THE IDEA
The part and the whole want different things
Every system contains a built-in tension. Each part has its own needs, goals, and logic. The whole has different ones. What’s efficient for a single department might be wasteful for the organisation. What’s rational for one person might be destructive for the group. What makes sense at the local level can be disastrous at the global level.
This isn’t a design flaw to be eliminated. It’s a structural tension to be managed. In any system where parts have some autonomy - which is to say, every interesting system - the interests of the part and the interests of the whole will sometimes conflict. The question isn’t how to eliminate the conflict. It’s how to create structures that balance local autonomy with global coherence.
The tension is everywhere. A cell that optimises its own reproduction without reference to the body becomes cancer. A business that optimises its own profit without reference to the commons produces pollution. A team member who optimises their own visibility without reference to team goals becomes a free rider. In each case, the local optimisation makes perfect sense from the local perspective. The damage is only visible from the global one.
IN PRACTICE
When doing your best makes things worse
Traffic. Every driver optimises locally - choosing the fastest route for themselves. When a faster route becomes known (through a navigation app, say), everyone diverts to it simultaneously. The route becomes congested, side streets are overwhelmed, and the overall traffic time for everyone increases. Each driver made a locally optimal choice. The global result is worse than if they’d all stayed on the main road. This is Braess’s paradox: adding capacity to a network can make everyone slower when individuals optimise locally.
A team has a budget for professional development. Each member chooses the course most relevant to their current role - a locally optimal decision. But the team now has five people with deeper versions of the same skill and nobody with the new skill the project needs next quarter. A globally optimal allocation would have sent two people to learn something different, even if it was less personally relevant. The team’s collective capability is weaker because each individual optimised for themselves.
A country subsidises its farmers to increase food production - locally optimal for food security. But the subsidised food is exported below cost, destroying farming in other countries that can’t compete. The global food system becomes less resilient, more dependent on a few producing nations, and more vulnerable to disruption. The local optimisation for one country’s food security undermined the global system’s food security.
WORKING WITH THIS
Bridging the gap between part and whole
The key is feedback. When local actors can see the global consequences of their choices, they naturally adjust. The problem is that most systems hide global consequences from local actors. The driver doesn’t see the overall traffic pattern. The team member doesn’t see the team’s skill gaps. The subsidy designer doesn’t see the effect on global markets.
Design feedback loops that make global consequences visible locally. Show teams how their decisions affect the whole. Make the system-level metrics available alongside the local ones. Create incentives that reward contribution to the whole, not just performance of the part.
When feedback alone isn’t enough, design constraints that prevent locally optimal but globally destructive behaviour. Speed limits on roads. Caps on individual use of shared resources. Budget processes that allocate based on system-level priorities, not just departmental requests. These constraints feel restrictive to the local actor. They’re protective of the global system.
THE INSIGHT
The line to remember
The best decision for the part is often the worst decision for the whole. The system’s job is to make the right trade-off visible, not to pretend it doesn’t exist.
RECOGNITION
When this is in play
You’re seeing local-vs-global tension when each component of a system is performing well but the system-level outcome is poor. When rational individual choices produce irrational collective outcomes. When a shared resource is being depleted by individually reasonable behaviour. When the phrase “we’re all doing our bit” is true at the local level and false at the global one. When adding more local optimisation makes the global problem worse, not better.