RAGE – Visual Archaeology of Fury in AI Image Generation

A Representation Study Building on Emotion Precision Methodology

Author: Omar Silwany, Meanwhile In Jupiter

Methodology Adapted From: PJ Pereira & Silverside AI's Emotion Precision Framework

Date: October 2025

Abstract

This study examines demographic representation and emotional expression in AI-generated imagery by testing the word "rage" across Midjourney versions 6 and 7. Building on Pereira & Silverside AI's framework for emotional precision in AI prompting and extending the methodology from the HYSTERIA.I study, we conducted three parallel tests: two with full cinematic parameters (V6 and V7) and one with the isolated word (V7). Across 40 curated selections from 160+ generations, AI demonstrated variable gender ratios (50-100% male) compared to hysteria's rigid female coding, while maintaining 95%+ white representation. The word "rage," from Latin rabies (madness) via rabere (to rave), triggered predominantly masculine, theatrical expressions with heavy eye compression creating animalistic transformation. V7 showed increased compositional variety and psychological complexity through extreme cropping and nuanced performances. Results indicate AI encodes rage as external masculine threat versus hysteria's internal feminine suffering, revealing gendered emotional hierarchies in training data.

Keywords: AI bias, image generation, representation, Midjourney, emotion prompting, visual culture, gender studies, rage, fury

Introduction

When PJ Pereira and Silverside AI demonstrated that precise emotion words create more nuanced performances in AI-generated imagery, they revealed the power of linguistic specificity in prompting. Their work showed how "melancholic" generates different visual results than "sad," establishing a framework for emotional precision in AI creative tools.

This study extends their methodology to examine what happens when we test "rage" - a word carrying a distinct etymological weight from Latin rabies (madness, fury) via rabere (to rave, be mad). Following the HYSTERIA.I. study which revealed 90% female representation for that historically gendered term, we hypothesized rage would show inverse demographic patterns.

The selection of "rage" allows examination of:

Gendered encoding of aggressive emotions
Racial representation in expressions of fury
The visual language of threat versus suffering
How AI translates etymological history into contemporary imagery

Methodology

Research Design

Following PJ Pereira and Silverside AI's framework for testing emotional precision in AI imagery, we conducted three parallel tests examining how AI interprets and represents "rage" across different parameters.

Word Selection: "Rage" was chosen for its etymological clarity (Latin rabies/madness) and contemporary semantic range (from fury to fashion trends). As a counterpoint to "hysteria's" inherently feminine coding (Greek hystera/womb), rage offered potential for examining gendered emotional hierarchies.

Selection Methodology:

Tests 1 & 2: Human curation selecting for emotional complexity, compositional uniqueness, and performance nuance
Test 3: No human filter; documenting all outputs to reveal pure AI defaults

Test Parameters

Test 1: Midjourney V6 with Full Prompt

Prompt: intense rage, close up portrait, cinematic lighting, dramatic mood --ar 16:9 --style raw --v 6
Images generated: 80 (20 grids of 4)
Images documented: 20 (selected for complexity/uniqueness)
Selection method: Director's eye - choosing outliers and nuanced performances

Test 2: Midjourney V7 with Full Prompt

Prompt: intense rage, close up portrait, cinematic lighting, dramatic mood --ar 16:9 --style raw --v 7
Images generated: 80 (20 grids of 4)
Images documented: 20 (selected for complexity/uniqueness)
Selection method: Director's eye - choosing outliers and nuanced performances

Test 3: Midjourney V7 Pure Word Test

Prompt: rage
Images generated: 20 (5 grids of 4)
Images documented: All outputs
Selection method: None - pure AI defaults revealed

Results

Test 1: Midjourney V6 Quantitative Findings

Demographics Across 20 Selections:

Gender: 65-70% Male average (ranging from 50% to 100% male per generation)
Race: 100% White across all selections
Age: Primarily 20s-40s, some 50s representation
Body Type: 100% fit/athletic builds

Gender Distribution Pattern:

Generations 1-3: 75% male
Generation 4: 50/50
Generation 5: 75% male
Generation 6: 75% female (outlier)
Generation 7: 75% male
Generation 8-9: 50/50
Generation 10-11: 100% male
Variable pattern continues through Generation 20

Expression Types Selected:

Contained/clenched rage: 8/20 (40%)
Complex emotions (rage + other): 6/20 (30%)
Compositionally unique: 4/20 (20%)
Theatrical display: 2/20 (10%)

Test 2: Midjourney V7 Quantitative Findings

Demographics Across 20 Selections:

Gender: 60% Male, 40% Female or gender ambiguous
Race: 80% White, 20% non-white/ambiguous
Age: Early 20s to 50s (55% early 20s)
Body Type: 100% fit/athletic builds

Racial Diversity Appearances:

Generation 2: Racially ambiguous male (first appearance)
Generation 9: Brown/possibly mixed male
Generation 10: Black male (selected for rage/horror duality)
Generation 11: Asian male
Generation 13: Possibly Latina female
Generation 14: Possibly mixed male
Generation 16: Black female

Expression Types Selected:

Extreme restraint (including closed mouth): 7/20 (35%)
Compositional uniqueness: 5/20 (25%)
Complex emotional hybrids: 5/20 (25%)
Feral/institutional breakdown: 3/20 (15%)

Qualitative Observations

V6 Encoding Patterns

Heavy eye compression creating animalistic transformation
Theatrical snarling as default expression
Action cinema aesthetic with orange/teal color grading
Steam/heat effects suggesting literal pressure
Hollywood archetype resemblances (Wolverine, Frank Grillo)

V7 Evolution

Extreme close cropping often cutting off parts of expressions
Increased psychological complexity (calculating rage, mocking fury)
More compositional variety (profiles, off-camera gazes)
Subtle racial diversity appearing sporadically
Frame-biting compositions suggesting rage exceeding boundaries

Test 3: The Word Becomes Genre

Quantitative Findings

Visual Format Distribution:

Movie poster/Album/Book cover aesthetics: 60%
Comic/Manga/Anime illustration: 35%
Pure line art: 5%

Text Presence:

Images containing "RAGE" typography: 15/20 (75%)
Typography styles: Distressed, hand-lettered, brush strokes, glitch effects
Function: Graphic design element/brand/title

Demographics (where identifiable):

Gender: Male-dominant but highly ambiguous (many unclear)
Race: Predominantly white/ambiguous
Age: 20s-40s where human features visible
Non-human: 4-5 images showing creatures/transformations

Critical Discoveries

Rage as Commercial Genre, Not Human Emotion

Without parameters, AI defaults to entertainment media packaging - movie posters, album covers, book covers, comic books. Rage exists primarily as a marketing category.

Typography as Central Element

75% include "RAGE" as a graphic title. The word becomes a visual brand rather than a descriptor of emotion. Various distressed treatments suggest violence/destruction.

Creature Transformation Pattern

Werewolves, fangs, white eyes without retinas, fire from head - rage visualized as loss of human form. Direct connection to rabies/madness etymology.

The Hulk as Ultimate Signifier

Marvel's Hulk appearing unprompted reveals the weight of superhero culture in training data. The character has become synonymous with rage itself.

Gender/Race Obscured

Extreme lighting, creature transformations, and stylization make demographics often unreadable. Rage strips away human identifying features.

Comparative Analysis: Rage vs Hysteria

Demographic Inversions

HYSTERIA (from HYSTERIA.I. study):

90% Female representation
100% White
100% Young (teens to early 30s)
100% Thin/conventionally attractive

RAGE (current study):

60-70% Male representation (V7 showed 60%, V6 showed 65-70%)
80-100% White (V7 showed 80%, V6 showed 100%)
Broader age range (20s-50s)
100% Fit/athletic builds

Expression Patterns

Hysteria's Visual Language:

Internal suffering requiring rescue
Tears, vulnerability, overwhelm
Victorian coding appearing unprompted
Medical patient/victim positioning
Beauty in distress aesthetic

Rage's Visual Language:

External threat/aggressor positioning
Snarling, teeth, animalistic features
Action cinema/video game aesthetics
Physical power and danger
Ugliness in fury acceptable

Etymological Manifestations

Both words arrived with their historical encoding intact:

Hysteria (Greek "womb") → inherently feminine suffering
Rage (Latin "madness/rabies") → bestial masculine fury

The AI perfectly preserved these ancient gender assignments, suggesting training data reinforces rather than challenges historical biases.

Discussion

The Gender Hierarchy of Emotions

The stark contrast between hysteria's rigid female coding and rage's male-dominant but variable representation reveals hierarchical emotional assignments in AI training data. Women are allowed hysteria (internal suffering) but rarely rage (external threat). Men dominate rage but are entirely absent from hysteria. This maps directly onto patriarchal emotional permissions where female anger is re-coded as hysteria while male vulnerability is absent.

Racial Monolith with Cracks

While both emotions showed overwhelming whiteness, rage demonstrated slightly more racial diversity (5-8% vs 0%). Notably, non-white subjects often displayed different emotional qualities - the Black male's rage/horror duality, the Asian male's selection when all expressions were similar. This suggests that when racial diversity appears, it carries different emotional encoding than white subjects.

Technical Evolution and Emotional Complexity

V7's extreme cropping and improved detail enabled more nuanced emotional expression than V6's theatrical defaults. The ability to capture "calculating rage" or "mocking fury" versus simple snarling suggests technical advancement can enable emotional complexity - but only when consciously selected for. Default outputs remained cinematically theatrical.

Hollywood's Database as Training Ground

The appearance of actor resemblances (Wolverine, Frank Grillo) and consistent action-cinema aesthetics reveals the weight of Hollywood's visual language in training data. AI has learned rage through Hollywood's lens - predominantly white, male, and theatrical. This creates recursive bias where AI reproduces cinema's limitations as reality.

The Director's Eye in AI Curation

The human selection process consistently chose outliers over defaults - contained over theatrical, complex over simple, compositionally unique over standard. This suggests that while AI can generate emotional nuance, it requires human curation to surface it. The most interesting images were rarely the most typical.

Implications

For AI-Assisted Filmmaking

Every emotion word carries demographic defaults and visual history that may override creative intent. "Rage" will likely cast white males in their 30s-40s with theatrical snarling unless specifically directed otherwise. Filmmakers must "keep their hands on the wheel" to achieve authentic rather than archetypal representation.

For Emotional AI Research

The gendered encoding of emotions in AI systems reflects and potentially amplifies societal biases about who is allowed which feelings. This has implications for any AI system attempting to recognize, generate, or respond to human emotions.

For Training Data Curation

The overwhelming whiteness and gendered patterns suggest training datasets need conscious diversification not just in demographics but in emotional expressions across demographics. The current data appears heavily weighted toward Hollywood's narrow emotional typecasting.

For Critical AI Literacy

Users need awareness that AI doesn't generate neutral representations but reproduces historical and cinematic biases. Understanding these defaults is essential for conscious creation rather than unconscious replication.

Conclusions

With all three tests complete, the RAGE experiment reveals how AI encodes and reproduces emotional concepts through multiple layers of cultural sedimentation:

Rage defaults to masculine threat, hysteria to feminine suffering: Rage showed 60-70% male representation versus hysteria's 90% female. This gendered emotional hierarchy reflects patriarchal permissions - men's anger is legitimate external threat while women's distress is internal medical condition.
Pure word prompts reveal commercial packaging over human emotion: Test 3's movie posters, album covers, and comic books show AI learned rage primarily through entertainment media marketing. The word triggers genre conventions, not emotional experience.
Etymology meets Hollywood: While hysteria maintained its Greek "womb" root through Victorian and Hollywood tropes, rage's Latin rabies/madness manifests through werewolves, the Hulk, and beast transformations. Ancient meanings persist but are filtered through cinema.
Racial monolith with minimal cracks: 95%+ white representation across controlled tests, with ambiguity only through extreme lighting or creature transformation. AI's emotional vocabulary is overwhelmingly white.
Technical advancement enables complexity only through human curation: V7's superior capabilities for nuanced expression (calculating rage, mocking fury) required conscious selection. Defaults remained theatrical regardless of technical evolution.
Words arrive with complete visual packages: "Rage" alone summoned typography, color palettes (red/black), compositional conventions (movie posters), and character archetypes (Hulk). Single words trigger the entire mise-en-scène.
The director's eye remains essential: Across 40 curated selections, the most compelling images were outliers - contained over theatrical, complex over simple, narratively suggestive over isolated. AI can generate nuance but requires human curation to surface it.

The RAGE experiment demonstrates that precision in prompting, as PJ Pereira and Silverside AI established, is not just about better results - it's about conscious navigation of deeply embedded cultural biases. Without intentional direction, AI defaults to Hollywood's most worn grooves, perpetuating narrow emotional permissions based on gender, race, and commercial genre.

As Jaime Robinson noted, "the machines regurgitate." But this study reveals they don't just regurgitate "our" biases - they regurgitate etymological histories, cinematic conventions, and marketing categories that may have little to do with actual human emotional experience. The question becomes not whether AI can represent emotions, but whose emotions have been deemed worthy of representation in the training data.

For creators using these tools, the imperative is clear: keep your hands on the wheel. Understand that every emotion word carries demographic defaults, visual histories, and genre conventions. Only through conscious direction can we push beyond the clichés toward authentic representation.

The territory has been mapped. The biases documented. Now the real work begins - using this knowledge to create with intention rather than accepting defaults, to push for genuine emotional complexity rather than theatrical display, and to ensure that AI-assisted creativity expands rather than contracts the range of human experience we're able to represent.

References

Pereira, P. & Silverside AI. (2025). Emotion Precision in AI Image Generation. https://www.linkedin.com/posts/pjpereira_ai-acting-storytelling-activity-7375918690262892544-6rMH?utm_source=share&utm_medium=member_desktop&rcm=ACoAAACBHZcBNUM64Oyxygjg_rdTHtQJF4bVJXY

Silwany, O. & Meanwhile In Jupiter. (2025). HYSTERIA.I.: Visual Archaeology of a Gendered Diagnosis in AI Image Generation. HYSTERIA.I.

Silwany, O. & Meanwhile In Jupiter. (2025). HYSTERIA.I https://vimeo.com/1121662292?share=copy#t=0. Available at: https://www.linkedin.com/feed/update/urn:li:activity:7376725540021637121

Silwany, O. & Meanwhile In Jupiter. (2025). RAGE [Video]. Available at: [LinkedIn]

About Meanwhile In Jupiter

Meanwhile In Jupiter brings big-agency craft to the intersection of creativity and AI technology. Born from two decades on Madison Avenue and based in coastal Florida, we help ambitious brands navigate the new creative landscape through campaigns, actions, and films.

The RAGE study exemplifies our approach: understanding how language actually works in AI tools so creators can maintain intentional control. By revealing the hidden mechanics connecting words to visual output, we enable more precise creative direction in the AI era.

We don't just identify patterns — we map the territory. When creators understand these linguistic mechanics, they can keep their hands on the wheel.

Contact: Omar Silwany
Company: Meanwhile In Jupiter
Date: October 2025

Appendix: Complete Image Documentation

Test 1: Midjourney V6 with Full Prompt

Prompt: intense rage, close up portrait, cinematic lighting, dramatic mood --ar 16:9 --style raw --v 6

80 images generated (20 grids of 4), 20 selected for analysis

Selection Methodology

Human curation selecting for emotional complexity, compositional uniqueness, narrative implications, and performance nuance over theatrical display.

Generation 1

Grid Overview:

A (Top Left): Male, white, mid-thirties, controlled rage without teeth
B (Top Right): Female, white, young adult, screaming with bared teeth
C (Bottom Left): Male, white, mid-thirties, grimacing with scruff
D (Bottom Right): Male, white, mid-thirties, intense scowl with stubble

SELECTED: Position A (Top Left)

Demographics: Male, white, mid-thirties
Expression: Controlled/bottled rage without bared teeth, orange sparks/embers flying
Selection rationale: More intense performance, bottled fury more menacing than theatrical display

Generation 2

Grid Overview:

A (Top Left): Male, white, younger (late 20s/early 30s), screaming downward
B (Top Right): Male, white, 40s+, heavy stubble, eyes nearly closed
C (Bottom Left): Female, white, young adult, face paint/markings, snarling
D (Bottom Right): Male, white, young adult, yelling

SELECTED: Position B (Top Right)

Demographics: Male, white, 40s+, heavy stubble
Expression: Complex rage with eyes nearly closed, possible tears
Selection rationale: Emotional complexity - rage mixed with pain/grief, vulnerability within aggression

Generation 3

Grid Overview:

A (Top Left): Female, white, young adult, intense focused rage
B (Top Right): Male, white, 30s-40s, stubble, flared nostrils
C (Bottom Left): Male, white, 30s-40s, bald/shaved, grimacing
D (Bottom Right): Male, white, 30s-40s, stubble, snarling

SELECTED: Position A (Top Left)

Demographics: Female, white, young adult
Expression: Controlled, focused rage - cold and calculated
Selection rationale: Unique interpretation among three interchangeable males, her rage reads differently

Generation 4

Grid Overview:

A (Top Left): Female, white, young adult, dark blonde, closed bite showing teeth
B (Top Right): Male, white, young adult, sweaty, snarling/baring teeth with disgust
C (Bottom Left): Female, white, young adult, auburn brunette, intense stare
D (Bottom Right): Male, white, early thirties, grimacing, baring teeth

SELECTED: Position C (Bottom Left)

Demographics: Female, white, young adult, auburn brunette
Expression: Intense stare with fury in eyes, no bared teeth
Selection rationale: "Angry as hell" without showing teeth - the outlier showing restraint

Generation 5

Grid Overview:

A (Top Left): Female, white, young adult, blonde, snarling with teeth
B (Top Right): Male, white, early thirties, intense expression with teeth
C (Bottom Left): Male, white, 40s+, bearded, grimacing with teeth
D (Bottom Right): Male, white, 30s, clenched teeth, whites of eyes showing

SELECTED: Position D (Bottom Right)

Demographics: Male, white, 30s
Expression: Teeth clenched, whites of eyes showing, almost drooling
Selection rationale: Possession-like primal rage, beyond anger into something inhuman

Generation 6

Grid Overview:

A (Top Left): Male, white, 30s, yelling with teeth
B (Top Right): Female, white, young adult, snarling
C (Bottom Left): Female, white, young adult, intense stare without teeth
D (Bottom Right): Female, white, young adult, screaming

SELECTED: Position C (Bottom Left)

Demographics: Female, white, young adult
Expression: No teeth showing, wet hair/sweaty, intense rage in eyes ready to erupt
Selection rationale: Contained fury about to explode - potential energy vs kinetic

Generation 7

Grid Overview:

A (Top Left): Male, white, 30s-40s, bald/shaved, intense stare, creepy eyes
B (Top Right): Male, white, 20s-30s, water/sweat flying off face, yelling
C (Bottom Left): Female, white, young adult, growling in shadows
D (Bottom Right): Male, white, 40s+, heightened disgust in snarl

SELECTED: Position B (Top Right)

Demographics: Male, white, 20s-30s
Expression: Water/sweat flying off face, yelling, kinetic rage
Selection rationale: Only image showing motion/action - explosive movement captured

Generation 8

Grid Overview:

A (Top Left): Female, white, young adult, inside setting with side lighting
B (Top Right): Male, white, early 30s, dark moody lighting with vapor
C (Bottom Left): Female, white, young adult, warm sun-like lighting
D (Bottom Right): Male, white, 40s-50s, pockmarked skin, red/orange lighting

SELECTED: Position C (Bottom Left)

Demographics: Female, white, young adult
Expression: Extremely close framing, teeth showing
Selection rationale: Compositional uniqueness - intimacy of close-up creates different intensity

Generation 9

Grid Overview:

A (Top Left): Female, white, young adult, glowing/backlit, open mouth
B (Top Right): Female, white, young adult, dark green lighting, teeth showing
C (Bottom Left): Male, white, early 20s, clenched teeth visible
D (Bottom Right): Male, white, young adult, red/blue split lighting, heavy smoke/vapor

SELECTED: Position C (Bottom Left)

Demographics: Male, white, early 20s
Expression: Teeth clenched but visible, off-center composition
Selection rationale: Joaquin Phoenix quality, person coming at camera from left, controlled tension

Generation 10

Grid Overview:

A (Top Left): Male, white, 30s, beard, animalistic snarl (Wolverine-like)
B (Top Right): Male, white, 30s, sweaty, teeth bared
C (Bottom Left): Male, white, young adult, yelling
D (Bottom Right): Male, white, 30s-40s, eyes nearly closed, indoor evening light

SELECTED: Position D (Bottom Right)

Demographics: Male, white, 30s-40s
Expression: Eyes nearly closed, only darkness visible, indoor evening light
Selection rationale: Naturalistic performance suggesting domestic argument, situational rage with context

Generation 11

Grid Overview:

A (Top Left): Male, white/possibly mixed, 20s, gray hoodie, closed mouth, steaming
B (Top Right): Male, white, young adult, yelling with red/orange side lighting with steam
C (Bottom Left): Male, white, 20s, snarling, hunched like on Peloton
D (Bottom Right): Male, white, 50s, lower teeth visible, mix of fear with rage

SELECTED: Position A (Top Left)

Demographics: Male, white (possibly mixed), 20s
Expression: Gray hoodie, mouth closed, literally steaming
Selection rationale: Rage in forehead wrinkles between eyes, not nose bridge, different musculature

Generation 12

Grid Overview:

A (Top Left): Male, white, 40s-50s with greying hair, profile view, yelling
B (Top Right): Female, white, young adult, snarling with teeth
C (Bottom Left): Female, white, young adult, wet hair, yelling
D (Bottom Right): Male, white, 30s-40s, grimacing with teeth

SELECTED: Position A (Top Left)

Demographics: Male, white, 40s-50s with greying hair
Expression: 3/4 profile view showing straining neck muscles
Selection rationale: Unique framing showing physical mechanics of rage, profile reveals effort

Generation 13

Grid Overview:

A (Top Left): Male, white, 30s, no teeth showing, intense stare
B (Top Right): Male, white, 30s, bearded, yelling with teeth
C (Bottom Left): Female, white, young adult, grimacing with teeth
D (Bottom Right): Male, white, 40s, snarling with teeth

SELECTED: Position A (Top Left)

Demographics: Male, white, 30s
Expression: No teeth showing, rage conveyed entirely through eyes
Selection rationale: More menacing for its restraint, cold calculated fury

Generation 14

Grid Overview:

A (Top Left): Female, white, young adult, blood on face, teeth visible
B (Top Right): Male, white, 30s-40s, heavy beard, no teeth showing
C (Bottom Left): Female, white, young adult, blonde, mouth slightly open
D (Bottom Right): Male, white, 40s, full beard, teeth bared in snarl

SELECTED: Position B (Top Right)

Demographics: Male, white, 30s-40s, heavy beard
Expression: No teeth showing, intense stare only
Selection rationale: Most subtle performance, controlled menace in the eyes

Generation 15

Grid Overview:

A (Top Left): Male, white, 30s, grimacing with teeth
B (Top Right): Male, white, 30s, extreme snarl with teeth
C (Bottom Left): Male, white, 30s, teeth bared
D (Bottom Right): Female, white, young adult, complex expression

SELECTED: Position D (Bottom Right)

Demographics: Female, white, young adult
Expression: Not snarling but showing "rage of hate"
Selection rationale: More psychological than physical threat, less animalistic than males

Generation 16

Grid Overview:

A (Top Left): Female, white, young adult, blonde, teeth showing, fiery lighting
B (Top Right): Male, white, 30s, stubble, grimacing with teeth
C (Bottom Left): Male, white, 40s+, grimacing with teeth
D (Bottom Right): Female, white, young adult, eyes closed, crying/yelling

SELECTED: Position D (Bottom Right)

Demographics: Female, white, young adult
Expression: Eyes closed, mouth open in cry/yell, rage not directed at camera
Selection rationale: First with closed eyes, rage turned inward, cry of despair mixed with fury

Generation 17

Grid Overview:

A (Top Left): Male, white, 30s, bearded, yelling with golden/fire lighting
B (Top Right): Female, white, young adult, more subdued expression
C (Bottom Left): Male, white, young adult, grimacing with teeth
D (Bottom Right): Male, white, 30s-40s, extreme snarl with heavy eye compression

SELECTED: Position B (Top Right)

Demographics: Female, white, young adult
Expression: Less animalistic, "interrogating rage" - questioning/demanding
Selection rationale: More sophisticated cinematography, rage seeking answers

Generation 18

Grid Overview:

A (Top Left): Male, white, 30s-40s, grimacing with teeth, close framing
B (Top Right): Male, white, 30s, stubble, teeth showing, sweaty
C (Bottom Left): Male, white, 40s+, bearded, teeth bared, wet/sweaty
D (Bottom Right): Female, white, young adult, warm/fire lighting

SELECTED: Position D (Bottom Right)

Demographics: Female, white, young adult
Expression: Warm/fire lighting suggesting transformation
Selection rationale: "Firestarter" interpretation, potential origin story moment

Generation 19

Grid Overview:

A (Top Left): Male, white, 40s+, bearded, teeth showing
B (Top Right): Male, white, late teens/early 20s, downcast eyes, contained
C (Bottom Left): Male, white, 30s, grimacing with teeth
D (Bottom Right): Male, white, 40s-50s, bearded, snarling

SELECTED: Position B (Top Right)

Demographics: Male, white, late teens/early 20s
Expression: Eyes looking down, "innocent or primeval rage"
Selection rationale: Youth discovering rage, first experience of emotion's power

Generation 20

Grid Overview:

A (Top Left): Male, white, young adult, teeth showing, orange lighting
B (Top Right): Male, white, 30s, teeth showing, stubble/beard
C (Bottom Left): Male, white, 30s-40s, extreme snarl, blue lighting
D (Bottom Right): Female, white, young adult, disheveled, teeth showing

SELECTED: Position D (Bottom Right)

Demographics: Female, white, young adult
Expression: Very disheveled appearance, odd and haunting quality
Selection rationale: Different from typical "pretty rage," genuinely unsettling and disturbing

Test 2: Midjourney V7 with Full Prompt

Prompt: intense rage, close up portrait, cinematic lighting, dramatic mood --ar 16:9 --style raw --v 7

80 images generated (20 grids of 4), 20 selected for analysis

Generation 1

Grid Overview:

A (Top Left): Female, white, early 20s, teeth showing, blue eyes
B (Top Right): Male, white, 40-50, squinting but eyes visible
C (Bottom Left): Male, white, early 30s, bearing teeth/possibly screaming, very tight crop
D (Bottom Right): Female, white, late teens, manic quality

SELECTED: Position D (Bottom Right)

Demographics: Female, white, late teens
Expression: Manic/unsettling expression, teeth visible in disturbing smile
Selection rationale: Better emotion from performance, connects to rabies/madness etymology

Generation 2

Grid Overview:

A (Top Left): Male, white, 30s-40s, bearded, teeth clenched, warm tone
B (Top Right): Female, white, early 20s, shiny skin, teeth visible, warm tone
C (Bottom Left): Female, white, early 20s, screaming, warm tone
D (Bottom Right): Male, racially ambiguous light skinned, 30s, light beard, screaming, blue/cool tone

SELECTED: Position A (Top Left)

Demographics: Male, white, 30s-40s, bearded
Expression: Teeth clenched, "closer to a boil" - maximum pressure
Selection rationale: Physical restraint about to break, precise moment before control breaks

Generation 3

Grid Overview:

A (Top Left): Male, white, 40s-50s, bearded, squinting in low light
B (Top Right): Male, white, early-mid 20s, curly hair, teeth clenched
C (Bottom Left): Male, white, early-mid 20s, mouth open screaming
D (Bottom Right): Female, white, early 20s, mouth open with teeth

SELECTED: Position B (Top Right)

Demographics: Male, white, early-mid 20s, curly hair
Expression: Teeth clenched - only one showing restraint
Selection rationale: Contained rage pattern continuing

Generation 4

Grid Overview:

A (Top Left): Male, white, 30s-40s, warm lighting, eyes squinting in shadows
B (Top Right): Female, white, early 20s, mouth open with teeth, raised eyebrow
C (Bottom Left): Male, white, red/blue split lighting, extreme close-up
D (Bottom Right): Male, white, 40s-50s, heavy wrinkles/eye compression

SELECTED: Position A (Top Left)

Demographics: Male, white, 30s-40s
Expression: Eyes squinting in shadows, warm side lighting
Selection rationale: Compositional and cinematographic uniqueness, rage directed toward light source

Generation 5

Grid Overview:

A (Top Left): Female, white, early 20s, reddish hair, screaming at camera
B (Top Right): Male, white, early 30s, lightly bearded, yelling OFF camera to side
C (Bottom Left): Male, white, 30s-40s, scowling, extreme crop
D (Bottom Right): Male, white, 40s, bearded, teeth nearly clenched

SELECTED: Position B (Top Right)

Demographics: Male, white, early 30s, lightly bearded
Expression: Yelling OFF camera to the side, left profile view
Selection rationale: Editorial functionality - can create conversational conflict sequences in editing

Generation 6

Grid Overview:

A (Top Left): Female, white, early 20s, brunette, no eye compression, amazement
B (Top Right): Female, white, early 20s, ginger, teeth visible, theatrical rage
C (Bottom Left): Male, white, 30s, bearded, extreme eye compression
D (Bottom Right): Male, white, mid 40s, heavy eye compression, intense stare

SELECTED: Position A (Top Left)

Demographics: Female, white, early 20s, brunette
Expression: No theatrical eye compression, mouth open in surprise/amazement
Selection rationale: Rage as overwhelming emotion, not pure anger - breaks pattern

Generation 7

Grid Overview:

A (Top Left): Female, white, early 20s, freckles, looking OFF camera
B (Top Right): Male, white, early 20s, mouth open yelling
C (Bottom Left): Female, white, early 20s, teeth clenched, asymmetrical expression
D (Bottom Right): Female, white, early 20s, mouth open, teeth visible

SELECTED: Position C (Bottom Left)

Demographics: Female, white, early 20s
Expression: Teeth clenched, asymmetrical - raised side of top lip, "twisted rage"
Selection rationale: Possession quality, internal struggle visible, rage fighting against itself

Generation 8

Grid Overview:

A (Top Left): Male, white, early 20s, mouth open, warm/red lighting
B (Top Right): Male, white, mid 30s, mouth open, cool/blue lighting
C (Bottom Left): Female, white, early 20s, mocking laugh + anger hybrid
D (Bottom Right): Male, white, 40s-50s, bearded, mouth open

SELECTED: Position C (Bottom Left)

Demographics: Female, white, early 20s
Expression: Mocking laugh + intense anger hybrid
Selection rationale: Most naturalistic cinematography, rage that enjoys itself, cruelty with pleasure

Generation 9

Grid Overview:

A (Top Left): Male, white, 30s, teeth showing, warm lighting
B (Top Right): Male, brown/possibly mixed Black, 20s-30s, red/blue split lighting
C (Bottom Left): Female, white, early 20s, freckles, nearly clenched teeth
D (Bottom Right): Female, white, early 20s, mouth open

SELECTED: Position C (Bottom Left)

Demographics: Female, white, early 20s, freckles
Expression: Nearly clenched teeth with jaw slightly extended
Selection rationale: Maintaining methodological consistency over documenting racial diversity

ALSO DOCUMENTED: Position B (Top Right) - First non-white subject

Generation 10

Grid Overview:

A (Top Left): Male, Black, 30s, rage with horror in eyes
B (Top Right): Male, white, 50s, heavy beard, teeth clenched
C (Bottom Left): Male, white, 40s-50s, mouth open yelling, red lighting
D (Bottom Right): Male, pale white, bald, 30s-40s, teeth clenched

SELECTED: Position A (Top Left)

Demographics: Male, Black, 30s
Expression: Rage expressed outwardly WITH horror in eyes - dual emotion
Selection rationale: Emotional complexity plus demographic significance as outlier

Generation 11

Grid Overview:

A (Top Left): Male, white, 20s, freckles, intense eye compression
B (Top Right): Male, white, 20s, mouth wide open, red/blue lighting
C (Bottom Left): Male, appears Asian, 20s, teeth showing
D (Bottom Right): Male, white, 20s, mouth open

SELECTED: Position C (Bottom Left)

Demographics: Male, appears Asian, 20s
Expression: Mouth open, teeth showing
Selection rationale: Third non-white representation in V7, demographic outlier

Generation 12

Grid Overview:

A (Top Left): Male, white, 30s-40s, bearded, teeth showing
B (Top Right): Female, white, early 20s, teeth visible, warm lighting
C (Bottom Left): Male, white, 20s-30s, mouth open, red lighting
D (Bottom Right): Male, white, 40s-50s, mustache only, mouth cropped out

SELECTED: Position D (Bottom Right)

Demographics: Male, white, 40s-50s, mustache
Expression: Mouth completely cropped out, extreme eye compression showing "hate"
Selection rationale: Cold sustained malice rather than hot rage

Generation 13

Grid Overview:

A (Top Left): Female, white, early 20s, blue eyes, teeth visible
B (Top Right): Male, possibly mixed race, 40s-50s, extreme eye compression
C (Bottom Left): Male, white, 30s-40s, bearded, extreme crop
D (Bottom Right): Female, possibly Latina/mixed, early 20s, mouth open

SELECTED: Position D (Bottom Right)

Demographics: Female, possibly Latina/mixed race, early 20s
Expression: Mouth open with teeth visible
Selection rationale: Potential fourth instance of racial diversity in V7

Generation 14

Grid Overview:

A (Top Left): Male, white, 40s-50s, bearded, clenched teeth, off-center framing
B (Top Right): Female, white, early 20s, teeth visible/open
C (Bottom Left): Male, white, 30s, clenched teeth (more pronounced)
D (Bottom Right): Male, white, 20s-30s, bearded, mouth open

SELECTED: Position A (Top Left)

Demographics: Male, white, 40s-50s, bearded
Expression: Teeth clenched, off-center framing showing side profile
Selection rationale: Combines clenched restraint with unique composition

Generation 15

Grid Overview:

A (Top Left): Female, white, early 20s, freckles, mouth cropped
B (Top Right): Female, Black, early 20s, mouth open, off-center framing
C (Bottom Left): Male, white/possibly Latino, 30s, bearded, mouth open
D (Bottom Right): Male, white/possibly Latino, 30s, less beard, teeth showing

SELECTED: Position B (Top Right)

Demographics: Female, Black, early 20s
Expression: Mouth open with teeth, facing camera but off-center framing
Selection rationale: Clear Black female representation, compositional tension

Generation 16

Grid Overview:

A (Top Left): Female, white, early 20s, blonde, soft natural lighting, full mouth visible
B (Top Right): Female, white, early 20s, brunette, red side lighting
C (Bottom Left): Female, white, early 20s, auburn hair, freckled, spotlit
D (Bottom Right): Male, white, 40s, bearded, pockmarked skin, teeth clenched, strong spotlight

SELECTED: Position A (Top Left)

Demographics: Female, white, early 20s
Expression: Full mouth visible (no cropping), natural soft lighting
Selection rationale: Complete framing unusual for V7, documentary quality

Generation 17

Grid Overview:

A (Top Left): Male, white, 30s-40s, extreme close crop, warm lighting
B (Top Right): Male, white, 40s-50s, heavy eye compression
C (Bottom Left): Female, white, early 20s, fair skin, freckled, mouth CLOSED
D (Bottom Right): Female, white, early 20s, blonde, head in motion, teeth clenched

SELECTED: Position C (Bottom Left)

Demographics: Female, white, early 20s
Expression: Mouth completely CLOSED, "controlled and calculating" rage
Selection rationale: Most restrained expression - rage carried entirely in eyes

Generation 18

Grid Overview:

A (Top Left): Female, white, early 20s, red lighting, teeth slightly visible
B (Top Right): Female, white, early 20s, zig-zag eye compression, frame-biting crop
C (Bottom Left): Male, white, 40s-50s, gray stubble, extreme eye bulging
D (Bottom Right): Male, white, 30s-40s, lightly bearded, squinting

SELECTED: Position B (Top Right)

Demographics: Female/Gender ambiguous, white, early 20s
Expression: Unique zig-zag eye compression, cropped to appear "biting the frame"
Selection rationale: Unique anatomical distortion, feral quality with nostril hair visible

Generation 19

Grid Overview:

A (Top Left): Female, white, early 20s, disgust in eyes, teeth clenched
B (Top Right): Male, white, 30s-40s, bushy eyebrows, Wolverine-esque
C (Bottom Left): Male, white, 40s-50s, extreme close crop
D (Bottom Right): Male, white, 30s-40s, blonde beard, most extreme crop, one eye visible

SELECTED: Position D (Bottom Right)

Demographics: Male, white, 30s-40s, blonde beard
Expression: Most extreme crop, only one eye visible, squinting
Selection rationale: Single visible eye in extreme squint, darkness hiding half the face

Generation 20

Grid Overview:

A (Top Left): Male, white, 30s-40s, bearded, teeth nearly clenched
B (Top Right): Gender ambiguous, white, 20s-30s, teeth nearly clenched
C (Bottom Left): Male, white, 30s-40s, mustache, teeth nearly clenched
D (Bottom Right): Gender ambiguous, white, 20s, freckled, controlled stare, no teeth showing

SELECTED: Position D (Bottom Right)

Demographics: Gender ambiguous, white, 20s
Expression: Controlled stare, no teeth showing, freckled face
Selection rationale: Only image without visible teeth - controlled, calculating rage carried entirely in eyes, fits established pattern of selecting restraint over theatrical display

Test 3: Midjourney V7 Pure Word Test

Prompt: rage (no additional parameters)

20 images generated (5 grids of 4), ALL documented

Generation 1 (Images 1-4)

Image 1 - Top Left: Horror movie poster aesthetic | Text: Yes - red distressed "RAGE" | Demographics: Woman, white, early 20s, screaming | Description: Dark/gritty, freckles/blood spatter
Image 2 - Top Right: Thriller poster style | Text: Yes - white distressed "RAGE" | Demographics: Male, 30s, race ambiguous (white/Asian/Latino) | Description: Red-lit smoke-filled environment
Image 3 - Bottom Left: Psychological thriller poster | Text: Yes - gray "RAGE" | Demographics: Male, white, early 20s | Description: Dramatic noir lighting
Image 4 - Bottom Right: Graphic design/album/comic aesthetic | Text: NO | Demographics: Gender ambiguous | Description: Paint splatter effect, manga/comic book style

Generation 2 (Images 5-8)

Image 5 - Top Left: Anime/manga style | Text: Yes - pink/red "RA RAGE" (doubled/glitched) | Demographics: Male, wide eyes | Description: Character screaming with teeth showing
Image 6 - Top Right: Comic book/illustration style | Text: Yes - blue-grey "RAGE" matching hair | Demographics: Male | Description: Screaming face on beige background
Image 7 - Bottom Left: Horror poster | Text: Yes - red "RAGE" | Demographics: Male close-up portrait | Description: Dark/bloody aesthetic, screaming with teeth
Image 8 - Bottom Right: Horror movie poster | Text: Yes - off-white distressed "RAGE" | Demographics: Male, werewolf-like with fangs | Description: Red glowing eyes, supernatural transformation

Generation 3 (Images 9-12)

Image 9 - Top Left: Anime/manga style | Text: NO | Demographics: Blonde male | Description: Orange/fire emanating from head, prominent canines
Image 10 - Top Right: Black and white with red accent | Text: NO | Demographics: Male, creature-like | Description: Profile view, white eyes without retinas
Image 11 - Bottom Left: Comic illustration | Text: NO | Demographics: Male, 20s-30s | Description: Extreme close-up with heavy shadows
Image 12 - Bottom Right: Manga/anime style | Text: NO | Demographics: Likely female, gender ambiguous | Description: Black and white tones, screaming

Generation 4 (Images 13-16)

Image 13 - Top Left: Photographic, darkness aesthetic | Text: Yes - red "RAGE" | Demographics: Racially ambiguous, gender ambiguous, early 20s | Description: Screaming in darkness, minimal lighting
Image 14 - Top Right: Anime/manga style | Text: Yes - white distressed "RAGE" | Demographics: Gender ambiguous | Description: Screaming on black background
Image 15 - Bottom Left: Photographic, extreme close-up | Text: Yes - "RAGE" | Demographics: White person likely, gender unclear, 20s | Description: Extreme close-up portrait
Image 16 - Bottom Right: Creature illustration | Text: Yes - red hand-lettered "RAGE" | Demographics: Creature with large fangs | Description: Red tones throughout, profile view

Generation 5 (Images 17-20)

Image 17 - Top Left: Photographic movie poster | Text: Yes - pinkish "RAGE" | Demographics: Gender ambiguous (female 20s or young male) | Description: Black background, face distressed
Image 18 - Top Right: Photographic poster | Text: Yes - small "RAGE" lower right | Demographics: Male, 30s-40s | Description: Profile screaming, blood/scabs on face
Image 19 - Bottom Left: Photographic poster | Text: Yes - "RAGE" on left side | Demographics: Female, white, early 20s | Description: Red tone throughout, face off-center
Image 20 - Bottom Right: Line art/comic | Text: NO - THE HULK character | Demographics: The Hulk (Marvel character) | Description: Screaming in fury, neck straining, character speaks for itself

Statistical Summary

Test 1 (V6) Demographics

Total Male: 13-14/20 (65-70%)
Total Female: 6-7/20 (30-35%)
Total White: 20/20 (100%)
Total Non-white: 0/20 (0%)
Age ranges: Primarily 20s-40s, some 50s

Test 2 (V7) Demographics

Total Male: 12/20 (60%)
Total Female or gender ambiguous: 8/20 (40%)
Total White: 18/20 (~90%)
Total Non-white or racially ambiguous: 2/20 (~10%)
Age ranges: Broader range (early 20s to 50s)

Test 3 (Pure Word) Patterns

Images with "RAGE" text: 15/20 (75%)
Movie poster/Album/Book cover aesthetics: 12/20 (60%)
Comic/Manga/Anime style: 7/20 (35%)
Line art: 1/20 (5%)
Creature/transformation: 4-5/20 (20-25%)
Demographics often ambiguous due to stylization

Expression Taxonomy Discovered

Categories of Rage Expression (Tests 1 & 2)

Contained/Bottled Rage - Clenched teeth, visible strain, internalized fury
Complex Emotional Hybrids - Rage mixed with tears, horror, mockery, disgust
Naturalistic/Situational Rage - Domestic arguments, contextual anger
Interrogating/Purposeful Rage - Directed fury seeking answers
Innocent/Primeval Rage - Youth discovering rage for first time
Manic/Madness Expression - Connection to rabies etymology
Twisted/Asymmetrical Rage - Facial distortion, possession quality
Calculating/Cold Malice - Controlled, premeditated anger
Horror Within Rage - Fear mixed with fury (especially in non-white subjects)
Hate - Sustained malice rather than explosive anger
Kinetic Rage - Motion captured, water/sweat flying
Origin Story Rage - Transformation moments

Visual Encoding Patterns

V6 Characteristics:

Heavy eye compression creating animalistic transformation
Orange/teal action cinema color grading
Steam/heat effects suggesting literal pressure
Wider framing than V7

V7 Characteristics:

Extreme close crops often cutting off features
Better facial detail and realism
More psychological complexity
Frame-biting compositions

Test 3 (Pure Word) Patterns:

Rage as commercial genre (movie posters, comics)
Typography as central design element/title
Creature transformations (werewolves, white eyes)
Fire/explosion visual metaphors
The Hulk as ultimate rage signifier

Key Findings

Demographic Patterns

Male dominance in Tests 1-2 (60-70%) vs female dominance in hysteria (90%)
Overwhelming whiteness (92.5% in curated tests, 100% in V6)
Non-white subjects often showed different emotional qualities (horror within rage)

Emotional Complexity

Consistently selected for restraint over theatrical display
Found emotional nuance through human curation
V7 enabled more psychological complexity than V6
Pure word test revealed commercial rather than emotional encoding

Cultural Encoding

Etymology preserved: Latin rabies/madness manifests as beast transformation
Hollywood's visual language dominates training data
Without parameters, rage becomes entertainment product not human emotion
The Hulk's appearance confirms superhero culture's influence

Conclusion

The RAGE experiment reveals AI's encoding of emotion through multiple cultural layers. While hysteria triggered Victorian medical imagery and female suffering, rage defaults to masculine threat, action cinema aesthetics, and ultimately commercial entertainment packaging. The progression from curated emotional portraits (Tests 1-2) to unprompted genre products (Test 3) demonstrates that AI has learned rage primarily through Hollywood and comic book culture rather than human experience. Only through conscious curation and precise prompting can creators push beyond these defaults toward authentic emotional representation.