Concepts · Humans and machines
Concepts
Foundational concepts for studying memory as a cognitive, behavioral, organizational, and computational phenomenon.
Summary Capsule
Explore ACT-R, retrieval, agent memory, and organizational context as related but distinct ideas.
This collection explains what cognitive architectures can contribute to AI-system design without treating engineering patterns as models of the human mind.

Canonical Memory Glossary
An up-to-date glossary of memory terms across science, computing, AI, and technology.

What Is ACT-R Memory Architecture?
How ACT-R connects declarative memory, buffers, activation, and production rules, with a worked example and practical lessons for AI builders.

What Is Soar Cognitive Architecture?
An introduction to Soar’s states, operators, preferences, and memory systems, with a support-case example and a reading path through the cluster.

ACT-R Memory for Agents
ACT-R-style agent memory uses goals, active context, retrieved chunks, and production rules to help AI agents choose what to do next.

What Is a Memory-Native Organization?
A delivery handoff shows how a memory-native organization carries lessons across people, shifts, and AI agents—and revises them when the work changes.

RAG vs ACT-R Memory for AI
RAG retrieves evidence for generation; AI memory lets prior experience change what becomes salient and how a system responds or acts next.
ACT-R Architecture Chapters
Read the ACT-R memory sequence
- 1What Is ACT-R Memory Architecture?A plain introduction to ACT-R memory architecture: goals, buffers, chunks, production rules, activation, and action selection.
- 2Ebbinghaus and ForgettingHow nonsense syllables, savings scores, and the forgetting curve made memory measurable.
- 3Chunks and Declarative MemoryHow ACT-R represents facts and experiences as retrievable chunks with graded activation.
- 4Production RulesHow procedural memory selects actions through condition-action rules and utility estimates.
- 5ACT-R and AI MemoryHow chunks and action rules can guide datasets for content and decision models.
SOAR Architecture Chapters
Read the SOAR problem-solving sequence
- 1What Is SOAR Cognitive Architecture?A simple guide to SOAR cognitive architecture: states, operators, goals, preferences, impasses, and learning.
- 2Working Memory in SOARHow SOAR keeps the current problem state readable through working memory elements.
- 3Operators in SOARHow SOAR represents possible next steps as operators that change the current state.
- 4Preferences in SOARHow preferences help SOAR choose between matching operators.
- 5Impasses in SOARHow SOAR handles stuck points by creating subgoals instead of guessing.
- 6Chunking in SOARHow SOAR learns new production rules from solved subproblems.
- 7Semantic and Episodic Memory in SOARHow SOAR separates general knowledge from remembered experience.
- 8SOAR and AI AgentsWhat agent memory can borrow from SOAR without pretending software has a human mind.
Concept Article Index
Latest concepts and chapters
22 published resources
- All Kinds of Memory in AI & SoftwareA layered guide to RAM, ROM, VRAM, bandwidth, context windows, KV caches, model weights, agent memory, and experience-shaped Emergent Memory.Concept article
- Canonical Memory GlossaryAn up-to-date glossary of memory terms across science, computing, AI, and technology.Core concept
- What Is a Memory-Native Organization?A delivery handoff shows how a memory-native organization carries lessons across people, shifts, and AI agents—and revises them when the work changes.Core concept
- How Achiral's ACT-R Memory Remembers and ForgetsA deployment goes wrong, a workaround succeeds, and then the process changes. How Achiral’s ACT-R memory uses past experience without getting stuck in it.Concept article
- Organic AI MemoryHow organic AI memory connects probabilistic recall with continuous or periodic learning, using selected experience to adjust a model’s future behavior.Concept article
- Emergent Memory vs Storage PrimitiveStorage preserves records; Emergent Memory changes their future influence. A concrete example connects ACT-R, probabilistic recall, and empirical evaluation.Concept article
- RAG vs ACT-R Memory for AIRAG retrieves evidence for generation; AI memory lets prior experience change what becomes salient and how a system responds or acts next.Core concept
- ACT-R Memory for AgentsACT-R-style agent memory uses goals, active context, retrieved chunks, and production rules to help AI agents choose what to do next.Core concept
- Declarative vs Procedural Memory for AI AgentsDeclarative memory stores what an AI agent knows. Procedural memory stores how it acts. Reliable agents need both facts and learned execution rules.Concept article
- Soar and AI AgentsHow AI agents can borrow Soar’s explicit state, operator selection, impasse handling, and learning, with a worked support-case example.SOAR chapter
- Soar Semantic and Episodic MemoryHow Soar stores general knowledge and recorded episodes, retrieves them into working memory, and keeps both distinct from learned production rules.SOAR chapter
- What Is Soar Chunking?How Soar learns conditional production rules from substate results, what those rules preserve, and how chunking differs from episodic memory and reinforcement learning.SOAR chapter
- Impasses and Subgoals in SoarHow Soar represents stalled decisions as impasses, creates substates to work on them, and returns results that let the original task continue.SOAR chapter
- Preferences in SoarHow Soar preferences distinguish candidate actions, comparisons, and requirements, and what happens when that knowledge cannot resolve a decision.SOAR chapter
- Operators in SoarHow Soar proposes, selects, and applies operators, with a password-reset example that separates candidate actions, selected actions, and observed results.SOAR chapter
- What Is Soar Working Memory?How Soar represents current state as working memory elements, updates that state across decisions, and distinguishes selected actions from observed results.SOAR chapter
- What Is Soar Cognitive Architecture?An introduction to Soar’s states, operators, preferences, and memory systems, with a support-case example and a reading path through the cluster.SOAR chapter
- ACT-R and AI MemoryHow ACT-R’s distinction between declarative knowledge and action rules can guide dataset distillation for content and decision models.ACT-R chapter
- Production Rules and Procedural MemoryHow ACT-R production rules match the current state, compete for selection, and learn from outcomes, with a worked support-agent example.ACT-R chapter
- Chunks and Declarative MemoryHow ACT-R represents knowledge as chunks, uses cues to request it, and models retrieval through activation, with a worked support-case example.ACT-R chapter
- Ebbinghaus Forgetting CurveWhat Ebbinghaus measured through relearning, what savings means, and how AI builders can distinguish retrieval decay from deleting a record.ACT-R chapter
- What Is ACT-R Memory Architecture?How ACT-R connects declarative memory, buffers, activation, and production rules, with a worked example and practical lessons for AI builders.ACT-R chapter
Memory and continuity in AI systems
Memory is not a single technical object. In cognitive science it concerns mechanisms of learning and recall; in AI engineering it often refers to persistent context, retrieval, or state management. This collection separates those meanings before asking how they can inform practical systems.
Achiral applies these ideas to a shared operational-context layer for AI systems. That layer is an engineering design, not a model of human memory or cognition.
| Perspective | Central question |
|---|---|
| Cognitive architecture | What must be represented, retrieved, activated, and acted on for a system to behave with continuity rather than isolated response? |
| Behavioral adaptation | How does repeated experience change future choices, priorities, habits, and interventions without requiring every rule to be restated? |
| Organizational memory | How do decisions, context, exceptions, and handoffs become durable shared state instead of disappearing across tools and conversations? |
| Computational retrieval | When should a system search, rank, reinforce, forget, or withhold context so memory remains useful instead of becoming noise? |
| Governance and safety | How can compounding memory stay permissioned, inspectable, reversible, and aligned with human review as it begins to shape action? |