Harnessing Intelligence with Frames

A Playbook for AI Life Games

(TBD)

Author
Affiliation

廖漢騰 (Han-Teng Liao)

獨立研究者

Published

July 23, 2026

Book cover showing the Illustrations of Alice and the Dodo from Alice's Adventures in Wonderland, by John Tenniel (1865)

Tip

📊 Character Counts / 字數統計

  • Chinese version: 8980 characters
  • English version: 21284 characters

🤗 Preface

“Divining with oracle bones?” 🐢 Can an AI Court Oracle really alter national destiny?

“Seeking a hug from ChatGPT after an exhausting day?” 👨‍👩‍👧‍👦 Can an AI Matchmaker truly change a person’s life?

From oracle bones to Large Language Models (LLMs), humanity has continuously sought ❝mental fill-in❞ to expand consciousness and solve problems.

How can we understand and master the latest mental fill-in technology—Artificial Intelligence (AI)?

Tip 1: Key Highlights of This Book💖
  • 👨‍👩‍👧‍👦 Target Audience: General readers, students, parents, and professionals interested in AI—suitable for both popular science education and career enhancement.
  • 🏮 Concise & Clear: Built around 66 key concepts1 to establish a complete scaffolding for knowledge and action, striving for clarity without being dogmatic.
  • 🤠 Dao / Artifice (道／盜): Highlighting the dual competitive and cooperative nature of intelligence: knowledge can establish rules, yet it can also challenge them2.
  • 🐛 problematics: Focusing on foundational problems such as the Symbol Grounding Problem and the Frame Problem, exploring their underlying problematics and solutions.
  • 🎄 Systematic Scope: Introducing AI’s main schools, paradigms, analytic forms, and orientations.
  • 💊 The AI Red Pill: Incorporating realist evolutionary and game-theoretic perspectives, such as Gestalt Psychology and Language Games, exposing manipulative mechanisms3.
  • 🪜 Personalized Approach: Guiding readers through “investigating things to extend intelligence” (格物致智 gé wù zhì zhì, i.e. a modern extension of Chinese saying of “Greater Learning” 格物致知 gé wù zhì zhī), building a personalized knowledge-action framework.
  • 🎍 Grounded Application Logic: Anchoring application logic in real-world scenarios while systematically selecting creative solutions.

↪️ For details, see Book Features.💖

🎲 The Dao & Artifice of Intelligence
An AI Game Playbook

This handbook is rooted in the foundational problematics (referring to foundational questions, frameworks, methods, and systemic awareness collectively established through a field’s development; see Part 1 ㉄) of artificial intelligence. Through rigorous fact-checking and logical tracing, it guides readers in constructing their own personalized 🪜 Knowledge-Action Scaffolding.

This scaffolding—capable of “framing intelligence” with holistic “perspective”—helps readers evaluate existing tech solutions while inspiring innovative approaches in AI Engineering (see Part 10 🌉).

The book focuses on the critical framing and perspective—both as righteous principle (“Dao”) and cunning maneuver (“Artifice”)—that humanity deploys when engineering artificial intelligence.

While anchored in rigorous knowledge, this book strives for clarity when addressing complex concepts, intentionally avoiding dogmatic assertions.

Tip🌌: Mind Map

A visual mind map of the book structure is available at Figure C.1. 🌌

📦 Structure of This Book

Anchored in core problematics at its origin (see Part 1 ㉄) and aiming toward innovative AI Engineering methodologies at its destination (see Part 10 🌉), this handbook traverses the following chapters:

  • Selecting knowledge and case studies focused on embodiment, game theory, and future foresight—ensuring systematic understanding of historical evolution and inspiring innovative future insights:
  • Systematically introducing schools and paradigms to integrate classical foundations with modern practices:
  • Highlighting systemic innovation and analytical depth by synthesizing the 5 “AI Orientations” (see Part 5 ☸) and 6 points of “Analytics & Decisions” (see Part 6 ❖).

The inclusion of Game‑oriented AI also echoes the “Dao / Artifice” theme because current LLM‑agent research uses simulated social interaction, cooperation, planning, and feedback as testbeds for strategic behavior, but finds persistent instability from prompt sensitivity, hallucinations, and control problems (Lu et al. 2024; Russell and Norvig 2021). While these studies show why games should be treated as AGI testbeds rather than proof of general intelligence, this book takes a step further by asking readers to be aware of the manipulation and power dynamics of AI design and use.

This chapter structure not only outlines a comprehensive framework but also encourages readers to build their own 🪜 Knowledge-Action Scaffolding, systematically generating essential Verb-Noun combinations via the author’s novel taxonomy Note B.3.

🖼️ Scope of This Book

From the earliest debates on the symbol grounding problem—how symbols gain meaning beyond human interpretation (Harnad 1990)—to the enduring frame problem of predicting what changes and what remains stable in a dynamic world (Shoham 1987; Schubert 1990), the foundations of AI have always been tied to both practical considerations and deep philosophical questions.

The scope of this book aims to curate absoulte minimum key terms (66 of them) to build a rich enough foundation for knowledge-action scaffolding and debates.

Contemporary surveys now emphasize that neither symbolic reasoning nor neural networks alone are sufficient; instead, hybrid approaches are emerging as the “third generation” of AI (Zhu and Su 2023; Ali and Dornaika 2025; Ye 2025; Russell and Norvig 2021). This shift is visible in the rise of embodied AI, where agents must perceive, reason, and act in real environments (Feng et al. 2025; Ma et al. 2024).

The book believes that despite the success of LLM in securing capital investments, it remains critical to be aware of the Schools & Paradigms, Games & Maths, Problematics & Engineering of AIs, and how they are woven into the wider understanding of not just “Artificial Intelligence”, but “Human Intelligence” as well.

🧭 “Three Linked Appendices”: A Framework for Knowing and Acting that works for Human and AI alike.

Supplementing and triangulating all the content chapters, this book presents three appendices that echo one another: “💪Action”, “🧠Brain”, and “🪜Capacity”.

Why do we need these? Recent conceptual work explicitly argues that effective human–AI collaboration depends on three linked questions: the ontological role of AI, how humans and AI jointly construct knowledge, and how systems practically enable that interaction (Apriyani and Patmanthara 2025). This maps closely to the Action–Brain–Capacity triad, and further supports treating both AI and human agency as a continuum rather than as separate categorie.

Together, these appendices provide taxonomies that serve as a critical roadmap for navigating the AI landscape in particular, and the broader learning journey in general. By unpacking “what to know” from “how to decide and act,” before recombining them for practical execution, the taxonomies deliberately refrain from presupposing any priority between “humans” and “machines.” Each appendix outlines parallel versions for both human and AI agents, enabling readers to choose any entry point.

  • First ask: “How should one act or prepare to act?”—See “💪Action”, with its verb grammar and quadrant of universal layers.
  • Next ask “Given the actions required or prepared, what is worth knowing?”—See “🧠Brain”, with its Triple Mind framework and concentric circles of nouns.
  • Finally ask : “Among many options, how can choices be converged?” —See “🪜Capacity”, with its Double Diamond Model and Scaffolding for constructive selection.

In academic terms, “Three Linked Appendices” roughly corresponds to Ontology, Epistemology, and Methodology, deliberately engineered to be substrate‑neutral. The same set of categories applies simultaneously to human and machine (AI) agents, rather than centering on human cognition and extrapolating to machines secondarily. It is important to note that this tripartite division follows the traditional philosophical order; it runs in parallel to the book’s ABC (Action–Brain–Capacity) framework but does not require identical sequencing.

Academic Dimension Corresponding Appendix Core Content
Ontology (What exists) “🧠Brain” Concentric Circles of Nouns (Core / Middle / Outer); ⟨Tri-Aspect Mind⟩ (Reactive / Emotive-Relational / Reflective-Symbolic)
Epistemology (How we know & categorize) “🧠Brain” (Primary), “💪Action” (Secondary) Human vs. Machine Learning contrasts (Behaviorism / Situatedness / Connectionism); AQAL Quadrant cognitive pathways (Subjective / Objective / Intersubjective / Interobjective)
Methodology (How we decide & act) “💪Action”, “🪜Capacity” Action syntax and dynamic spiral evolution across AQAL Quadrants; Double Diamond divergence and convergence («Options» → «Choices»)

Thus, the three appendices are not isolated topics, but three complementary facets of a single agent-neutral framework: establishing ontology and epistemology first (Brain), unfolding the option space of action (Action), and finally converging via methodology into executable choices (Capacity). This structure underpins the book’s parallel treatment of humans and machines. Readers interested in specific theoretical contexts (such as dual-process theory, triune brain theory, social brain hypothesis, or the three major AI paradigms) will find source citations and extended links within the corresponding appendix.

Synthesizing selected classical and contemporary authorities across Eastern and Western traditions through selective socio-technical integration, this framework moves beyond listing foundational problematics to cultivate agents’ structural scrutiny—the ability to identify, navigate, and reconstruct system frameworks in real-world application domains. Readers are encouraged to gather their own data sources and core action verbs and nouns to construct alternative framework expressions.

Readers are encouraged to start their exploration from either of these two entry points:

Tip 2: Zen Poem: 🎲 The Game vs. 🏁 The Framing

This book is titled Framing Intelligence, also subtitled The Dao & Artifice of Intelligence. Beyond emphasizing the core AI problem—The Frame Problem—it attempts a touch of Zen philosophy to illuminate the game-theoretic dilemma of contemporary AI:

  • 🏆 “Within the game, victory and defeat feel real” ~ Inside the play, win and loss are palpable, yet remain relative phenomena.
  • 🧭 “Contemplating the frame, true composure is revealed” ~ Transcending game boundaries to observe the whole (framing) brings genuine grounding.
  • ❤️ “When the mind chases the game, it stays trapped within it” ~ Fixating on outcome binds the mind to the rules of play.
  • 🪷 “When wisdom observes the frame, its intent is to break it” ~ Seeing the totality reveals the frame’s existence and purpose, opening paths to new horizons.
  • ⚡ “Breaking the frame is not destruction, but revealing truth beyond” ~ Shattering a rigid game shouldn’t destroy, but uncover grounded reality outside.
  • 🏯 “Establishing a frame is not attachment, but nurturing good within” ~ Building a new boundary shouldn’t constrain, but create space for truth, goodness, and beauty.
  • 🌑 “Knowing only how to break leads into void” ~ Destruction without construction devolves into nihilism.
  • 🌞 “Mastering both breaking and building fulfills grand purpose” ~ Balancing critique with creation brings complete fulfillment.

🎲🏁

✨ Key Features of This Book

The primary features of this book include:

  1. A Knowledge Framework Focused on Perspective: Starting with “investigating things to extend intelligence” (格物致智), using 66 key terms to build a rich knowledge-action scaffolding that directly addresses core problematics (e.g., Symbol Grounding Problem, Frame Problem, and Language Games), facilitating systematic mastery of various schools, analytic forms, and underlying principles (“Dao”) to enable efficient cross-disciplinary dialogue and critical inquiry.
  2. Grounded Application Logic: Aiming to “extend intelligence” through actionable application paths, incorporating emerging subjects like Game-Theoretic AI, Embodied AI, AI Mathematics, and AI Engineering to creatively assimilate diverse developmental trajectories into practical usage.
  3. Reinterpreting Classical Concepts: Connecting classic knowledge points with modern topics like Large Language Models (LLMs) under the new idiom “investigating things to extend intelligence”, constructing a self-improving, expandable knowledge-action scaffolding. 讀者能習得謀定而後動的「框智力」，並擴展到組織的動態能力@sec-frame-intelligence 與模控治理 🌐🦾 Scaling Frame Intelligence: Organizational Dynamic Capabilities and Cybernetic Governance 。
  4. Integrating Science & Game Theory: Introducing concepts like Gestalt Psychology and Language Games to present the author’s evolutionary and game-theoretic perspectives on “investigating things to extend intelligence”, emphasizing a systems science view from human language to AI technology.
  5. The Dual Dynamics of Dao & Artifice: Utilizing the pun of “Dao” (righteous principles) and “Artifice” (cunning tactics) to highlight intelligence’s dual nature of competition and cooperation in human society. Examining practices from “divining with oracle bones” to “seeking a hug from ChatGPT” through evolutionary and game-theoretic lenses: knowledge creates rules, but it can also challenge them.
  6. Respecting Authority Without Blind Following: Striking a balance between humility and conviction, citing classic and contemporary authorities from both Western and Eastern traditions for comprehensive understanding. Quotation does not imply blind adherence—the goal is to help readers grasp the full picture while retaining room to form their own judgment and framework.

As this handbook accompanies readers through understanding AI’s varied problematics and solutions, it aspires to empower readers to embark on their own journey of practice and design rooted in systematic comprehension.

🎓 A Note for Rigorous Readers

A rigorous reader might reasonably ask: by casting human-made discipline as “Artifice” (盜) via Laozi, does this book advocate an anti-technology Daoist romanticism?

No. The pun on “Dao” and “Artifice” is a diagnostic critical heuristic, not a call to abandon technology—it points out that human technologies, from language to statecraft to AI, all carry the power to “reframe the game.”

This book’s core goal is to cultivate readers’ Meta-Framing Capabilities, developing their “Knowledge-Action Scaffolding Construction Ability.” Mastering modern AI requires holding two forms of thinking in tension: using Confucius’s rigorous “rectification of names” (必也正名乎) with one hand to construct executable engineering systems, while using Laozi’s and Wittgenstein’s “Dao/Artifice” insights and language games with the other hand to predict system boundary drift, prevent specification gaming (reward hacking), and dynamically reframe the game when rigid structures fail.

The author’s “Dao / Artifice” reminder shows that the AI Red Pill4 and AI Blue Pill5 are not a binary choice, but a dynamic tension to be held between both positions:

  • The deconstructive, diagnostic “AI Red Pill”6 (Laozi / later Wittgenstein): exposes how over-rigid rules and manufactured discipline invite systemic loopholes and gaming (Artifice/盜)—an early articulation of what modern AI alignment calls Goodhart’s Law and specification gaming (reward hacking), warning against dogmatic technical hubris and brittle hard-coded specifications.
  • The constructive, grounding “AI Blue Pill”7 (Confucius): real-world AI engineering cannot run on deconstruction alone. A deployable system still depends on “rectification of names”—precise entity grounding, schema contracts, ontologies, and enforceable protocols (Li).

This book’s actual position: Meta-Framing Capabilities.

📐 Mathematics & Code

This handbook provides basic mathematical definitions and code snippets, allowing readers without technical backgrounds to grasp the underlying mechanics through plain-text explanations before pursuing deeper study according to their needs.

🧮 Mathematical Examples

This book includes concise mathematical formulas primarily for illustrative purposes, requiring no advanced mathematical background. For example, consider the mathematical definition of Nash Equilibrium below:

\[ u_i(s_i^*, s_{-i}^*) \;\geq\; u_i(s_i, s_{-i}^*) \quad \forall i \in N, \; \forall s_i \in S_i \]

Designed as a popular science reference, the mathematical formulas in this book aim to lower the threshold for conceptual understanding.

🛠 Python Examples

This book contains minimal code snippets to demonstrate how core concepts are implemented in practice and to illustrate available choices.

For example, the code below demonstrates how context engineering achieves factual disambiguation and world grounding, showing how MCP tools complement LLMs while offering flexibility across different LLM providers:

# World Grounding Example: Preventing LLM hallucinations in ambiguous queries like "Georgia".

from langchain_community.chat_models import ChatOllama
from langchain_mcp_adapters import initialize_agent_executor_with_mcp

# 1. Define LLM Source
llm = ChatOllama(model="qwen2.5:1.5b", temperature=0)   # Local Qwen model via Ollama
# llm = ChatOpenAI(model="gpt-4o-mini", temperature=0)  # Cloud-based GPT-4o-mini
# llm = ChatOllama(model="claude3-haiku", temperature=0) # Local Claude small model via Ollama
# llm = ChatOllama(model="llama3.2:3b", temperature=0)   # Local Llama3.2 small model via Ollama

# 2. Register MCP Server (e.g., Country Lookup)
mcp_servers = {
    "country_lookup": {
        "command": "python",
        "args": ["./mcp_server_country.py"],  # MCP server providing ISO country codes
        "transport": "stdio",
    }
}

# 3. Initialize Agent Executor
executor = initialize_agent_executor_with_mcp(
    llm=llm,
    mcp_servers=mcp_servers,
)

# 4. Disambiguation Query requiring entity grounding
result = executor.invoke({"input": "Disambiguate Georgia and provide candidate ISO country codes."})
print(result)

📌 Notes

  1. A tribute to Japanese AI writer Youichiro Miyake’s book of 68 core concepts (三宅∙陽一郎（Youichiro　Miyake） 2017).↩︎

  2. The AI Red Pill: Borrowing the metaphor from the classic sci-fi film The Matrix (“swallow the red pill to see the harsh truth”). In this book, it refers to bringing in Laozi’s “Dao/Artifice” insight, the later Wittgenstein’s “language games,” and evolutionary game theory to help readers see through rigid rules, wordplay, and the loopholes, specification gaming (reward hacking), and common manipulation tactics that arise under system over-alignment.↩︎

  3. The AI Red Pill: Borrowing the metaphor from the classic sci-fi film The Matrix (“swallow the red pill to see the harsh truth”). In this book, it refers to bringing in Laozi’s “Dao/Artifice” insight, the later Wittgenstein’s “language games,” and evolutionary game theory to help readers see through rigid rules, wordplay, and the loopholes, specification gaming (reward hacking), and common manipulation tactics that arise under system over-alignment.↩︎

  4. The AI Red Pill: Borrowing the metaphor from the classic sci-fi film The Matrix (“swallow the red pill to see the harsh truth”). In this book, it refers to bringing in Laozi’s “Dao/Artifice” insight, the later Wittgenstein’s “language games,” and evolutionary game theory to help readers see through rigid rules, wordplay, and the loopholes, specification gaming (reward hacking), and common manipulation tactics that arise under system over-alignment.↩︎

  5. The AI Blue Pill: Borrowing the metaphor from The Matrix (“swallow the blue pill to remain within stable order”). In this book, it refers to Confucius’s constructivist engineering practice of “rectification of names”(必也正名乎)—that is, in real AI deployment, establishing explicit schema contracts, ontologies, system prompts, and normative protocols (Li) to prevent hallucination and ensure predictable, safe system behavior.↩︎

  6. The AI Red Pill: Borrowing the metaphor from the classic sci-fi film The Matrix (“swallow the red pill to see the harsh truth”). In this book, it refers to bringing in Laozi’s “Dao/Artifice” insight, the later Wittgenstein’s “language games,” and evolutionary game theory to help readers see through rigid rules, wordplay, and the loopholes, specification gaming (reward hacking), and common manipulation tactics that arise under system over-alignment.↩︎

  7. The AI Blue Pill: Borrowing the metaphor from The Matrix (“swallow the blue pill to remain within stable order”). In this book, it refers to Confucius’s constructivist engineering practice of “rectification of names”(必也正名乎)—that is, in real AI deployment, establishing explicit schema contracts, ontologies, system prompts, and normative protocols (Li) to prevent hallucination and ensure predictable, safe system behavior.↩︎