A systematic approach to answering the easy problems of consciousness based on an executable cognitive system

This paper proposes a systematic approach to Chalmers' "easy problems" of consciousness by demonstrating how an executable cognitive system, grounded in Kantian conceptual knowledge, can computationally derive abilities such as discrimination, attention, and reportability through its learning, emotional, and information-manipulation mechanisms.

The Gist

Here is an explanation of the paper, translated into simple language with creative analogies.

The Big Idea: Building a "Conscious" Robot to Solve a Mystery

Imagine you are trying to understand how a human mind works. For decades, scientists have been stuck on a big puzzle. Philosopher David Chalmers split the puzzle into two parts:

1. The Hard Problem: Why does it feel like something to be you? (Why do you feel the warmth of the sun, not just measure it?)
2. The Easy Problems: How does the brain actually do the work? (How does it see a cat, decide to pet it, and tell you "that's a cat"?)

The author of this paper, Qi Zhang, says: "Let's stop guessing and start building." He created a computer program called "The Adder." It's not a super-intelligent AI like ChatGPT; it's a simpler, more honest machine designed to show exactly how a system can learn, think, and report what it sees.

By watching "The Adder" work, the author claims we can solve all the "Easy Problems" of consciousness.

---

The Blueprint: How "The Adder" is Built

Think of "The Adder" as a tiny, digital brain with three main rooms:

1. The Library (Semantic Memory): This is where the machine stores "facts" and "concepts." It's like a library where books aren't just words, but bundles of features.

   • The Analogy: Imagine a book about a "Square." It doesn't just say "Square." It has a tag that says: 4 sides, equal length, 90-degree corners.
   • Kant's Insight: The author uses an old idea from the philosopher Immanuel Kant. He says a "concept" is just a list of things that different objects have in common. To learn what a "dog" is, you ignore the fact that one is brown and one is black, and you focus only on what they share (four legs, fur, tail). The Adder does exactly this.

2. The Movie Reel (Episodic Memory): This is the machine's diary. It stores specific moments in time, like a video camera recording a day in the life.

   • The Analogy: If you see a red ball roll across the floor, the Movie Reel saves that exact sequence. Later, the machine can play this reel back, either in order or randomly.

3. The Boss's Office (The Operator): This is the most important room. It's the equivalent of the human Prefrontal Cortex (the part of the brain behind your forehead that plans and decides).

   • The Analogy: This is the manager. It looks at the facts from the Library and the videos from the Movie Reel. It decides what to do, counts things, adds numbers, and tells the machine what to say or draw.

(Note: There is also an "Emotion" room, but in this specific machine, it's just a placeholder to remind us that feelings help us pay attention.)

---

Solving the "Easy Problems" with the Adder

Here is how the Adder explains the seven "Easy Problems" using simple metaphors:

1. Discrimination, Categorization, and Reaction

• The Problem: How do we tell a cat from a dog, and then react to it?
• The Adder's Solution: It uses a "Match-Game."
  • When the Adder sees a shape, it checks its Library. Does this shape have "4 equal sides"? Yes. Does it have "90-degree corners"? Yes.
  • It matches these features to the concept of "Square."
  • Because it matched "Square," it automatically ignores the concept of "Circle."
  • The Magic: You don't need a separate brain part for "seeing" and another for "deciding." The act of matching features is the act of recognizing and reacting. It's like a key fitting into a lock; once it fits, the door opens.

2. Integrating Information

• The Problem: How does the brain combine sight, sound, and memory into one thought?
• The Adder's Solution: It builds a "Tower of Blocks."
  • First, it sees a line (Level 1).
  • Then it sees four lines making a box (Level 2).
  • Finally, the Boss's Office (Operator) counts the boxes.
  • The Magic: Integration happens at every step. You aren't just adding numbers at the end; you are combining features as you go up the tower.

3. Reportability (Talking about what you see)

• The Problem: How can we say, "I see a red ball"?
• The Adder's Solution: You can only report what you have processed.
  • The Adder can only "speak" (output a symbol or draw a picture) after the Boss's Office has looked at the data and organized it.
  • The Magic: If the Boss hasn't processed the info yet, the machine stays silent. This explains why we can't report things we haven't fully "thought" about yet.

4. Accessing Internal States

• The Problem: How do we know what we are thinking?
• The Adder's Solution: The Boss's Office is the only place where information becomes "public."
  • Before the Boss looks at the data, it's just raw noise in the Library. Once the Boss processes it, it becomes "accessible."
  • The Magic: You only know what you are thinking when your "inner manager" has reviewed the memo.

5. Attention

• The Problem: How do we focus on one thing and ignore the rest?
• The Adder's Solution: It uses a "Bouncer at the Door."
  • The Boss's Office has a door (an interface). It only lets in the information that is relevant to the current task.
  • Also, the "Emotion" room acts like a magnet, pulling the Boss's attention toward exciting or scary things.
  • The Magic: Without this bouncer, the Boss would be overwhelmed by a million thoughts at once (white noise). Attention is just the bouncer saying, "You're in, you're out."

6. Deliberate Control

• The Problem: How do we choose to do something on purpose, like solving a math problem?
• The Adder's Solution: The Boss uses "Rules of the Road."
  • The machine has learned rules (like "add these two numbers"). When the Boss decides to use a rule, it manipulates the information on purpose.
  • The Magic: Deliberate control isn't a magic spark; it's just the Boss deciding to apply a specific rule from the Library to the current situation.

7. Wakefulness vs. Sleep (Dreaming)

• The Problem: What is the difference between being awake and dreaming?
• The Adder's Solution: It's all about where the data comes from.
  • Wakefulness: The Boss gets data from the outside world (cameras, sensors). It has full control.
  • Sleep (Dreaming): The outside world is cut off. The Boss gets data from the "Movie Reel" (Episodic Memory), but the clips are played randomly.
  • The Magic: In a dream, the Boss is still working hard, trying to make sense of random clips from the past. It's like watching a movie where the scenes are shuffled. The Adder shows that even in this chaotic state, the machine is still learning and organizing memories.

---

The Big Takeaway

The paper argues that we don't need to invent a magical "soul" or a mysterious "consciousness chip" to explain how we think.

Instead, consciousness (at least the "easy" parts) is just a very sophisticated information processing system.

• Learning is just finding common patterns.
• Thinking is just the Boss's Office organizing those patterns.
• Dreaming is just the Boss's Office playing with old patterns when the outside world is quiet.

By building a machine that does this step-by-step, the author proves that these complex human abilities are actually just the result of simple, mechanical rules working together. It turns the mystery of the mind into a blueprint we can actually build and test.

Technical Summary

Here is a detailed technical summary of the paper "A systematic approach to answering the easy problems of consciousness based on an executable cognitive system" by Qi Zhang.

1. Problem Statement

The paper addresses the "easy problems" of consciousness as defined by David Chalmers (1995). These are phenomena that can be explained in terms of computational or neural mechanisms, including:

• Discrimination, categorization, and reaction to stimuli.
• Information integration.
• Reportability of mental states.
• Access to internal states.
• Focus of attention.
• Deliberate control of behavior.
• The difference between wakefulness and sleep.

The Gap: Despite three decades of research, these problems lack a systematic, unified explanation. Existing theories (Global Workspace Theory, Integrated Information Theory, Higher-Order Thought, etc.) are largely conceptual and lack sufficient executable computational mechanisms to demonstrate how these functions arise. Furthermore, the search for Neural Correlates of Consciousness (NCC) has been hindered by the difficulty of distinguishing core correlates from prerequisites and consequences in biological data.

Objective: The author proposes an alternative approach: using an executable artificial cognitive system to systematically demonstrate and explain these mechanisms within a single framework, thereby providing a computational proof-of-concept for the easy problems.

2. Methodology

The study utilizes a custom-developed executable cognitive system named "adder." The methodology involves:

• System Architecture: The "adder" is built upon a cognitive construct inspired by Kantian philosophy (concepts as common features) and Hebbian learning ("what fires together, wires together").
• Core Components:
  • Semantic Memory: Composed of a Feature Subsystem (FS) and a Symbol Subsystem (SS). The FS extracts common features from sensory input, while the SS learns symbolic representations. They are synchronized to link symbols with meanings.
  • Episodic Memory: Stores sequential segments of experience, allowing for replay (both sequential and random).
  • Mental Operator: Acts as the system's equivalent of the Prefrontal Cortex (PFC). It extracts operational rules (cause-and-effect), manipulates information, and integrates data from memory and emotion.
  • Emotion Block: A non-functional but conceptual module representing the limbic system's role in biasing attention and decision-making.
• Experimental Approach: The author addresses each of the seven easy problems by:
  1. Reviewing relevant empirical findings and theories.
  2. Explaining how the "adder's" specific mechanisms solve the problem.
  3. Demonstrating the solution via the system's actual output (semantic reports, image generation, and arithmetic operations) in response to stimuli.

3. Key Contributions

The paper's primary contribution is a unified computational framework that derives complex cognitive functions from a single learning mechanism based on common features.

• Unified Mechanism for Discrimination and Categorization: The study argues that discrimination, categorization, and reaction are not distinct mechanisms but emergent properties of a system learning common features. If an input matches a stored set of common features, it is categorized; if it lacks features, it is discriminated against.
• Operational Knowledge as "How": The system distinguishes between descriptive knowledge (what things are) and operational knowledge (how things work). The Mental Operator extracts operational rules (e.g., addition, subtraction) from sequences of events, enabling deliberate control.
• Reportability and Internal Access: The system demonstrates that reportability is not a magical property but a functional stage where processed information (via the Operator) is projected back to the Symbol and Feature subsystems to generate output. This aligns with the PFC's role in human reportability.
• Attention and Deliberate Control: Instead of a dedicated "attention" module, the system uses filtering interfaces at region boundaries and emotional bias to select information for the Operator. Deliberate control is shown to be the application of acquired operational rules within the Operator.
• Wakefulness vs. Sleep: The paper posits that the difference between wakefulness and dream sleep lies in the source of stimuli. Wakefulness processes external sensory input, while dream sleep processes randomly activated segments of episodic memory. Both states utilize the same learning mechanisms (consolidation), explaining why sleep aids memory.

4. Results and Demonstrations

The "adder" system successfully demonstrated the following capabilities, serving as evidence for the proposed mechanisms:

• Discrimination & Categorization: The system correctly identified geometric shapes (lines, squares, rectangles) in images based on common features (e.g., "4 equal sides," "perpendicular adjacent sides"). It successfully inhibited incorrect categorizations when features did not match.
• Information Integration:
  • Low Level: Integrated sensory features to recognize objects.
  • High Level: Performed arithmetic (addition/subtraction) by integrating numeric counts and operational rules within the Operator.
• Reportability: The system generated semantic reports (e.g., outputting the symbol "5" or "Count") and image reports (drawing 5 horizontal lines) based on its internal state, proving that reportability is a function of processed knowledge.
• Deliberate Control: The system executed specific tasks (e.g., "Count to 5") by retrieving operational rules and manipulating data, demonstrating goal-oriented behavior without external step-by-step programming.
• Sleep Simulation: Referencing previous work, the system demonstrated that random firing of episodic memory (simulating REM sleep) facilitates learning and memory consolidation, supporting the activation-synthesis model.

5. Significance

• Bridging Theory and Implementation: The paper moves consciousness research from purely theoretical or purely biological observation to computational implementation. It proves that the "easy problems" can be solved by a system with specific, defined architectures.
• Validation of Kantian Concepts: It provides a modern computational validation of Kant's view that concepts are abstractions of common features, showing how this abstraction leads to discrimination and categorization.
• Unified Explanation: It offers a single framework where attention, control, memory, and reportability are not isolated modules but interconnected functions of a semantic knowledge system and an executive operator.
• Implications for AI: The study highlights that current Generative AI (like LLMs) lacks these specific cognitive constructs (grounded semantic memory and operational operators), explaining their failure in tasks requiring true counting or logical consistency, unlike the "adder."
• Reframing Sleep: It offers a mechanistic explanation for the cognitive benefits of sleep, suggesting that the lack of external input during sleep allows the system to reorganize internal knowledge structures (consolidation) without the noise of the external world.

In conclusion, the paper argues that the "easy problems" of consciousness are essentially inquiries into the structure and utilization of semantic knowledge in a cognitive system. By building a system that learns, integrates, and reports based on common features, the author provides a systematic, executable answer to how these cognitive phenomena arise.