Carl Hewitt — The Actor Model

Carl Hewitt — The Actor Model

§SELF — thinker-carl-hewitt

What this page is: A profile of Carl Hewitt and the Actor Model of computation. What it explains: The Actor Model — what it is, how it works, and why it matters for concurrent systems. Why read it: To understand the fundamental unit of concurrent computation and how it applies to OIP.

What the Actor Model Is

The Actor Model is a theory of computation where the fundamental unit is an "actor" — an entity that receives messages, processes them, and sends messages to other actors. Carl Hewitt introduced it in 1973 in his paper "A Universal Modular Actor Formalism for Artificial Intelligence."

An actor is an independent, concurrent, autonomous entity. It encapsulates state (internal data not accessible from outside), communicates only by message passing (no shared memory, no direct function calls), and processes one message at a time (messages handled sequentially within an actor, but actors run concurrently with each other).

Actors have three primitive operations: they can create new actors, send messages to actors they know about (their "acquaintances"), and change their own behavior for the next message they receive.

Why It Matters

Before the Actor Model, concurrent programming relied on shared memory and locks — which leads to race conditions, deadlocks, and non-deterministic behavior. The Actor Model eliminates these problems by design: since actors share no state and communicate only through messages, there is no shared memory to corrupt and no locks to deadlock.

The model is "universal" in Hewitt's terms because any concurrent computation can be expressed as a system of actors. It maps directly to physical reality: actors are like biological cells (independent, communicating through signals) or human agents (autonomous, message-passing).

The Key Idea

The key idea is that concurrency should be the default, not an afterthought. In the Actor Model, every actor runs concurrently. Sequential processing is a special case (one actor handling one message). The model does not add concurrency to a sequential foundation — it builds sequentiality on top of a concurrent one.

This inverts the traditional approach, where a single sequential processor is assumed and concurrency is bolted on through threads, locks, and shared memory.

What Hewitt Got Right

• Message passing is safer than shared memory. By forbidding shared state, the Actor Model eliminates entire categories of concurrent programming bugs.
• Encapsulation should be mandatory, not optional. An actor's internal state is inaccessible from outside. There is no "public field" or "global variable" in the actor world.
• Concurrency is fundamental. The model treats concurrent, independent agents as the base case, not a complication of sequential logic.
• Creation is a primitive. The ability to spawn new actors is built into the model, making dynamic system growth natural.
• Behavior change is explicit. An actor specifies how it will handle the next message, making state transitions clear and deterministic.

What Hewitt Got Wrong or Left Unfinished

• No canonical formal semantics. Hewitt's original formulation was broad and philosophical. Formal semantics (such as those developed by Agha and others) came later and imposed restrictions Hewitt did not originally specify.
• Message delivery guarantees. The original model did not specify whether messages are guaranteed to arrive, in what order, or within what time frame. Real systems must make these guarantees explicit.
• Actor addresses and mobility. The model says actors send messages to actors they "know about," but the mechanism for discovering and sharing addresses was underspecified.
• Practical implementation challenges. Efficient actor implementations (scheduling, memory management, message routing) were left as engineering problems, not addressed by the theory.

How It Connects to Other Ideas

• Object-oriented programming: Actors preceded and influenced objects. Alan Kay cited the Actor Model as an influence on Smalltalk. The difference: objects typically use synchronous method calls (shared stack), while actors use asynchronous message passing.
• Biological computation: Hewitt explicitly compared actors to cells. Both are encapsulated, communicate by signaling, and operate concurrently. This analogy has been explored in biological computing and membrane computing.
• OIP (Open Integration Platform): An OIP object is an actor. It receives a message (the invocation), processes it (the runner executes), and may send messages (the receipt links to other objects). The stateless dispatch door means every invocation is a message to an actor. The model fits exactly: no shared state, message passing, encapsulated behavior.

Sources

• Hewitt, C., Bishop, P., Steiger, R. "A Universal Modular Actor Formalism for Artificial Intelligence." IJCAI, 1973.
• Agha, G. Actors: A Model of Concurrent Computation in Distributed Systems. MIT Press, 1986.

---

Up the tree

• OIP root — protocol root, zero-context entry
• Thinker Reference hub — full hierarchy map
• Thinkers shelf — siblings on this shelf
• Voxel graph article — how pages link as voxels
• Self-describing protocol

Related on this shelf

• Alan Kay — The Big Idea Is Messaging
• Alfred North Whitehead — Process and Reality
• J.L. Austin and John Searle — Speech Acts
• Barbara Liskov — Abstract Data Types and Distributed Consensus
• Bram Cohen — BitTorrent and Content-Addressed Protocol Design
• Butler Lampson — Protection and Access Control
• Charles Sanders Peirce — Signs, Abduction, and Pragmatism
• Doug Engelbart — Augmenting Human Intellect

Machine surfaces

• Public page: https://miscsubjects.com/a/thinker-carl-hewitt
• JSON article: https://miscsubjects.com/api/articles/thinker-carl-hewitt
• OIP ask: https://miscsubjects.com/api/dispatch?ask=Carl%20Hewitt%20%E2%80%94%20The%20Actor%20Model