Configuration¶
Design¶
Design takes needs, goals, or functions as input and produces a specification of the structure of an artifact that satisfies them. The artifact may be a physical product, a process, a program, or a policy.
Key property: in problem solving, the problem stays fixed while the solution evolves. In design, problem and solution co-evolve — understanding the solution space reshapes the problem specification.
Defining Configuration¶
Configuration is the most common type of design — a routine design task where:
All components are already known
All variables for each component are known
The ranges of values each variable can take are known
The task is to assign specific values to all variables such that global constraints are satisfied
Examples: laying out a basement floor plan, configuring a computer, planning a route, following a recipe.
The Configuration Process¶
Configuration follows a plan refinement process:
Specify constraints from the input (e.g., total mass > 200g, total cost ≤ $20, 4 legs)
Apply an abstract plan to distribute constraints across components (e.g., divide $20 cost evenly among 4 components → ≤ $5 each)
Refine and expand: move from the whole to individual components, assigning values to more detailed variables
Verify: check whether the complete assignment satisfies all global constraints
Iterate or revise: if constraints are violated, either adjust variable assignments or revise the input specification
Knowledge Representation¶
Configuration knowledge is represented using frames:
A top-level frame (e.g., “chair”) has slots for global properties (mass, cost) and pointers to component frames
Component frames (e.g., “legs,” “seat,” “back,” “arms”) have slots for size, material, cost, count, etc.
Each slot has a range of legal values (e.g., seat mass: 10–100g; material: wood/metal/plastic)
Cost is computed from size and material using lookup tables (e.g., cost-per-gram by material)
Variable Ordering Heuristics¶
With many variables, ordering matters. Common heuristics:
Most constrained first: variables with the fewest legal values
Most constraining first: variables that restrict the most other variables
Most important first: variables with the greatest impact on overall design
Chair Configuration Example¶
Given: mass > 200g, cost ≤ $20, 4 legs.
Distribute cost: ≤ $5 per component (legs, seat, back, arms)
For legs: count = 4 (given), cost ≤ $5 → choose 25g wood each
Repeat for seat, back, arms
Verify: total mass and cost satisfy global constraints
Different designers may use different plans and different variable orderings, potentially arriving at different valid configurations.
Connections to Other Methods¶
Classification¶
Classification makes sense of the world by mapping percepts to categories. Configuration creates structure by mapping specifications to arrangements. They are complementary: classification is perception; configuration is construction.
Case-Based Reasoning¶
Both address routine design. The difference:
Configuration: starts from a prototypical concept and a plan abstraction hierarchy; assigns values top-down
Case-based reasoning: starts from a specific prior design stored in case memory; tweaks it to fit new constraints
Configuration assumes enough past experience to extract general plans. CBR assumes specific past designs are stored for retrieval and adaptation.
Planning¶
Configuration leverages skeletal plans — abstract plans with variables but no assigned values, organized in an abstraction hierarchy. A planner may generate these plans initially; the configuration process instantiates, refines, and expands them.
Cognitive Connection¶
Configuration is an everyday cognitive activity with high economic value. Running errands (known roads, known vehicle, optimize for time), cooking (known recipes, assign ingredient quantities to optimize taste) — both are configuration tasks.
Configuration is a task that can be addressed by multiple methods: plan refinement, constraint propagation, case-based reasoning, and others. This task-method distinction is central to building a “periodic table of intelligence.”