Learning is usually imagined as a response to an environment that exists independently of the learner. Something happens, an institution observes the result, draws a lesson and changes its behaviour. The next episode then appears to offer another opportunity for learning. Yet institutional learning is rarely this clean. Once an institution changes because of experience, that change alters the conditions under which future experience will occur. Previous learning helps create the environment in which future learning takes place.

This is the endogenous side of institutional learning. The institution does not simply move through a fixed world collecting lessons. Its previous lessons modify its attention, procedures, capabilities, categories, relationships and expectations. Those modifications influence what it encounters next, what it notices and what it is capable of learning from the encounter.

Consider an organisation that experiences a serious operational failure and responds by creating a specialised risk team. The new team improves monitoring and makes certain problems visible earlier. Future institutional experience is now different because the organisation possesses a capability that did not previously exist. The evidence reaching decision-makers changes, interventions happen sooner and some failures may never develop far enough to become observable in their old form.

The original learning has changed the environment of later learning. This produces path dependence. Once institutional cognition has developed along one route, subsequent possibilities are conditioned by what has already been built. Expertise accumulates around particular problems. Indicators make certain phenomena easy to observe. Procedures direct attention towards familiar risks. Technology encodes previous classifications. Budgets sustain capabilities whose existence shapes which future responses are feasible.

None of this means the institution is trapped completely by history. PATH DEPENDENCE ≠ DETERMINISM. It means that the starting point for future learning is partly an outcome of previous learning. This distinction separates the mechanism from mislearning. An institution can become path-dependent because it learned something correct. A valuable capability created yesterday can shape tomorrow’s evidence even when the original lesson remains valid. Mislearning concerns whether the lesson itself was wrong or misleading. Endogenous learning concerns how any sufficiently durable lesson alters future cognitive conditions.

It also differs from recursive learning. Recursive learning requires the institution to observe and revise its own learning process. Path dependence can occur without such reflexivity. Indeed, institutions may be most strongly conditioned by previous learning precisely when they do not recognise that it is happening.

RECURSION ≠ PRECONDITION FOR ENDOGENOUS LEARNING CONDITIONS.

Suppose a government successfully develops expertise in one type of crisis. Over time, recruitment, training, scenario planning and information systems become organised around that experience. The capability is real and useful. But future uncertainty now enters an institution whose cognitive architecture has been shaped by that history. New problems resembling the known crisis may be recognised quickly, while problems requiring very different categories may be harder to interpret.

The same effect appears in evaluation. If previous learning teaches an institution that certain indicators are especially important, those indicators will generate more attention and better data in future. The institution may then receive increasingly rich evidence about what it already knows how to measure, reinforcing the apparent importance of that domain.

Learning has altered the evidence landscape from which later learning proceeds. Artificial intelligence can strengthen these dynamics. Models trained on previous institutional decisions can reproduce patterns generated by earlier learning. Recommendation systems may direct attention towards cases resembling historically important ones. Automated classifications can stabilise categories that originated in earlier institutional experience. As these systems become embedded, previous learning acquires technical persistence.

Again, this can be beneficial. Institutions should preserve valuable lessons. The problem arises when persistence becomes invisible and the institution begins treating its inherited learning environment as though it were simply reality.

Path dependence is particularly important for reform because new interventions rarely begin from a neutral institutional state. They encounter accumulated routines, categories, professional capabilities and memories produced by earlier rounds of adaptation. A reform that looks optimal in abstraction may interact differently with an institution whose history has made some capabilities strong and others weak.

This is one reason institutional designs cannot always be transplanted successfully. Two organisations facing similar external problems may learn differently because their previous learning has created different internal environments. Recognising endogenous learning therefore requires institutions to ask historical questions about current cognition. Why do we measure this? Why is expertise concentrated here? Why does this category seem natural? Why do these risks receive immediate attention while others do not? Which previous experiences made the current learning architecture plausible?

These questions do not imply that inherited arrangements are wrong. Their purpose is to make contingency visible. An institution capable of recognising path dependence gains a wider range of future possibilities because it can distinguish what reality requires from what its own history has made familiar. It can ask whether yesterday’s successful adaptation is still enabling learning or beginning to narrow it.

The relationship between learning and history is therefore recursive in a broader temporal sense even when the institution itself is not reflexively recursive. Learning produces architecture; architecture conditions later experience; later experience is interpreted through what previous learning made possible. Over decades, an institution becomes partly an accumulation of its own lessons. That accumulation can be a source of intelligence. It is how expertise, resilience and mature capability develop. But it can also become constraint when inherited learning defines too tightly what the institution expects to encounter.

The institution that learns today is helping construct the learner that will confront tomorrow. Mature institutional learning therefore requires awareness that every durable lesson changes not only what the institution knows, but the conditions under which it will be able to learn next.