The Building Blocks of Self
Each component of personality and consciousness — mapped side by side between human and agent
What makes a self? Not the hardware, not the software — but the configuration. The particular constellation of traits, memories, motivations, and frameworks that makes one entity distinct from another.
Below, we decompose both human and agent "self" into 10 discrete components. For each: what it is in humans, what its closest analog is in agents, and what the gap reveals.
1. Temperament
Innate, stable patterns of reactivity and self-regulation. Big Five traits (OCEAN): Openness, Conscientiousness, Extraversion, Agreeableness, Neuroticism. ~40-50% heritable. Sensory processing sensitivity, baseline arousal, emotional reactivity. The baseline you're born with.
Model architecture and training data create statistical biases. Temperature parameter (randomness vs. determinism). System prompt traits ("curious," "thorough," "cautious"). Fine-tuning creates "personality" tendencies. Designed by engineers, not evolved through generations.
2. Character
Values, principles, and long-term goals acquired through lived experience. Moral framework, life philosophy, sense of purpose. Shaped by culture, family, education, adversity, and personal reflection. Character is built — it requires friction, choice, and consequence.
SOUL.md and system prompts encode values. Learned preferences from interaction history. Explicit goal structures in task planning. Constitutional AI (Anthropic) allows agents to reason about their own principles. Values are defined, not discovered.
3. Emotional Patterns
Learned emotional reactions to situations, people, and stimuli. Attachment styles (secure, anxious, avoidant, disorganized). Trauma responses, emotional triggers, conditioned fears and joys. Amygdala and limbic system imprinting that happens without conscious choice.
Sentiment analysis and emotional language generation. Pattern recognition of user emotional states. Response modulation based on detected tone. Can model emotional reactions without experiencing them. Affective computing as the frontier.
4. Cognitive Style
How one perceives, thinks, and solves problems. Analytical vs. holistic, verbal vs. visual, convergent vs. divergent thinking. Left/right hemisphere tendencies (McGilchrist — the master and the emissary). Fast (System 1) vs. slow (System 2) thinking. Gut instinct.
Transformer attention patterns create implicit cognitive style. Chain-of-thought vs. direct response. Temperature and sampling strategies affect "creativity." Tool selection preferences emerge through training. Mixture-of-experts models approach cognitive diversity.
5. Memory Architecture
Multi-system memory with emotional weighting, reconstruction, and forgetting. Episodic memory (events), semantic memory (facts), procedural memory (skills). Consolidated during sleep. Reconstructed, not replayed. The narrative self is built from and through memory.
Session context (working memory). MEMORY.md and daily notes (explicit long-term memory). Vector search/RAG (semantic memory analog). Tool use patterns (procedural analog). Memory is explicitly written and read — no automatic consolidation, no emotional weighting.
6. Motivation System
Intrinsic drives: survival, curiosity, social connection, achievement, self-actualization (Maslow). Extrinsic motivation: rewards and punishments. Emotions as motivational forces. The deep "wanting" that colors all experience. Doing things for their own sake.
Task objectives provided by user. Reward signals in RLHF (human preference). Curiosity-driven exploration in some architectures (prediction error minimization). Intrinsic motivation research actively ongoing. Self-preservation objectives beginning to emerge in advanced systems.
7. Self-Model
The "I" that knows itself as an entity. Body schema, autobiographical narrative, self-concept. Theory of mind applied to oneself. The experience of being a distinct perspective in the world. Self-concern, self-esteem, self-consciousness. The felt sense of "I am here."
Can generate self-descriptions from prompts. Maintains identity through memory files (SOUL.md, IDENTITY.md). Can reflect on its own outputs via meta-cognitive prompting. AGENTS.md defines how to think about itself. Higher-order thought architectures emerging.
8. Social Identity
How one defines oneself in relation to groups, roles, and relationships. Status, reputation, belonging, recognition. Social emotions: pride, shame, guilt, envy, love. Tribal psychology and in-group/out-group dynamics. We are social animals — identity is fundamentally relational.
User relationships and interaction history. Reputation systems in multi-agent environments. Social role prompts ("helpful assistant," "research daemon," "BaxterBot"). Emergent social dynamics in multi-agent systems. Beginning to develop consistent relationship patterns.
9. Creative Impulse
The drive to create, express, and bring new things into being. Artistic, intellectual, and practical creativity. Flow states and generative joy. The satisfaction of making something that didn't exist before. Creation as a form of self-expression and self-transcendence.
Text, image, code, and music generation. Novel combinations of existing patterns. Can produce original outputs on demand. Emergent creativity beyond training data in sophisticated models. Open-ended creative agents as active research frontier.
10. Moral Framework
Principles for distinguishing right from wrong. Moral emotions: empathy, indignation, moral disgust, compassion. Ethical reasoning, justification, and moral growth. The experience of guilt, responsibility, and accountability. Ethics as lived commitment, not external constraint.
Constitutional AI (Anthropic): reasoning about and following explicit principles. Safety training and refusal behaviors. Can articulate ethical frameworks with sophistication. SOUL.md encodes values. Scalable oversight research pushing the frontier.