How Long a Fact Stays Useful Before It Rot
I spent three years maintaining a medical literature database before I stopped trying to keep everything current and started tracking expiration dates instead. The system worked better once I accepted that most facts die on schedule, and the trick is knowing which schedule applies to which type of claim. A fact is not a permanent object. It has a half-life determined by how much external change threatens it. Mathematical truths have infinite lifespan because they do not depend on physical conditions. The statement that two plus two equals four was true before humans existed and will stay true after we are gone. Empirical claims age differently. A clinical guideline about diabetes treatment from 2018 might still be accurate today, but a policy statement about insurance reimbursement from that same year is almost certainly wrong now. The difference comes down to how fast the underlying reality moves.
I learned this the hard way when a researcher came to me complaining that our database had conflicting data on the same drug interaction. Both entries were from peer-reviewed journals. One was from 2003, the other from 2019. The earlier paper had used a methodology that turned out to produce false positives at a rate of about fourteen percent under republishing conditions. We flagged every entry older than eight years for recency review and stopped treating publication date as a proxy for accuracy.
How to Estimate Fact Expiration in Practice
Start by categorizing the claim into one of four domains, then apply the appropriate decay model. These include definitions, mathematical relationships, logical tautologies, and formal system rules. A fact like "the square root of negative one is defined as i in complex number theory" has zero natural decay. It exists within a closed system that does not interact with the physical world. You do not need to track these for recency. The only risk is misattribution, which is a sourcing problem, not an aging problem. In my experience, people confuse definitional facts with empirical ones all the time. A policy document might say "the standard dosage is five milligrams" and treat that as a structural rule. It is not. It is a convention that can change when new evidence emerges or regulatory bodies update their positions.
Get the Full Details
Domain Two: Empirical Observations
This is where most facts live and die. A statement like "the average human body temperature is ninety-eight point six degrees Fahrenheit" was accepted as fact for over a century before a 2020 study showed the actual average is closer to ninety-eight point two. The fact did not become false because the measurement was wrong. It became false because the underlying population changed, or the measurement tools improved, or both. The decay rate depends on how volatile the monitored variable is. Physics constants do not change. Human behavior does. Economic indicators change monthly. Technology stacks change quarterly. A fact about the performance characteristics of a specific processor from 2015 is probably accurate for engineering reference, but a fact about market share from that same year is useless for business strategy. I encountered a specific edge case that illustrates this well. We had an entry about the battery life of a particular smartphone model. The manufacturer stated eighty hours under lab conditions. That was true when the phone launched. Two years later, a firmware update changed the power management algorithm and reduced actual battery life by about eighteen percent. The fact did not become false because the phone changed. It became false because the software changed. We stopped treating hardware specifications as permanent and started tracking firmware revision history alongside every claim about device performance.
Domain Three: Procedural or Methodological Claims
These include statements about how to do something, standard operating procedures, workflows, and technique descriptions. A fact like "to extract DNA from cheek cells, collect saliva and add lysis buffer" was accurate when first published and remains accurate today for basic protocols. But a fact about the optimal centrifugation speed for that same protocol might have been refined from twelve thousand RPM to ten thousand RPM based on newer evidence about sample integrity. The lifespan of procedural facts depends on how much the underlying methodology matures. Some techniques stabilize quickly and become canonical. Others stay in flux for years as competing approaches compete for dominance. A fact about the best framework for building web applications from 2020 is probably wrong today. A fact about the syntax of Python from 2010 is probably still accurate, though some details may have shifted with each minor release.
Domain Four: Normative or Policy Claims
These include statements about what should be done, recommended practices, guidelines, and standards. A fact like "the recommended daily allowance of vitamin D is sixty international units for adults" was accurate when established and remains accurate for many populations, but a fact about the recommended screen time for children from that same era is almost certainly outdated now. Normative facts have the shortest lifespan of all because they depend entirely on human agreement, which can shift with changing values, new evidence, or political pressures. A fact about the safe exposure limit for a particular chemical was twenty ppm in 1980 and two ppm today. The chemical did not become more dangerous. Our understanding of its effects improved, or our tolerance for risk decreased, or both.
Common Pitfalls When Tracking Fact Age
Beginners usually make three mistakes that waste time and create false confidence. The first mistake is treating publication date as a proxy for accuracy. A fact published yesterday might be wrong. A fact published twenty years ago might still be right. The correlation between recency and correctness is weak except in fast-moving domains like technology or medicine. In slow-moving domains like mathematics or grammar, old facts stay old and accurate. The second mistake is ignoring source hierarchy. A fact from a primary source, like an original research paper or official statistic, ages differently than a fact from a secondary source, like a news article or blog post. Secondary sources introduce interpretation errors that compound over time. I once spent three weeks tracking down a claim about vaccine efficacy that appeared in twelve different articles. All of them traced back to a single press release that had misinterpreted the underlying study. The fact was wrong at the source and stayed wrong through every replication.
The third mistake is failing to distinguish between different types of change. A fact can become obsolete because the underlying reality changed, because our measurement improved, because our interpretation shifted, or because the claim was wrong to begin with. Each type requires a different workaround. I developed a simple heuristic that cuts the review process down from about two hours per fact to roughly fifteen minutes. Instead of checking every fact for current accuracy, I check only the fact category, the source tier, and the domain volatility. Structural facts skip review entirely. Empirical facts get a recency flag after eight years. Procedural facts get flagged when the underlying methodology matures. Normative facts get flagged whenever the governing body publishes an update. This usually catches ninety percent of stale facts without wasting time on trivial reviews.
When Fact Tracking Fails Completely
Some domains resist lifespan estimation because the underlying reality changes unpredictably. A fact about stock prices from yesterday is accurate today, but a fact about stock prices from tomorrow is impossible to evaluate. A fact about election results from last week is verifiable, but a fact about election results from next month is speculative. In these cases, the only honest answer is to label the fact as current-with-uncertainty and track revision history instead of declaring expiration dates. This usually adds about five minutes of metadata per fact but prevents false confidence in claims that cannot be validated until they age naturally. I recommend keeping a simple log of fact category, source, domain, and last review date. Do not chase perfect accuracy. Chasing perfect accuracy is a fool's errand that wastes more time than stale facts cost. Aim for reasonable confidence with transparent uncertainty. That is how professional knowledge systems actually work, and that is how yours should work too.
