Skip to content

The Deep Layers of the Readable Code Debate: Human Anxiety and Values Behind Technical Arguments [2025 Edition]

What You'll Understand from This Article

Three human factors behind technical arguments The fundamental fear of black boxes Generational anxiety about skill value The complexity of intertwined values

"Does code written by AI need to be readable by humans?"—On the surface, this is a technical debate. But deeper down lie fears of losing control, anxiety about accumulated skills, and clashing values. This article unravels the controversy from a multi-layered human perspective.

Target Audience

  • Those who feel something off about the intensity of the debate
  • Developers who want to sort out their own anxieties and positions
  • Those who sense 'something' that can't be explained by technical arguments alone

Key Points

  1. Understanding the human background behind why the debate gets so heated
  2. Ability to objectively view your own anxieties and values
  3. Perspective to make judgments while respecting diverse viewpoints

Perspective of This Article

The purpose is not to judge technical correctness, but to understand the "human feelings" behind the controversy.

For Those Who Want Practical Guidance

For specific decision criteria and checklists for projects, see the companion article "AI Era 'Readable Code Unnecessary Theory' Practical Decision Guide".

The Core: Surface is Technical, Depths are Human Anxiety

Observation: Why Does This Get So Heated?

This debate has characteristics different from typical technical discussions.

Observed FactImplication
Sometimes becomes personal attacksNon-technical factors involved?
Extreme polarization (complete acceptance vs complete rejection)Difficult to find middle ground?
Tendency to doubt opponents' intentionsNot purely technical debate?
Opinions split along generational linesExperience and position influence views?

Hypothesis: Various human factors are hiding behind technical arguments

"Quality and Risk Technical Debate"

  • Effectiveness of readable code
  • AI reliability assessment
  • Maintainability and cost

"Complex Tangle of Various Anxieties and Values"

  • Fundamental fear of black boxes
  • Anxiety about skill value
  • Generational value differences
  • Psychological resistance to losing control

Human Factor 1: Fundamental Anxiety About Black Boxes

Psychological Resistance to "Can We Trust What We Don't Understand?"

Humans have an instinctive resistance to trusting what they cannot understand.

"What You Can't See Inside Is Scary"

  • Don't know what AI is doing
  • Feeling of lack of control
  • Fear of "being helpless when something goes wrong"
  • Nightmares of past obfuscated code

Fear of Losing Control

Understandable = Controllable = Safe
Not understandable = Uncontrollable = Dangerous

This equation forms the psychological foundation for many people

Specific Fear Scenarios
  • "What if AI randomly writes rm -rf /?"
  • "What if security holes are embedded?"
  • "What if no one can fix bugs?"
  • "What if AI service ends and everything needs rewriting?"

However: PCs Are Also Black Boxes But Are Trusted

Question: Can you explain how a transistor works?

Most people cannot. Yet they use PCs.

LayerPeople Who UnderstandTrusted
Transistor physics≈0.01%✅ Nobody cares
Memory cell operation≈0.1%✅ Nobody cares
OS kernel implementation≈1%✅ Nobody cares
Cloud physical configuration≈5%✅ Nobody cares

In other words: Not understanding and trusting have coexisted

Insight

The essence of trust is not "understanding" but "predictability" and "track record". PCs are not understood, but trusted through the predictability of "move mouse, screen moves" and 50 years of track record.

Why Is Anxiety About AI Special?

Deterministic vs Probabilistic

  • PC: Same input → Always same output
  • AI: Same input → Different outputs possible

Difference in Track Record

  • PC: 50 years of stable operation
  • AI: Only about 5 years, examples of "unintended behavior" exist

Most Dangerous Is "Half-Baked Trust"

  • Over-reliance on autonomous driving → Accidents
  • Blind faith in AI diagnosis → Misdiagnosis
  • Unvalidated deployment of AI code → System failures

Dangerous to depend before trust is established

Current Position

The problem with AI is not that it's "incomprehensible", but that "predictability and track record are still insufficient". Feeling anxious during this transition is extremely rational.

Human Factor 2: Anxiety About Skill Value

First Generation: Fear That Accumulated Skills Become Worthless

Attachment to Skills Cultivated Over Years

  • Polished code for 10, 20 years
  • Pride in writing beautiful code
  • Joy of understanding algorithms
  • Identity as a craftsperson

"Feeling That One's Value Is Being Denied"

  • Humiliation of "losing to AI copy-paste"
  • Sense of loss: "What have I been doing all this time?"
  • Impatience of "being overtaken by juniors"
  • Feeling of helplessness: "Experience is useless"

Superficially spoken as technical arguments:

  • "Quality will decline"
  • "Long-term maintainability is important"
  • "Won't work in enterprise"
  • "Can't apply without knowing fundamentals"

But behind may lie attachment to skills and anxiety

Real Voices (One Interpretation)

"I spent years learning readable code. Being told it's 'unnecessary' feels like denying all my efforts"

Third Generation: Vague Anxiety About Skipping Fundamentals

Experience of Entry via New Tools

  • Could create without learning programming basics
  • Could solve problems even if "bad at coding"
  • Drastically reduced learning costs
  • Joy of quickly shaping "what you want to make"

Vague Anxiety of "Is This Really Okay?"

  • "Is it okay not to know the basics?"
  • "Will I be helpless without AI?"
  • "Am I really becoming a proper engineer?"
  • But learning basics now is...

Superficially spoken as efficiency arguments:

  • "No need to read, just needs to work"
  • "Don't want to be bound by old conventions"
  • "AI just needs to read it"
  • "Speed is important"

But behind may lie anxiety about "lacking basics"

Generational Mutual Distrust (One Interpretation)

First Generation's View of ThirdThird Generation's View of First
"Without even knowing basics...""Clinging to old rules..."
"Disregarding quality""Don't understand new tools"
"Being complacent""Afraid of change"
"Will be helpless when it matters""Just trying to protect skill value"

Structure of Mutual Distrust

Both sides only react to each other's "surface arguments", not "anxieties". Result: heated debates and personal attacks.

Human Factor 3: Complex Tangle of Various Values

This debate is not a simple binary opposition.

Technical Arguments, Risk Arguments, and Human Aspects Are "Complexly" Intertwined

  • Effectiveness of readable code
  • AI accuracy assessment
  • Maintainability vs development speed
  • Risks of AI dependency
  • Risks of human costs
  • Optimism vs caution
  • Anxiety about skill value
  • Fear of black boxes
  • Generational value differences
  • Resistance to losing control

All these are complexly intertwined and cannot be simplified

Contradictory Emotions Within the Same Person

Real Internal Conflicts

First Generation's Interior (One Example):

  • Intellectually understands "AI is convenient"
  • But emotionally "code I can't understand is scary"
  • Competitive spirit of "don't want to lose to juniors"
  • Simultaneously curiosity to "learn new tools"
  • Sense of mission to protect "quality"
  • But doesn't want to be seen as "outdated"

Third Generation's Interior (One Example):

  • Claims "basics are unnecessary"
  • But anxious inside: "Is this really okay?"
  • Champions "efficiency focus"
  • Simultaneously desires "to become a proper engineer"
  • Criticizes "old rules"
  • But secretly also "respects" them

Insight

Debates get heated because people hold "contradictory emotions" while discussing on the surface as technical arguments. No space to speak true feelings.

History of Black-Boxing: Four Layers

Having understood human anxieties, let's look at historical patterns.

Layer 1: Hardware (1940s-1970s)

Anxiety ThenSolutionNow
"Electric circuits randomly delete data?"ECC memory, redundancy, massive testingNobody cares
"Bit flip errors are scary"Parity check, error correctionTrusted

Layer 2: OS (1970s-1990s)

Anxiety ThenSolutionNow
"If OS bugs, everything crashes?"Memory protection, process isolationNobody cares
"Kernel panics are scary"Permission management, safety improvementsTrusted

Layer 3: Cloud (2000s-2020s)

Anxiety ThenSolutionNow
"AWS randomly deletes data?"SLA, audit trailsNobody cares
"Vendor lock-in is scary"Standardization, certification systemsMany trust

Layer 4: AI (2020s-20XXs) ← We are here

Current AnxietyPath to SolutionFuture
"AI writes destructive code?"Guardrails under construction
"Unpredictable behavior is scary"Safety evaluation standardization ongoing

Historical Pattern

Same anxieties repeated each generation. Eventually, trust built through technical guarantees, track records, and social systems. AI likely follows same path.

The Essence of Trust: Understanding vs Predictability

PC Trust Structure (Achieved)

FactorStatusBasis of Trust
Understanding❌ Almost nobody understands-
Predictability✅ Deterministic behaviorMove mouse, screen moves
Track Record✅ 50 years stable operation"Random deletion" almost never
Constraints✅ No unintended behaviorFile deletion etc. doesn't happen

AI Trust Structure (Under Construction)

FactorStatusNeeded Evolution
Understanding❌ Interior invisible(Unnecessary - black box OK)
Predictability⚠️ ProbabilisticStrengthen guardrails
Track Record⏳ Only 5 yearsBuild up over time
Constraints⏳ Under constructionEstablish "won't destroy" guarantee

Core Insight

The essence of trust is not "understanding" but "predictability". Human anxiety comes not from "not understanding" but from "unpredictable and unconstrained".

Technical Supplement: AI Also Assumes Readable Code

An important technical fact in the debate: Current AI (GitHub Copilot, Claude, ChatGPT, etc.) is trained on readable code.

AI Training Data CharacteristicsMeaning
GitHub public repositoriesMostly code following readable code principles
Open source projectsReviewed, high-quality code
Documented codeRich comments and explanations

Result: When handling non-readable code, AI performance likely drops.

Correcting Misconception

"AI optimization" doesn't mean "ignoring humans", but "structures AI can understand easily". Ironic that for current AI, readable code is optimal.

Timeline Scenarios: When Will Anxieties Be Resolved?

2025 (Present): Transitional Anxiety at Maximum

AI Imperfect, Anxiety Rational

  • AI accuracy 80-90% (remaining 10-20% can be fatal)
  • Track record still limited
  • Social systems undeveloped
  • Predictability insufficient
  • Fear of black boxes: ★★★★★
  • Anxiety about skill value: ★★★★★
  • Generational conflict: ★★★★☆

Feeling anxious is extremely rational. Readable code has clear value

2030-2040 (Mid-term): Anxiety Gradually Easing

AI Accuracy Improves, Track Record Accumulates

  • AI accuracy improves to 95%+
  • 10-20 years track record
  • Social systems begin to develop
  • Guardrails standardized
  • Fear of black boxes: ★★★☆☆
  • Anxiety about skill value: ★★★☆☆ (adaptation progressing)
  • Generational conflict: ★★☆☆☆ (new generation becomes mainstream)

Tier system standardized. Only critical parts reviewed by humans

2040+ (Long-term): Anxiety Disappears?

AGI Achieved, Society Accepts

  • AI reliability far exceeds humans
  • 50+ years track record
  • Legal framework complete
  • "Won't do destructive behavior" guarantee established
  • Fear of black boxes: ★☆☆☆☆ (don't care like with PCs)
  • Anxiety about skill value: ❓ (migrated to new skill sets)
  • Generational conflict: ☆☆☆☆☆ (old generation retired)

Major Incident Occurs, Regulations Strengthen

  • Large-scale failure due to AI defects
  • "Human comprehensibility" becomes legal requirement
  • Readable code re-evaluated
  • Hybrid period continues long-term

Uncertainty

Which scenario occurs depends on future AI evolution and society's choices. Need to prepare for both.

Integrated Perspective: Understanding Various People's Feelings

Inner Thoughts from Each Position (One Interpretation)

"Want to Protect Quality"

  • Surface: Technical quality arguments
  • Deep: Attachment and anxiety about accumulated skills
  • True feelings: Change is scary, but want to adapt
  • Wish: Want places where experience can be utilized

"Want to Pursue Efficiency"

  • Surface: Speed-focused efficiency arguments
  • Deep: Vague anxiety about lacking basics
  • True feelings: Actually want to learn properly
  • Wish: Want new approaches recognized

"Want to Find Balance"

  • Surface: Pragmatism
  • Deep: Understand both sides
  • True feelings: Conflict of not being able to side with either
  • Wish: Want to bridge generations

Mutual Understanding

Behind each argument lie various feelings and anxieties. This is why debating only on technical grounds leads to parallel lines.

Summary: Understanding the Complex Debate Multi-Dimensionally

  • Surface is technical, depths are human anxieties: Fear of black boxes, anxiety about skill value, clash of values
  • Essence of trust is predictability: PCs aren't understood but are trusted. AI likely follows same path
  • Transitional anxiety is rational: Feeling anxious about AI now is extremely rational. May ease over time
  • Generational mutual distrust: Important to understand feelings behind surface arguments, not just the arguments
  • Complexity that can't be simplified: Technical, risk, and human aspects complexly intertwined, no single correct answer

Realistic Conclusion

Rather than extremes (completely unnecessary vs absolutely essential), "necessary but priority and scope change" is a realistic middle approach. Leave automatable parts to tools, focus on parts requiring human judgment (WHY comments, design decisions). Since AI is trained on readable code, "AI optimization" doesn't mean "ignoring humans".

Constructive Attitude

Viewing the debate not as "who's right" but "what are each side's anxieties and feelings" can generate new dialogue. Precisely because it's a transition period, wisdom to respect diverse perspectives and respond flexibly is needed.

Next Steps

  • Sort out your own anxieties and values (nothing to be ashamed of)
  • Imagine the "feelings" behind others' arguments
  • Promote intergenerational dialogue (not just technical, but true feelings too)
  • Make decisions with time horizon in mind (current anxieties aren't eternal)

For Those Who Want Practical Project Decisions: The companion article "AI Era 'Readable Code Unnecessary Theory' Practical Decision Guide" provides checklists, layer-based strategies, and 6 practical guidelines usable in the field.


This article is an analysis as of October 2025. The situation will change as AI technology and society evolve.