mala.dev
Home/Decision Graphs/Decision Graph vs Knowledge Graph
DECISION GRAPHS · DECISION GRAPH VS KNOWLEDGE GRAPH

Decision Graph vs Knowledge Graph

Knowledge graphs map what entities exist and how they relate. Decision graphs map what decisions were made and why. For AI governance, you need the second — and most AI stacks only have the first.

Knowledge graphs and decision graphs are both graph-structured data models, but they capture fundamentally different things. A knowledge graph represents entities (people, places, organizations, concepts) and the relationships between them. A decision graph represents decision events — what an AI agent decided, in what context, under which policy, and with what cryptographic proof. Understanding this distinction is critical for AI teams building governance infrastructure.

What is a Knowledge Graph?

A knowledge graph is a structured representation of facts: entities, their attributes, and their relationships. Knowledge graphs power search engines, recommendation systems, and RAG (retrieval-augmented generation) architectures. They answer the question: 'What do we know about X, and how does X relate to Y?'

What is a Decision Graph?

A decision graph is a structured representation of decision events: what was decided, by which agent, in what context, under which policy, and sealed with cryptographic proof. Decision graphs answer the question: 'What did the AI agent decide at this moment, and can you prove it?' They are dynamic — a new node is created for every production AI decision — and immutable once sealed.

How They Work Together in AI Systems

In a sophisticated AI agent architecture, knowledge graphs and decision graphs are complementary. The knowledge graph provides context for agent decisions — what the agent knows about a customer, a patient, or a regulatory requirement. The decision graph records what the agent decided given that context — creating the accountability record. Think of the knowledge graph as the agent's memory, and the decision graph as the audit trail of what the agent did with that memory.

Why Enterprises Need Both

Most enterprise AI teams have invested heavily in knowledge graphs for RAG — improving the quality of AI agent outputs by grounding them in organizational knowledge. But they have not yet built the decision graph layer — the accountability infrastructure that makes those improved outputs defensible. As AI agents take on consequential tasks in regulated industries, the decision graph becomes as essential as the knowledge graph. One provides context; the other provides proof.

Frequently Asked Questions

Can a knowledge graph serve as an AI audit trail?
No. A knowledge graph records facts about the world — entity relationships and attributes. It does not record decision events — what an AI agent decided at a specific moment, with what context, and under which policy. For audit trail purposes, you need a decision graph: a sealed, timestamped record of each individual AI decision.
Does Mala use knowledge graphs internally?
Mala's Wisdom Graph — our institutional memory component — is a precedent graph that stores organizational decision history. When an agent is about to make a decision, the Wisdom Graph surfaces relevant precedents and applicable policies. The decision the agent makes is then recorded in the decision graph. The Wisdom Graph is context infrastructure; the decision graph is the accountability record.
What makes a decision graph different from a workflow graph?
A workflow graph defines the steps and transitions in a business process — what happens in what order under what conditions. A decision graph records actual decision events within those workflows — what the AI agent decided at each decision point, with full context and cryptographic seal. Workflow graphs are process definitions; decision graphs are execution records.