▘ Intro~ “Mental Fill-in” as Cognitive Shortcuts for Actions

Keywords

mental fill-in, cognitive shortcuts for actions

🥶🥵 When you see the Chinese idiom 「水深火熱」 (Shui-Shen-Huo-Re — literally “deep water and scorching fire”), what comes to your mind instantly?

🌬️🌞 When you see the the Chinese idiom 「水火不容」 (Shui-Huo-Bu-Rong — literally “irreconcilable as water and fire”), what comes to your mind?

⚔️🐉 If we pivot to Western pop culture and you hear the phrase “A Song of Ice and Fire,” what intricate plots, looming battles, and epic character arcs does your brain immediately start reconstructing out of thin air?

💧🔥 Ultimately, are “water” and “fire” (or “ice” and “fire”) even comparable things?

These seemingly “common-sense” questions actually involve multiple (and distinct) interwoven mental capacities.

Get ready to embark on a journey of mental and cognitive capacities using these primal elements as our guide, and witness how humans naturally perform ❝Mental Fill-in❞.

💧🔥 On Water and Fire

Regarding “water” and “fire”:

  • 🥶🥵 What is your “reactive” physical sensation when encountering extreme heat or biting cold?
  • 🌬️🌞 What is your “emotional-relational” memory when encountering relevant Chinese idioms or the fiction title of A Song of Ice and Fire?
  • 💧🔥 What does your “reflective-symbolic” scientific knowledge tell you about what “water” and “fire” actually are?

Furthermore, consider the evolution of human society:

  • 🪵🔥 What are the mythical stories and practical skills surrounding the use of fire?
  • 🌊🏗️ What are the mythical stories and contemporary engineering projects regarding water management and governance?
  • 🚂💨 How did the “fusion of water and fire” in the Industrial Revolution shape modern urban civilization and the contemporary AI Industrial Revolution?

Reflect and take notes: What answers can you ❝Mentally Fill-in❞ as responses to the questions above?

🤯 “Mental Fill-in” as Cognitive Shortcuts for Action

“Mental fill-in” (詳見 ?def-mental-fill-in 及 Important D.1) refers to the human behavior of relying on experience and imagination to automatically fill in complete “plots” or “facts” in the mind when information is incomplete.

It manifests in countlessly everyday moments:

  • 💌 A partner sends a text message with an ambiguous tone, and your mind immediately acts out: “Is he angry? Is she implying something? Or did I do something wrong?”
  • 📺 When background music swells over news commentary about two nations, your emotions overflow: “Ah, another conflict! One side must have provoked first, forcing the other to retaliate.” Verdicts on “who is right or wrong” and “where responsibility lies” are already drawn.
  • 🎭 Watching a drama unfold to a critical climax, you anticipate twist after twist: “This character will definitely betray them! No, wait, he is actually an undercover agent! Hold on, is the director going to flip it again?”

These mental activities of “mental fill-in” are both our instincts and potential sources of misreading reality. Emotions, stances, narratives… anything can be filled in.

🧠 Mental and Cognitive Capacity 🐸🐘🧘

By observing the different facets of “water” and “fire,” we have unconsciously mobilized multiple levels of mental and cognitive resources—ranging from instinctive reactions to emotional memories, and finally to symbolic knowledge reasoning.

  • 🐸⚡ 🙶Reactive🙷 Mind: The tactile sensation and temperature changes of “water,” and the instinctive avoidance of the heat and light of “fire.” These belong to rapid, thought-free physiological reflexes.
  • 🐘💞 🙶Emotional-Relational🙷 Mind: Idioms and scenarios like “water and fire are merciless,” “in deep water and scorching fire,” “like a person drinking water (knowing its temperature naturally),” or “getting along like a house on fire” evoke emotional memories and social interaction experiences.
  • 🧘☸️ 🙶Reflective-Symbolic🙷 Mind: Recognizing that “water” is \(H_{2}O\) and “fire” is a phenomenon of combustion requiring oxygen \(O_{2}\), allowing for further deductions of their physical and chemical conditions.

These levels do not operate in isolation; instead, they intertwine into a collaborative cognitive capacity system.

To analyze and apply these capacities more clearly, we can abstract and categorize them to form a cognitive framework that compares human intelligence with artificial intelligence.

It is precisely these cognitive frameworks that elevate the empirical rules of thumb (see also ?sec-heuristic ) of ❝Mental Fill-ins❞ into higher-level cognitive capacities. Frames work to elevate mental and cognitive capacity.

Important 1: 🧠 ⟨Tri-Cognitive Model⟩

The cognitive capacity classification originally integrated in this book provides a refined starting point for classifying noun <>:

  • 🐸⚡ 🙶Reactive🙷 Mind: Dominated by real-time perception and rapid response modules.
  • 🐘💞 🙶Emotional-Relational🙷 Mind: Centered around human-computer interaction, affective computing, and social simulation.
  • 🧘☸️ 🙶Reflective-Symbolic🙷 Mind: Focused on high-level reasoning, planning, ethics, and alignment.

Reflecting on human mental activities based on their processing speed, social functions, and symbolic capabilities allows us to distinguish our own cognitive capacities from those of machines.

For details, see 🧠 ❝Tri-Aspect Mind❞: Integrated Learning With a Taxonomy.

⚡💞☸️

Understanding how our innate cognitive abilities (e.g., “filling in the blanks” or “connecting the dots”) operate can unlock the key to understanding and innovating in Artificial Intelligence (AI):

Humans rely on experience and imagination to fill in the gaps in incomplete information, outlining “facts” that drive our “actions.” How do these “cognitive shortcuts for actions” understanding contribute to the development of AI?

🤯 ❝Mental Fill-ins❞: Cognitive Shortcuts

❝Mental Fill-ins❞ (Mental Fill-in) refers to the human behavior of automatically filling in complete “plots” or “facts” in the mind using experience and imagination when information is incomplete.

It manifests in countless daily moments:

  • 💌 Your partner sends a text message with an ambiguous tone, and your mind immediately stages a drama: “Is he angry? Is she implying something? Did I do something wrong?”
  • 📺 When the news mentions two countries over background music, your emotions overflow: “Ah, another conflict! One side must have provoked first, forcing the other to retaliate.” “Who is right and who is wrong” is already decided in your mind.
  • 🎭 When watching a drama at a critical climax, you eagerly anticipate twist after twist: “This character is definitely going to betray them! No, wait, he’s actually an undercover agent! Hold on, is the director going to pull off another twist?”

These mental activities of ❝Mental Fill-ins❞ are both our instincts and potential sources of misreading reality. Emotions, stances, stories… anything can be filled in.

It is this very mechanism that unlocks the key to understanding and innovating in Artificial Intelligence (AI):

The instinct for ❝Mental Fill-ins❞ is the human “cognitive shortcut for actions.” Relying on experience and imagination, we fill in the blanks within incomplete information, sketch out a “fact,” and drive ourselves to “take action.”

🖼️ Addressing AI Problematics ㉄

By sketching “facts” and driving “actions,” this structured view of cognitive capacities regarding “water” and “fire” can be utilized when building artificial intelligence systems to:

  • Address the Frame Problem:
    • Define “perception-decision-action” patterns or modules, such as:
      • What to do when it’s too dark? 🔥
      • What to do when too thirsty? 💧
  • Address Gestalt Psychology: Provide cognitive shortcuts and heuristics for “mental fill-ins,” such as: * What kind of “mental fill-in” is 🍏🤤 “slaking thirst by thinking of plums”? * —— Triggering physiological responses through imagination and memory to compensate for a lack of current resources. * What kind of “mental fill-in” is 🧨🔥 “a single spark starting a prairie fire”? * —— Inferring an imminent massive disruption from minute signs. * What kind of “mental fill-in” is 🌬️👤 “catching at shadows and chasing the wind”? * —— Constructing elaborate plots and causal relationships out of thin air when information is scarce.
  • Address the Symbol Grounding Problem:
    • Plan the system’s “information needs and resource allocation” based on response time scales, such as:
      • Why and how should combustible resources be allocated to support the “water and fire” mechanism that sparked the 🚂💨 Industrial Revolution?
      • For 🌡️☁️ hyperscale data centers, “water and fire” refers to the balance between power consumption (fire) and cooling systems (water). How can a data center balance computing power (fire) with cooling (water) needs to plan its extreme-scale deployment of computing resources and energy consumption, ensuring stability and sustainability?
      • For 🏭🤖 Dark Factories, “water and fire” denotes the synergy between automated operations (flowing continuously like water) and energy supply (fire) to achieve a highly autonomous, human-free closed-loop production system.
      • Looking ahead to 🌐 Embodied AI and Smart Cities, “water and fire” will extend to the integration of the energy grid (fire) and social infrastructure (water representing circular resources). How should these resources be collaboratively allocated?
  • Address the AI Alignment & Control Problem:
    • Design ethical or behavioral constraints to maintain controllability and predictability.
      • 🥸🤖 Can an AI State Grandmaster truly audit a nation’s energy mix, control \(CO_{2}\) emissions, and alter the country’s destiny?
      • 🥳👨‍👩‍👧‍👦 Can an AI Matchmaker truly transform lives or even alter the birth rate of a nation or a community?

When addressing these AI Problematics, issues regarding temporal and spatial scales, as well as human socio-technical scaling, naturally surface.

While human ❝mental fill-in❞ capacity is a useful “cognitive shortcut” that can be polished through experience and learning, the foundational issues of artificial intelligence—such as the Frame Problem, Gestalt Psychology, and the Symbol Grounding Problem—constitute the empirical and engineering discipline focused on understanding and simulating these human “fill-in” experiences.

🦾⚡ Case in Point: ❝Mental Fill-ins❞ and VLA Robots

In 2026, a “two-camp” rivalry emerged in the robotics and embodied-AI industry, and it happens to offer a living, technical proof of the ❝mental fill-in❞ concept — the industry’s route dispute turns out to be, at its core, two modern forms of this same old mental habit.

One camp champions Vision-Language-Action (VLA) models: the robot need not “understand” the world first — it only needs to learn, from massive volumes of demonstration data, the mapping and correlation between “seeing a situation” and “outputting an action” directly. This is the machine version of ❝mental fill-in❞ — it never asks “why,” only “how,” relying on trained “routines” to rapidly fill the gap between perception and action. It is scalable, good enough, fast, and cheap, and extremely effective in familiar situations (Zitkovich et al. 2023; Black et al. 2024).

The other camp champions the World Model: before a machine acts, it should first build an internal model capable of simulating “if I do this, how will the world change?” — this is closer to the deliberate foresight of the human “reflective-symbolic” mind than to a bare ❝mental fill-in❞ shortcut. Leading figures include Yann LeCun (LeCun 2022; Assran et al. 2023) and Fei-Fei Li (Gupta et al. 2021; Krishna et al. 2022; World Labs 2026). This camp holds that VLA’s “look-and-act” mapping is precisely the kind of shortcut that breaks down outside its training distribution — a difficulty echoed in broader systematic reviews of embodied AI (Liu et al. 2025) — and that only a world model built for internal simulation can extend reliably into genuinely novel situations.

Note 1: 🦾⚡🙶 ❝Mental Fill-in❞ vs. VLA, 🙶Constructive Fill-in🙷 vs. World Model

  • VLA ≈ the engineering of ❝mental fill-in❞: like a “routine” — fast, cheap, effective in familiar situations, but prone to “missing the forest for the trees,” or even locking in a single, confidently wrong read, once it meets a situation outside its training distribution.

  • World Model ≈ the engineering of 🙶constructive fill-in🙷: like “technological foresight” — it builds a model capable of internal simulation before acting. This costs more, but copes far better with the Frame Problem’s core difficulty: situations where information is insufficient or falls outside prior experience. Notable research includes Yann LeCun (LeCun 2022; Assran et al. 2023) whose Joint Embedding Predictive Architecture, JEPA, draws on cybernetics and cognitive science rather than image generation, and Fei-Fei Li (Gupta et al. 2021; Krishna et al. 2022; World Labs 2026) whose long-standing research on embodied intelligence and spatial understanding underlies this camp’s outlook

This maps neatly onto the contrast drawn earlier in “🖼️ Addressing AI Problematics ㉄”: the human ❝mental fill-in❞ failure mode of “excluding too fast” corresponds to VLA’s “strategy lock-in”; the AI Frame Problem’s failure mode of “excluding too slowly” corresponds to the World Model’s pursuit of “exhaustive simulation,” which can come at the cost of “losing your bearings.”

📊 VLA vs. World Model: Side-by-Side

Route Characteristic Leading Figures Representative Work Strengths Risks
VLA (fill-in style) Perception → action, directly linked Kim et al.(Kim et al. 2025); Zitkovich et al. (Zitkovich et al. 2023); Black et al. (Black et al. 2024) OpenVLA, RT-2, π0 series Fast, cheap, effective in familiar settings Prone to hallucination, poor generalization
World Model (constructive style) Internal simulation → then action LeCun (LeCun 2022; Assran et al. 2023); Li Fei-Fei (Gupta et al. 2021; Krishna et al. 2022; World Labs 2026) JEPA, World Labs Handles unfamiliar situations, capable of reasoning High cost, slow to deploy

🧠 Echoes of the Threefold Mind

  • 🐸⚡ The Reactive Mind → corresponds to VLA’s “rapid mapping,” emphasizing instant reflex and routinized action.

  • 🐘💞 The Affective-Bonding Mind → in human-machine interaction, VLA’s “mental fill-in” can invite situational dependence and misreading.

  • 🧘☸️ The Reflective-Symbolic Mind → corresponds to the World Model’s “constructive simulation,” emphasizing symbolic reasoning, planning, and ethics.

This alignment shows that the 2026 AI industry’s multi-hundred-billion-dollar capital bet on this route dispute is not a brand-new philosophical question — it is the same ❝mental fill-in❞ vs. 🙶constructive fill-in🙷 debate that opened this book’s first chapter, restaged in a new technical vocabulary.

㉄ A Problem Set or a Discipline? 🤖

Artificial Intelligence (AI) is an independent discipline, yet it also resembles a collection of problems that attract cross-disciplinary efforts—much like Nuclear Engineering—requiring fresh insights from various fields. Although AI is viewed as an independent field within electrical engineering and computer science, AI Engineering centers strictly on industry real-world implementation.

Category Field Core Focus Application Focus
Application Field Nuclear Engineering Applications of nuclear fission and fusion processes, commonly used for power generation.
(💡 Focuses on high-risk application domains.)		 Problems are real-world systems (e.g., designing safe reactors, managing waste, ensuring non-proliferation), requiring expertise in physics, chemistry, materials science, mechanical engineering, and international policy.
Foundational Discipline Artificial Intelligence (AI) Building systems capable of performing tasks that typically require human intelligence.
(📚 Focuses on the universality of theories and algorithms, transcending specific application fields.)		 Problems are broad real-world applications (e.g., medical diagnosis, autonomous vehicles, language processing, strategic stability), requiring expertise in computer science, ethics, cognitive science, political science, and domain-specific knowledge (such as medicine or law).
Applied Discipline AI Engineering Designing, developing, and deploying AI systems to solve practical problems in a reliable, scalable, and ethical manner.
(🛠️ Focuses on engineering and operationalization challenges from lab to market.)		 Problems lie in operationalizing AI (e.g., model version control, MLOps, robust evaluation, latency management, ensuring data quality), combining principles of computer science, software engineering, and machine learning.

In both academia and industry, AI is simultaneously:

  1. A foundational discipline focused on theory.
  2. 🧭 A high-risk, multidisciplinary application arena spanning a cross-disciplinary problem set, similar to nuclear engineering. This highlights that solving AI’s critical challenges (cognitive capacity, safety, ethics, regulation) must be a multidisciplinary effort.
  3. 🏗️ The emergence of “AI Engineering” reinforces this “cross-disciplinary problem set” view, but shifts the focus to system reliability and operationalization, requiring a concurrent understanding of how ML models function (data science) and the operational challenges of large-scale software systems (software engineering).

Therefore, this book initiates its journey with 01 ㉄ AI Problematics.

🌌 Deconstructing the Book 📑

Before diving into the chapters, this book provides a “deconstruction” guide to help readers understand how to flexibly utilize the entire text.

“Deconstruction” carries a dual meaning here: on one hand, the book serves as a knowledge map that can be oriented and navigated via longitudes and latitudes (the book’s 🌌 Mindmap «AI Knowledge Scaffolding»); on the other hand, the book functions as a reassemblable toolbox. Readers can dismantle and recombine it according to their needs to form their own personal scaffolding for knowledge and action (The 💪 Action, 🧠 Brain, and 🪜 Cognitive Capacity «Knowledge-Action Scaffolding» Practice Manual).

The following “longitudes and latitudes” offer distinct reading and application perspectives.

Important 2: 🌐🧭 Navigating the Book’s Longitudes and Latitudes

The 10 parts and individual chapters of this book are perfectly suited for readers to dismantle and apply.

🧭 Longitudes and Latitudes 🔗

To quickly master the primary knowledge points of this book, you can start with the 🌌 Mindmap and map out your coordinates along the vertical axis (the «data-driven decision» process) and the horizontal axis (various problematics, schools, paradigms, orientations, etc.):

  • X-Axis (Longitude) ~ Illustrates the «data-driven decision» process, detailed in 《06 ❖ Analytics & Decisions》📍?fig-mindmap-06, mapped to ✍️💡 Bloom’s 6 Cognitive Capacities:
    • 🤓📘 From the vast world, perform 『Descriptive Analysis』 to establish a context-bound «Memory» of the present.
    • 😷🩺 Within relevant memories, perform 『Diagnostic Analysis』 to correct fallacies through factor-related «Understanding».
    • 🤠🔮 Among relevant factors, perform 『Predictive Analysis』 to aid the «Application» of model-based knowledge.
    • 🧐🧭 Within knowledge models, perform 『Prescriptive Analysis』 to gauge the «Evaluation» and «Creation» of system intelligence.
    • Alongside these 4 core pillars of analytics, there are innovative divergent and convergent auxiliary paths:
      • 🙀🎨 Utilizing collective large models, leverage 『Generative AI』 to generate a diverse, divergent pool of possible «Options».
      • 😽🪄 Out of these divergent options, utilize 『Decision-making Algorithms』 to achieve a convergent, autonomous selection for the ultimate «Choice».
  • Y-Axis (Latitude) ~ Outlines various problem sets and schools across multiple entry chapters:

The remaining 6 parts cover the following contents:

In short, the book’s 🌌 Mindmap and 📖 Table of Contents serve as an initial draft of the «AI Knowledge Scaffolding» designed by the author to help readers rapidly grasp their learning direction.

Important 3: 🧭 Mindmap Navigation and Interactive Features

The book’s 🌌 Mindmap condenses complex knowledge into a neat grid. It not only 🧭 navigates the process of intelligent generation but also helps distinguish different technologies and application scenarios:

  • Each knowledge point node box contains embedded 🔗 hyperlinks, allowing readers to jump, compare, and cross-reference instantly.
  • The connecting lines define relationships between knowledge points.
  • Select entries also incorporate timelines for historical context.

🎍 Divergence and Convergence 🪕

The goal of orienting along longitudes and latitudes is to transform AI knowledge into innovative action: how do we break down knowledge, diverge our thinking, and converge on innovative ideas given limited resources?

  • 🔗 Explore Conceptual Connections: By browsing the 🌌 Mindmap (where each node is a clickable hyperlink) and utilizing the hyperlinks throughout the text, readers can jump between sections, compare concepts, and build a cross-chapter knowledge network.
  • 📚 Study Whole Systems: Follow the overviews of the ten parts, browsing them according to the longitudes and latitudes of the mindmap to gradually differentiate and integrate problem sets, schools, paradigms, orientations, and applications.
  • 🛠️ Create Intelligent Systems: Readers can construct their own personalized “Knowledge-Action Scaffolding” 🟰 Verbs ➕ Nouns. Begin by listing “Actions” (verbs: what to do, what can be done, what need not be done), then categorize “Knowledge” (nouns: what must be known, what can be known, what is temporarily unknown). Weave these into matrices to initiate your ❝constructive fill-ins❞ and build your capabilities step by step.

🌱 In short, the mission of this section is to help readers fuse “navigation” with “practice”: - Longitudes & Latitudes → Assist with orientation; - Dismantling, Divergence, & Convergence → Drive action.

Equipped with these, readers will carry both a knowledge map and a practice manual.

🪜 From ❝Mental Fill-ins❞ to 🙶Constructive Fill-ins🙷 🌌

This introduction explains the concept of ❝mental fill-ins❞ and AI, and demonstrates how to navigate this toolbook. Readers can quickly find information using search functions and mindmap browsing; the book’s systematic arrangement provides distinct indexical and reference value.

The next section, 📑 Notes 《❝Constructive Fill-ins❞: Cognitive Unity of Knowledge and Action》, shifts focus to the practice of ❝constructive fill-ins.❞ Once readers deploy their own «Knowledge-Action Scaffolding» and map out matrices based on processed «Options» and «Choices», they will begin constructing their own operational knowledge systems, expanding the boundaries of action and knowledge. That next chapter will also take this chapter’s personal, individual-mind contrast of “❝Mental Fill-in❞ vs. VLA, 🙶Constructive Fill-in🙷 vs. World Model” and scale it up to the organizational and societal level, showing how “framing intelligence” echoes both management theory’s “dynamic capabilities” and political science’s “cybernetic governance.”

If readers are willing to invest more effort, they will find that this book is a highly interconnected system: its structure—consisting of one introduction, one set of notes, and multiple appendices—helps readers explore methodology deeply. The associated appendices on 💪 Action, 🧠 Brain, and 🪜 Cognitive Capacity serve as the author’s carefully curated practice manual for your personal «Knowledge-Action Scaffolding»!

Tip 1: 💬 Cognitive Shortcuts Tip: 🌟 TLDR

🌟 TLDR (Too Long; Didn’t Read): This book is your knowledge map and action guide for the AI era.

Mode Objective Critical Operations
🔍 Quick Retrieval Find isolated pieces of information or definitions. Use the 🔍 search function and browse the 📚 left sidebar for rapid scanning.
🔗 Conceptual Linkage Establish cross-domain knowledge networks. Use the clickable nodes on the 🌌 Mindmap to perform quick comparisons and cross-referencing.
📚 Systematic Learning Master AI schools, paradigms, problems, and applications. Follow the ten parts and systematically differentiate and integrate concepts using the mindmap’s 🧭 coordinates.
🏗️ Practical Action Construct a personalized Knowledge-Action Scaffolding for AI systems. List your “Actions” (verbs) and “Knowledge” (nouns), weave them into custom matrices, and systematically execute ❝constructive fill-ins❞.
🦾 Frontier Verification Verify the framework against the 2026 industry route dispute. Contrast VLA (the engineering of ❝mental fill-in❞) with the World Model (the engineering of 🙶constructive fill-in🙷).

Ultimately, this book aims to help readers transform the human instinct for ❝mental fill-ins❞ into a systematic, deliberate practice of ❝constructive fill-ins❞ for innovation.

🦾💪🔥💧