Notes: Philosophy of Agency and Agentic AI
Key Concept: Diachronic Organization
Definition: The coordination of rational behavior and choices across time (as opposed to synchronic, which is a single moment in time).
The Core Problem: How does an entity ensure its past decisions, present actions, and future goals align without constant re-evaluation?
The Human Solution: Plans. Humans use stable, commitment-based plans to anchor their behavior, preventing decision paralysis and allowing long-term tracking of complex goals.
The Bratman Hierarchy of Agency
1. Diachronic Organization (Individual)
Managing your own continuity across time.
Example: Deciding on Monday to attend a lecture on Friday, and your Friday-self honoring that commitment.
2. Shared Action (Social / Interpersonal)
Meshing individual sub-plans with another independent actor to achieve a mutual goal.
Example: Two people coordinating to paint a room or write a paper together.
3. Institutional Organization (Structural)
Scaling planning up to massive, enduring structures that maintain policy, intent, and identity over generations.
Example: Universities, corporations, or legal frameworks.
The BDI Framework
(Belief-Desire-Intention)
Element | Description | Role in Planning
Beliefs | Information about the world. | Provides the factual baseline.
Desires | Broad, potential options or wants. | Can be conflicting or passive.
Intentions | Commitments to action. | The fixed points that drive execution.
Takeaway: Intentions provide mental inertia. They resist reconsideration unless significant new data arises, allowing agents to remain stable over time.
Why It Matters for AI Agents
The Gap: Current LLMs are fundamentally synchronic (stateless, reactive prompt-by-prompt).
The Goal: True agentic AI requires diachronic architecture—the ability to hold a long-term objective, break it down into sequential sub-tasks, monitor its own progress over days or weeks, and self-correct without losing its original intent.
