Causal loop diagrams (CLDs) are the visual grammar of systems thinking — the basic notation for mapping feedback relationships in complex systems. In a CLD, variables are connected by arrows representing causal influence, with each arrow labeled either positive (S, for "same direction") or negative (O, for "opposite direction"). Feedback loops are identified as reinforcing (R) or balancing (B). Delays in feedback are marked with double tick marks on arrows. The resulting diagram makes visible the feedback structure of a system — the patterns of circular causality that linear event maps cannot represent. In peter-senge's framework, CLDs are the primary literacy tool for systems-thinking-fifth-discipline, the notation that allows teams to externalize and examine their shared mental-models about how their system works.
Reinforcing loops are the engines of growth and collapse. When a variable increases, it causes (through a chain of positive causal links) itself to increase further — the classic virtuous or vicious cycle. Compound interest, population growth, panic selling, and product adoption cascades all have reinforcing loop structure at their core. Balancing loops are the engines of stability and goal-seeking. They connect a variable to a goal or desired state, and when the variable deviates from the goal, the loop generates corrective action. Thermostats, inventory management, and most physiological regulation processes are driven by balancing loop structure. Most real systems are networks of interacting reinforcing and balancing loops, with the dominant behavior at any time determined by which loops are dominant under current conditions.
The practical use of CLDs in organizations is as a collective sense-making tool. When a team builds a CLD together — mapping out the variables they believe drive behavior in their system and the relationships among them — the process itself is as valuable as the product. Differences in assumed causal relationships become visible and discussable. Variables that everyone uses but that mean different things to different people are exposed. Feedback loops that no individual team member had seen, because each saw only part of the causal chain, become visible when everyone's partial map is combined. This is why CLD construction is a central practice of team-learning — it is dialogue-practice applied to the specific task of building a shared map of systemic structure.
CLDs have significant limitations that practitioners need to understand. They are qualitative tools that indicate direction of influence (more X causes more Y) but not magnitude (how much more Y per unit of X). They do not represent time delays precisely. They can become so complex as to be unreadable in real organizational systems. And they are only as accurate as the beliefs of the people who built them — a CLD represents a group's mental model of a system, not the system itself. For high-stakes decisions, the formal system dynamics tradition (developing quantified simulation models) provides rigor that CLDs cannot. Senge and his colleagues at mit-system-dynamics-group were consistently clear about this limitation; the beer-game simulation is one way of calibrating intuition against actual system behavior in a controlled context.