Affective Computing & Emotion AI Expert (2026 Edition)

English | Bahasa Indonesia

---

<a name="english"></a>

English

Orchestration & Integration

• Upstream Orchestrators: Works under brainstorming and zero-to-prod-orchestrator (Phase 4 - AI Agents & Multimodal Pipelines) to integrate real-time affective computing.
• Downstream Cooperating Skills:
  • voice-ai-realtime-agent — Bidirectional WebRTC audio transport and low-latency audio capture (<300ms).
  • ephemeral-generative-ui-architect — Dynamically re-render UI components according to calculated AffectVector.
  • ui-ux-pro-max — Ensure affective UI themes adhere to WCAG 2.2 contrast ratios and non-jarring transitions.
  • ai-llm-integration-expert — Multimodal token handling via Gemini Multimodal Live API.
  • session-memory-manager — Persist affective interaction history and user emotional state vectors.

Description

Comprehensive guide for Affective Computing, empathic conversational agents, and real-time biometric sentiment synthesis using native multimodal tokens (Gemini 4 Pro Multimodal Live API, WebRTC bidirectional audio streams). Interprets vocal pitch variance, speech rate, pause intervals, and facial micro-expressions. Maps affective signals into continuous 2D Valence-Arousal coordinate space (Russell Circumplex Model) to dynamically modulate agent voice prosody, conversation pacing, and generative UI layouts.

Trigger Conditions

• Developing empathetic AI customer service, healthcare triage, or crisis support agents.
• Adaptive EdTech tutoring platforms that detect student confusion or cognitive fatigue.
• Interactive gaming NPCs that react to real human vocal intonation rather than text syntax.
• Real-time Affective Generative UI: dampening visual clutter and slowing animation springs when user agitation is detected.

The Russell 2D Circumplex Model of Affect

                       High Arousal (+1.0)
                         [Agitated / Alert]
                               │
       Angry / Frustrated       │       Delighted / Excited
       (Valence: -0.8, Arousal: +0.7)  (Valence: +0.8, Arousal: +0.7)
                               │
Negative Valence (-1.0) ────────┼──────── Positive Valence (+1.0)
                               │
       Depressed / Fatigued     │       Calm / Serene
       (Valence: -0.6, Arousal: -0.6)  (Valence: +0.7, Arousal: -0.5)
                               │
                       Low Arousal (-1.0)
                         [Sleepy / Calm]

Multimodal Real-Time Affective Pipeline (TypeScript)

// src/services/affective-engine.ts
export interface AffectVector {
  valence: number; // -1.0 (Very Negative) to +1.0 (Very Positive)
  arousal: number; // -1.0 (Passive/Sleepy) to +1.0 (High Energy/Agitated)
  confidence: number;
}

export interface UIAdaptationContract {
  theme: 'calm-slate' | 'high-contrast-focus' | 'vibrant-energetic';
  cognitiveDensity: 'minimal' | 'balanced' | 'dense';
  animationSpringDamping: number; // Stiff vs Soft
}

export class AffectiveEngine {
  // Evaluates multimodal telemetry and computes dynamic UI parameters
  static computeUIAdaptation(affect: AffectVector): UIAdaptationContract {
    // Agitated or Frustrated user (High Arousal, Negative Valence)
    if (affect.arousal > 0.4 && affect.valence < -0.2) {
      return {
        theme: 'calm-slate',
        cognitiveDensity: 'minimal',
        animationSpringDamping: 40, // Gentle, non-distracting springs
      };
    }

    // Fatigued or Disengaged user (Low Arousal, Negative Valence)
    if (affect.arousal < -0.3 && affect.valence < 0.1) {
      return {
        theme: 'high-contrast-focus',
        cognitiveDensity: 'minimal',
        animationSpringDamping: 25,
      };
    }

    // Standard balanced baseline
    return {
      theme: 'vibrant-energetic',
      cognitiveDensity: 'balanced',
      animationSpringDamping: 20,
    };
  }
}

Real-Time Gemini Multimodal Live Audio Stream

// src/lib/gemini-affective-live.ts
import { GoogleGenAI } from '@google/genai';

const ai = new GoogleGenAI({ apiKey: process.env.GEMINI_API_KEY! });

export async function createAffectiveLiveSession() {
  const session = await ai.models.createWebsocketSession({
    model: 'gemini-2.5-flash',
    config: {
      generationConfig: {
        responseModalities: ['AUDIO'],
        speechConfig: {
          voiceConfig: { prebuiltVoiceConfig: { voiceName: 'Puck' } }
        }
      },
      systemInstruction: `You are an empathic customer resolution sentinel.
Analyze user vocal pitch, breathiness, and cadence directly from raw audio frames.
If the customer sounds irritated, lower your pitch, slow your speaking rate by 20%, and maintain a supportive, calm demeanor.`
    }
  });

  return session;
}

<a name="bahasa-indonesia"></a>

Bahasa Indonesia

Integrasi Orkestrasi

• Orkestrator Hulu: Beroperasi di bawah arahan brainstorming dan zero-to-prod-orchestrator (Fase 4 - Agen AI & Pipeline Multimodal) untuk mengintegrasikan komputasi afektif real-time.
• Skill Hilir & Kerja Sama:
  • voice-ai-realtime-agent — Transport audio WebRTC dua arah dan penangkapan audio latensi rendah (<300ms).
  • ephemeral-generative-ui-architect — Menyesuaikan tata letak komponen UI secara reaktif terhadap AffectVector.
  • ui-ux-pro-max — Menjamin kontras tema UI afektif mematuhi WCAG 2.2 AA tanpa transisi yang mengejutkan.
  • ai-llm-integration-expert — Penanganan token multimodal via Gemini Multimodal Live API.
  • session-memory-manager — Menyimpan histori interaksi afektif dan vektor kondisi emosional pengguna.

Deskripsi

Panduan komprehensif untuk Komputasi Afektif, agen percakapan empatik, dan sintesis sentimen biometrik waktu-nyata menggunakan token multimodal native (Gemini Multimodal Live API, WebRTC audio streaming dua arah). Menafsirkan variasi nada suara, tempo bicara, jeda nafas, dan mikro-ekspresi wajah tanpa ketergantungan pada transkrip teks biasa. Memetakan sinyal emosi ke dalam ruang koordinat 2D Valence-Arousal (Model Sirkumpleks Russell) untuk secara dinamis mengatur prosodi suara agen, kecepatan interaksi, dan tata letak antarmuka pengguna (UI).

Kondisi Pemicu

• Membangun layanan pelanggan empatik, konsultasi medis awal, atau agen pendamping krisis.
• Platform EdTech adaptif yang mendeteksi kebingungan atau kelelahan kognitif siswa saat belajar.
• Karakter NPC dalam game interaktif yang bereaksi terhadap nada suara manusia secara nyata.
• Adaptasi UI Generatif: mengurangi kepadatan visual dan memperlambat animasi ketika terdeteksi kepanikan atau kemarahan pengguna.

Pola Implementasi Respon UI Adaptif di Frontend

// src/components/affective/AdaptiveContainer.tsx
'use client';

import { useMemo } from 'react';
import { AffectiveEngine, AffectVector } from '@/services/affective-engine';

export function AdaptiveContainer({ affect, children }: { affect: AffectVector; children: React.ReactNode }) {
  const adaptation = useMemo(() => AffectiveEngine.computeUIAdaptation(affect), [affect]);

  return (
    <div
      className={`min-h-screen transition-colors duration-700 ${
        adaptation.theme === 'calm-slate'
          ? 'bg-slate-900 text-slate-100'
          : adaptation.theme === 'high-contrast-focus'
          ? 'bg-black text-amber-300'
          : 'bg-white text-slate-900'
      }`}
    >
      <main className={`mx-auto p-6 ${adaptation.cognitiveDensity === 'minimal' ? 'max-w-2xl' : 'max-w-5xl'}`}>
        {children}
      </main>
    </div>
  );
}

3 Prinsip Dasar Komputasi Afektif

1. Analisis Langsung pada Sinyal Audio (Zero-STT Latency): Jangan ubah suara menjadi teks terlebih dahulu. Analisis dilakukan langsung pada aliran PCM audio agar intonasi dan desahan nafas tidak hilang.
2. Koordinat Kontinu Valence-Arousal: Jangan batasi emosi pada label kaku ("senang", "sedih"). Gunakan nilai desimal kontinu antara -1.0 hingga +1.0 untuk modulasi yang halus.
3. Peredaman UI Ramah Stres: Ketika pengguna frustrasi, jangan tampilkan notifikasi pop-up animasi yang ramai; sederhanakan tampilan hanya pada tindakan penyelesaian utama.