Intelligence Architecture
The Systems No One Wrote Down
Mapping a company's unwritten order is not preparation for AI work. It is the design work — and skipping it lets a second invisible order form.
·
Oliver Wrede
Editor · Intelligence Architecture
Oliver Wrede writes and teaches on interface design, knowledge systems, and the architecture of intelligence in organizations. He is interested in how humans, institutions, and machines reason together — and how design shapes the quality of that reasoning.
13 pieces by Oliver Wrede
Foundations
How an AI Output Acquires Institutional Weight
State v. Loomis shows how reports, rules of use, and accountable roles give a risk score institutional weight.
·
Foundations
Testing a New Definition of Intelligence Against AlphaZero
Bernard Jennings draws an architectural boundary around intelligence. His treatment of AlphaZero makes the proposal concrete and exposes its unresolved burden of proof.
·
Foundations
Stuart Russell and the Conditions for Useful Correction
The Off-Switch Game explains why an agent may benefit from waiting. Evidence from a working decision system shows what an organization must add.
·
Foundations
Stafford Beer and What Project Cybersyn Could Control
A telex network proved useful during Chile's October 1972 strike. The larger cybernetic system associated with Cybersyn never became routine organizational practice.
·
Foundations
What Must Be Decided Before Choosing an AI Model
Three bodies of theory with very different standing clarify the work that comes first: setting purpose, detecting material deviation, and making correction effective.
·
Commentary
When Does a System Count as Intelligent?
Bernard Jennings proposes a test based on a system's internal architecture. Organizations still have to decide what authority, if any, its output should carry.
·
Artificial Intelligence
How a Risk Score Changed Hotline Work
Allegheny County tied the AFST to a default rule requiring investigation and reorganized screening around that rule. Policy made the model consequential.
·
Artificial Intelligence
What Makes an AI Output Worth Trusting
In the Horizon case, a disputed accounting shortfall became a debt and eventually a suspicion. What mattered was whether it could still be examined and corrected effectively.
·
Design Theory
When False Calculations Become Facts
Robodebt distributed aggregate income across periods that had not been observed. The process then treated the resulting figures as facts.
·
Foundations
What Is Intelligence Architecture?
A working term for useful and responsible collaboration between people and AI systems inside organizations.
·
Artificial Intelligence
When a Sepsis Score Becomes an Alert
A hospital committee turned a prediction into a page. External validation shows why every step from score to clinical work needs deliberate design.
·
Organizations
Columbia: Why a Warning Failed to Prompt a New Mission Decision
The investigation shows how a debris strike was recorded and analyzed without remaining an open safety question in the mission process.
·