DataTerminology.comNEW
Also known as: Agent design
The structural design of components and control mechanisms that govern an intelligent agent’s perception, reasoning, and action.
DESCRIPTION
Agent architectures specify how agents sense their environment, reason about goals, and execute actions. Examples include reactive (stimulus-response), deliberative (model-based planning), hybrid (layered), and cognitive architectures like BDI, Soar, and ACT-R. Choices impact autonomy, interpretability, and robustness. Modern LLM-based agents integrate planners, memory, and tool-use modules orchestrated by prompts and policies. Good architectures align with tasks, latency constraints, and safety requirements.
ALTERNATIVE DEFINITIONS
Wikipedia
A cognitive architecture is a blueprint for an intelligent agent and a hypothesis about fixed structures that provide a mind.
Frames architecture as both design and scientific hypothesis.
Soar Cognitive Architecture
Soar is a general cognitive architecture for developing systems that exhibit intelligent behavior.
Defines architecture through a concrete, influential implementation.
RELATED TERMS
Agent BDI Architecture Reactive Architecture Deliberative Architecture Planning Reinforcement Learning Knowledge Representation
COMPETING TERMS
Monolithic Architecture
RELATIONS
- Narrower: Cognitive Architecture
KNOWLEDGE AREAS
AI Data Architecture Software Engineering
TARGET AUDIENCE
AI Engineer Software Developer Data Architect MLOps Engineer
EXTERNAL SOURCES
- Wikipedia - Cognitive architecture
- Soar - General-purpose cognitive architecture
- ACT-R - Cognitive architecture and resources
- Wikipedia - BDI agent model overview
TOP QUESTIONS
What is the purpose of an Agent Architecture?
It organizes perception, memory, reasoning, and action, enabling agents to act autonomously and reliably.
Category: Benefits
How do reactive and deliberative Agent Architectures differ?
Reactive are fast, stimulus-driven; deliberative plan using models. Hybrids combine both.
Category: Concepts
What risks exist when designing Agent Architectures?
Overcomplexity, poor latency control, and unsafe behaviors without guardrails or verification.
Category: Risks
How do LLMs change Agent Architecture design?
They add components for prompting, tool use, memory, and planning, requiring orchestration and safety layers.
Category: Topic
What is BDI in Agent Architecture?
A model where beliefs, desires, and intentions structure reasoning and action selection.
Category: Example
How to evaluate an Agent Architecture?
Assess task success, sample efficiency, robustness, interpretability, latency, and safety compliance.
Category: Testing
When to choose a reactive Agent Architecture?
For strict real-time constraints and simple dynamics where planning overhead is prohibitive.
Category: Use cases
What governance applies to Agent Architectures?
Define responsibility boundaries, audit logs, and fail-safe mechanisms for autonomous decisions.
Category: Governance
How does memory factor into Agent Architecture?
Episodic and semantic memory modules support context, learning, and longer-horizon planning.
Category: Concepts