What are leverage points?

Donella Meadows ranked twelve kinds of intervention by how much each moves a system, and the ones people reach for first are at the bottom of the list. Adjusting numbers is the weakest. Changing the goal, or the paradigm the system arises from, is the strongest.

The order, in short#

At the weak end: constants, parameters and numbers, then buffer sizes, then the physical structure of stocks and flows. In the middle: the length of delays, the strength of balancing and reinforcing feedback loops, the structure of information flows. At the strong end: the rules of the system, the power to change those rules, the goals of the system, the paradigm out of which the system arises, and above all of it the ability to hold any paradigm lightly.

The uncomfortable part is the correlation with authority. The interventions most people are allowed to make are parameters and buffers. The interventions that move a system are goals and paradigms, which almost nobody has permission to touch, and that is the general reason organisations work hard and move slowly.

Meadows' own warning#

She said in the essay that the list should not be taken as precise, that leverage points are counter-intuitive, and that people usually find them and then push in the wrong direction. That caution is part of the framework and is routinely dropped when the list gets reproduced, which is worth keeping because the list is otherwise easy to use as a ranking of importance rather than as an argument about structure.

What it says about altitude#

This is the sharpest existing answer to the question of which level to work at, and it answers a different question from the altitude frameworks. Kanter, Woodward and Heifetz are about the level of attention a person adopts. Meadows is about the level of the system an intervention touches, which is a property of the change rather than of the observer. A leader can be at fifty thousand feet and still be adjusting a parameter.

Used together, the leverage ranking is a test for whether an intervention chosen from a given altitude is worth making, and the three filters at the Altitude Lens are how the altitude gets chosen in the first place.

Where most AI programmes sit on the list#

Near the bottom, and this is the most useful thing the ranking does here. Automating a process changes the speed of an information flow. Adding a dashboard changes information availability. Deploying a model to make a decision faster changes a delay. All of those are real and all of them sit in the weaker half.

The strong interventions are the goals of the system and the paradigm behind it. An organisation deciding which decisions machines may make, which stay with people, and what the people must remain capable of, is operating at the rules and goals level, which is where this research argues the work actually is. The related arguments are at how AI decision rights should be allocated and custody.

What this does not establish#

The ordering is argued rather than measured, which Meadows states herself. No study ranks interventions empirically, and nothing establishes that an organisation acting at higher leverage points performs better. What the framework supplies is a way of noticing that an intervention is weaker than the effort going into it, and that is a diagnosis rather than a finding.

Key sources

• Meadows, D. H. (1999). Leverage points: places to intervene in a system. The Sustainability Institute, developing an essay first published in 1997.

Related SuperSkills research#

• Models of big picture thinking
• The Altitude Lens
• The shape of the organisation after AI