Knowledge can fail us in more than one way. New evidence can change what we believe. A changing environment can make a once-useful method stop fitting. And sometimes the lesson remains correct, but never reaches the decision where it is needed.
This study asks how to tell those failures apart—and how to build systems that retain useful knowledge without giving old assumptions permanent authority.
Status: working synthesis. The distinctions below are a practical research model, not a universal scientific taxonomy.
There is no single clock
The phrase half-life of knowledge is tempting because it suggests a clean decay curve. The evidence is messier. Different kinds of knowledge age for different reasons, on different timescales, and sometimes do not meaningfully decay at all.
In medicine, one study of systematic reviews found a median 5.5 years before a signal that an update was needed. That does not mean half of medical knowledge becomes false every 5.5 years. It means a specific cohort of reviews reached a defined update signal on that timescale.
Organizational research shows a different phenomenon: knowledge gained through experience can depreciate when production stops, people turn over, routines weaken, or learning fails to transfer. Other work shows knowledge can remain embedded more durably in routines or technology.
Four ways useful knowledge can lose its effect
1. Epistemic revision. New evidence changes what is justified.
2. Applicability drift. The original claim may still be sound, but the environment, version, market, workflow, or decision has changed.
3. Retrieval failure. The right lesson exists somewhere, but it is not found at the point of use.
4. Activation failure. The lesson is available—even understood—but does not change the consequential action or receiving outcome.
That fourth failure is easy to miss. A system can “remember” perfectly and still fail to learn operationally.
Old does not mean stale
A recent source is not automatically better than an older one. Older work can remain foundational. A newer source can be weaker. Historical evidence can also explain why a current rule exists even after the old rule itself has been superseded.
The more useful question is not How old is this? but What would make its authority change?
That can include new evidence, changed dependencies, changed context, contradictory results, a different receiving environment, or evidence that the knowledge no longer improves outcomes.
A related but different question: how long do mistakes survive?
If a known mistake keeps recurring, that is not necessarily the same thing as knowledge becoming outdated. The lesson may still be correct. The failure may be that it was never retrieved, never activated, overruled by another incentive, or lost at an interface between systems.
So I am treating two directions separately: how long useful knowledge keeps its value, and how quickly a known failure stops recurring. They may eventually support comparable measurements, but they should not be forced into one number before the evidence earns it.
What I am watching next
I am looking for cases where a lesson was known before a decision, then tracing what happened next: Was it retrieved? Did it still apply? Did it have authority? Did it change the action? Did the receiving system actually benefit?
The study will change as those cases accumulate. Updates here should represent substantive changes in evidence or interpretation, not a fresh timestamp applied to the same idea.
Selected research
Shojania et al. (2007), How quickly do systematic reviews go out of date?
Darr, Argote & Epple (1995), The Acquisition, Transfer, and Depreciation of Knowledge in Service Organizations
Argote, Beckman & Epple (1990), The Persistence and Transfer of Learning in Industrial Settings
Sinyor et al. (2023), Does knowledge have a half-life?
Cochrane Handbook, Updating a review