Files
2nd/10_Wiki/Topics/AI_and_ML/Plutchiks-Wheel-of-Emotions.md
T
2026-05-10 22:08:15 +09:00

213 lines
7.3 KiB
Markdown
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
---
id: wiki-2026-0508-plutchiks-wheel-of-emotions
title: Plutchik's Wheel of Emotions
category: 10_Wiki/Topics
status: verified
canonical_id: self
aliases: [plutchik-wheel, emotion-wheel, primary-emotions]
duplicate_of: none
source_trust_level: A
confidence_score: 0.9
verification_status: applied
tags: [emotion, psychology, affective-computing, nlp]
raw_sources: []
last_reinforced: 2026-05-10
github_commit: pending
tech_stack:
language: Python
framework: transformers
---
# Plutchik's Wheel of Emotions
## 매 한 줄
> **"매 8 primary emotions × 3 intensities + dyads = 매 emotion taxonomy"**. 매 Robert Plutchik (1980) 매 evolutionary-grounded model 의 propose. 매 8 primary: joy, trust, fear, surprise, sadness, disgust, anger, anticipation. 매 affective computing + emotion-aware NLP 의 widely used.
## 매 핵심
### 매 8 primary emotions (4 axes)
1. **매 joy ↔ sadness**
2. **매 trust ↔ disgust**
3. **매 fear ↔ anger**
4. **매 surprise ↔ anticipation**
### 매 intensity (radial)
- 매 joy: ecstasy (high) → joy (mid) → serenity (low).
- 매 anger: rage → anger → annoyance.
- 매 fear: terror → fear → apprehension.
- (매 8 primary × 3 levels = 24 emotions)
### 매 dyads (combinations)
- **매 primary dyads** (adjacent): joy + trust = love, trust + fear = submission, fear + surprise = awe.
- **매 secondary dyads** (one apart): joy + fear = guilt, anger + joy = pride.
- **매 tertiary dyads** (two apart): joy + surprise = delight, anger + surprise = outrage.
- **매 opposite dyads** (across): rare/conflicting (e.g. joy + sadness = bittersweet).
### 매 vs other models
- **매 Ekman (6 basic)**: anger, disgust, fear, happiness, sadness, surprise — 매 universal facial expression-based.
- **매 GoEmotions (27, Google 2020)**: 매 Reddit-derived, 매 fine-grained.
- **매 dimensional (VAD)**: valence-arousal-dominance — 매 continuous space.
- **매 Plutchik 매 advantages**: 매 structured (axes + intensity + combinations), 매 evolutionary grounding.
### 매 응용
1. 매 emotion classification (NLP).
2. 매 sentiment analysis (richer than pos/neg).
3. 매 chatbot empathy modeling.
4. 매 mental health monitoring.
## 💻 패턴
### Plutchik labels — Python enum
```python
from enum import Enum
class Plutchik8(Enum):
JOY = "joy"
TRUST = "trust"
FEAR = "fear"
SURPRISE = "surprise"
SADNESS = "sadness"
DISGUST = "disgust"
ANGER = "anger"
ANTICIPATION = "anticipation"
INTENSITY = {
"joy": ["serenity", "joy", "ecstasy"],
"anger": ["annoyance", "anger", "rage"],
"fear": ["apprehension", "fear", "terror"],
# ...
}
PRIMARY_DYADS = {
("joy", "trust"): "love",
("trust", "fear"): "submission",
("fear", "surprise"): "awe",
("surprise", "sadness"): "disapproval",
("sadness", "disgust"): "remorse",
("disgust", "anger"): "contempt",
("anger", "anticipation"): "aggressiveness",
("anticipation", "joy"): "optimism",
}
```
### Multi-label emotion classifier (transformers)
```python
import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer
# 매 j-hartmann/emotion-english-distilroberta-base — 7 emotions (Plutchik-aligned subset)
MODEL = "j-hartmann/emotion-english-distilroberta-base"
tok = AutoTokenizer.from_pretrained(MODEL)
model = AutoModelForSequenceClassification.from_pretrained(MODEL).eval()
def emotion_scores(text):
inputs = tok(text, return_tensors="pt", truncation=True)
with torch.no_grad():
logits = model(**inputs).logits
probs = logits.softmax(-1)[0]
labels = model.config.id2label
return {labels[i]: float(probs[i]) for i in range(len(probs))}
print(emotion_scores("I just got promoted!"))
# {'joy': 0.94, 'surprise': 0.04, ...}
```
### LLM emotion classification (Claude)
```python
from anthropic import Anthropic
client = Anthropic()
PROMPT = """Classify the emotion in this text using Plutchik's 8 primary emotions
(joy, trust, fear, surprise, sadness, disgust, anger, anticipation).
Return JSON: {"primary": [...], "intensity": "low|mid|high", "dyad": "..."}.
Text: {text}"""
def classify(text):
msg = client.messages.create(
model="claude-opus-4-7",
max_tokens=200,
messages=[{"role": "user", "content": PROMPT.format(text=text)}],
)
return msg.content[0].text
```
### Wheel visualization (matplotlib polar)
```python
import matplotlib.pyplot as plt, numpy as np
emotions = ["joy", "trust", "fear", "surprise", "sadness", "disgust", "anger", "anticipation"]
colors = ["#FFEB3B", "#8BC34A", "#4CAF50", "#00BCD4", "#3F51B5", "#9C27B0", "#F44336", "#FF9800"]
angles = np.linspace(0, 2*np.pi, 8, endpoint=False)
fig, ax = plt.subplots(subplot_kw={"projection": "polar"}, figsize=(8, 8))
ax.bar(angles, [1]*8, width=2*np.pi/8, color=colors, alpha=0.7)
for a, e in zip(angles, emotions):
ax.text(a, 1.1, e, ha="center")
ax.set_yticks([]); ax.set_xticks([])
plt.show()
```
### Dyad combination
```python
def combine(e1, e2):
key = tuple(sorted([e1, e2]))
for (a, b), dyad in PRIMARY_DYADS.items():
if tuple(sorted([a, b])) == key:
return dyad
return f"{e1}+{e2} (uncommon)"
print(combine("joy", "trust")) # love
```
### GoEmotions → Plutchik mapping
```python
GOEMOTIONS_TO_PLUTCHIK = {
"admiration": "trust", "amusement": "joy", "anger": "anger",
"annoyance": "anger", "approval": "trust", "caring": "trust",
"confusion": "surprise", "curiosity": "anticipation", "desire": "anticipation",
"disappointment": "sadness", "disapproval": "disgust", "disgust": "disgust",
"embarrassment": "fear", "excitement": "joy", "fear": "fear",
"gratitude": "joy", "grief": "sadness", "joy": "joy",
"love": "joy", "nervousness": "fear", "optimism": "anticipation",
"pride": "joy", "realization": "surprise", "relief": "joy",
"remorse": "sadness", "sadness": "sadness", "surprise": "surprise",
}
```
## 매 결정 기준
| 상황 | Model |
|---|---|
| 매 structured 8-class (interpretable) | 매 Plutchik |
| 매 facial expression | 매 Ekman 6 |
| 매 fine-grained social media | 매 GoEmotions 27 |
| 매 continuous (intensity gradient) | 매 VAD dimensional |
| 매 product reviews (binary) | 매 sentiment pos/neg |
**기본값**: 매 Plutchik 8 매 emotion classification baseline (interpretable + structured).
## 🔗 Graph
- 부모: [[Affective-Computing]] · [[Emotion-Theory]]
- 변형: [[Ekman-Basic-Emotions]] · [[GoEmotions]] · [[VAD-Model]]
- 응용: [[Sentiment-Analysis]] · [[Emotion-Classification]] · [[Chatbot-Empathy]]
- Adjacent: [[Mental-Health-AI]] · [[Conversational-AI]]
## 🤖 LLM 활용
**언제**: 매 emotion taxonomy 의 prompt 매 LLM 의 give → 매 zero-shot Plutchik classification, 매 chatbot empathy module.
**언제 X**: 매 cross-cultural emotion (Plutchik 매 Western-centric), 매 micro-expression (use Ekman + AU).
## ❌ 안티패턴
- **매 mutually exclusive assumption**: 매 emotions 매 co-occur — 매 multi-label.
- **매 ignore intensity**: 매 "anger" vs "rage" 매 different.
- **매 universalism**: 매 cultural variation 매 exists (Plutchik 매 Western bias).
## 🧪 검증 / 중복
- Verified (Plutchik 1980 "A general psychoevolutionary theory of emotion", standard in affective computing literature).
- 신뢰도 A.
## 🕓 Changelog
| 날짜 | 변경 |
|---|---|
| 2026-05-08 | Phase 1 |
| 2026-05-10 | Manual cleanup — 8 primary + dyads + classifier patterns + GoEmotions mapping |