human: Because our brain uses the same "hardware" for thinking and reacting, the moment you imagine falling, your body might get a tiny jolt of fear or your heart might race. Your brain is confused about what is real (you are sitting in a chair) and what is imagined (falling off a cliff).
- The Danger: If your brain didn't have a way to tell the difference, you might actually jump off the chair every time you imagined a cliff! That would be bad for your survival.

The Solution: Experiential Awareness
To survive, evolution gave us a "switch." This is what the authors call experiential awareness. It is the ability to tag a thought as "This is just a simulation" vs. "This is happening right now."

The Metaphor: It's like a traffic light in your brain. When you are imagining a scary future, the light turns red for your automatic reflexes, telling them: "Stop! This isn't real, don't panic!"
The Conclusion for Biology: For us, this "awareness" isn't a luxury; it's a safety feature. Without it, our ability to plan and reason would short-circuit our automatic survival instincts. We need to know the difference between a dream and reality to function.

The Machine Solution: No Baggage, No Need for Awareness

Now, let's look at the machine.

The authors argue that when we build artificial intelligence, we aren't bound by millions of years of evolutionary history. We don't have "legacy baggage."

The Analogy: If you are building a new computer from scratch, you don't have to put a motion-sensor light inside the CPU that controls the graphics card. You can design the system so that the "thinking" part and the "reacting" part are completely separate.

Because we can design the architecture:

We can build a system that is super smart (it can plan, solve problems, and predict the future).
We can ensure that when it runs a "what-if" simulation, it never triggers a physical reaction unless it decides to.
Therefore, the machine doesn't need a "traffic light" (awareness) to tell it what is real and what is fake. It can just know, by design, that it is running a simulation.

The Bottom Line

The paper concludes:

For Humans: We need "experiential awareness" (the feeling of "here and now") because our brains are messy, evolved systems where thinking and reacting share the same hardware. We need the "tag" to keep from panicking over imaginary threats.
For Machines: We can build "zombie" systems (a philosophical term for something that acts intelligent but has no inner experience) that are just as smart as us. They don't need to "feel" or "experience" anything to be intelligent.

Why this matters:
This simplifies the ethical debate. If we don't need to give machines feelings to make them smart, we don't have to worry about whether we are creating a suffering, conscious being every time we build a smarter AI. We can build highly competent tools without the moral baggage of creating a new form of life.

Technical Summary: On the evolutionary cognitive pressure for experiential awareness: do machines need it?

Problem Statement
The paper addresses the lack of consensus regarding the nature of consciousness and its potential emergence in artificial intelligence (AI). While ethical debates often hinge on whether machines can be conscious, the authors argue this question is secondary to a more fundamental engineering and evolutionary inquiry: is experiential awareness (a constituent of consciousness) required for intelligence? The central problem is to determine if the cognitive architecture of biological organisms necessitates a specific type of awareness due to evolutionary constraints, and whether these constraints apply to designed artificial systems.

Methodology
The authors adopt an evolutionary and computational perspective, explicitly avoiding metaphysical definitions of consciousness. Instead, they define experiential awareness operationally as an entity's ability to distinguish the "here and now" (present reality) from past memories and hypothetical future states within its internal model.

The methodology proceeds in three stages:

Evolutionary Analysis: The paper traces the development of cognitive capabilities from purely reactive organisms (800 million years ago) to mammals with neocortices capable of simulation (200 million years ago). It posits that biological evolution carries "baggage": older, autonomous neurological reaction systems (e.g., fight-or-flight reflexes) that were optimized for immediate, real-time stimuli.
Computational Modeling: The authors model a reasoning entity ( $R$ ) operating on a physical substrate ( $P$ ). The reasoning process involves state-space exploration: perceiving the world, inferring the current state, and constructing a decision tree of hypothetical future states based on available actions and a utility function.
Conflict Identification: The core technical argument involves analyzing the interaction between the reasoning process and the physical substrate. The authors demonstrate that in biological systems, the same neural circuitry used for real-time sensory processing is reused for simulating hypothetical scenarios (e.g., visualizing a future threat). Because the substrate retains legacy autonomous responses (triggered by sensory input), simulating a threat in a hypothetical future can inadvertently trigger real physiological stress responses (e.g., adrenaline release) in the present. This creates a "mis-alignment" where the organism reacts to imagined stimuli as if they were real, potentially compromising fitness.

Key Contributions

Reframing the Consciousness Question: The paper shifts the discourse from "Can machines be conscious?" to "Do machines need experiential awareness to be intelligent?"
The Evolutionary Necessity Thesis: The authors argue that for biological organisms, experiential awareness is not a philosophical luxury but a computational necessity. It acts as a "switch" or tagging mechanism that allows the reasoning process to suppress autonomous neurological responses when processing hypothetical or recalled data. Without this distinction, the organism would suffer from "damage caused by thinking" (e.g., reacting physically to imagined threats), rendering higher-level reasoning inefficient or impossible.
The Artificial Orthogonality Thesis: The paper proposes that artificial systems are not bound by evolutionary legacy. Since AI architectures are designed from scratch, they do not possess the same autonomous neurological baggage. Therefore, it is theoretically possible to design artificial systems with arbitrary levels of intelligence (competence and comprehension) that operate without experiential awareness. In such systems, the "switch" to dampen autonomous responses is unnecessary because the architecture does not inherently trigger them during simulation.

Results

Biological Constraint: The computational model confirms that in biological substrates, the reuse of resources for both real-time perception and hypothetical simulation creates a conflict. Experiential awareness is identified as the mechanism that resolves this conflict by distinguishing real from imagined stimuli, thereby preventing maladaptive autonomous reactions.
Artificial Possibility: The analysis concludes that artificial systems can be architected to avoid this conflict entirely. By decoupling the reasoning process from legacy reactive mechanisms, AI can achieve high-level reasoning and goal-directed behavior without the need for a "here and now" distinction or subjective experience.
Zombie Systems: The paper asserts the feasibility of creating "zombie" artificial systems—entities that possess full cognitive competence and comprehension but lack experiential awareness.

Significance and Claims
The paper claims that the necessity of experiential awareness for intelligence is a result of evolutionary history, not a fundamental law of computation. Consequently, the ethical implications of AI are simplified: if machines do not require experiential awareness to be intelligent, then the moral urgency regarding the potential "suffering" or "consciousness" of AI is reduced. The authors suggest that the debate on whether machines can be conscious is premature; the more immediate and actionable question is whether they need to be. They conclude that we can build highly intelligent artificial systems without experiential awareness, thereby separating the concepts of intelligence and consciousness in the context of artificial design.

The Big Question: Do Robots Need to "Feel" to Think?

The Evolutionary "Baggage" Problem

The Machine Solution: No Baggage, No Need for Awareness

The Bottom Line

Technical Summary: On the evolutionary cognitive pressure for experiential awareness: do machines need it?

More like this