🧪 This platform is in early beta. Features may change and you might encounter bugs. We appreciate your patience!
inspection_paradox
The Inspection Paradox occurs when observing a process at a random moment makes you more likely to land in a longer interval. This means experienced wait times, class sizes, or lifespans systematically exceed their true averages. It is a form of length-biased sampling.
A bus company schedules buses every 10 minutes on average, but actual headways vary. If you arrive at a random time, you are more likely to arrive during a long gap than a short one, so your average wait exceeds the expected 5 minutes.
A commuter starts a new job and notices her subway train always seems packed when she boards. She concludes the line is overcrowded, not realizing she tends to board during long gaps between trains — the very gaps that allow more passengers to accumulate on the platform, making each train she catches unusually full.
A hospital administrator samples patients currently in beds to estimate average length of stay and gets a figure of 8 days. The actual average stay is only 3 days — but because long-stay patients occupy beds for more days, any random snapshot disproportionately captures them, skewing the estimate upward.
Binary (yes/no) questions an LLM must answer to identify this aspect:
Is an observation being made by sampling at a random point in time rather than at the start of an interval?
Type: binaryAre longer intervals more likely to be observed simply because they occupy more time?
Type: binaryDoes the reported experience differ systematically from the scheduled or average interval?
Type: binaryIs the conclusion drawn from a length-biased sample rather than from the full distribution of intervals?
Type: binaryThe Inspection Paradox occurs when observing a process at a random moment makes you more likely to land in a longer interval. This means experienced wait times, class sizes, or lifespans systematically exceed their true averages. It is a form of length-biased sampling.
When you sample a process at a random point, longer intervals are proportionally more likely to contain your observation point. A 20-minute gap is twice as likely to be 'hit' by a random arrival as a 10-minute gap, so the experienced distribution is biased toward longer durations.
Distinguish between the distribution of all intervals and the distribution of intervals experienced by random arrivals. Use forward recurrence time calculations rather than naive averages. Collect data from interval start points, not from random observation points.
This paradox affects transit planning, where riders perceive service as worse than scheduled. It also arises in class size surveys (students disproportionately report large classes), hospital length-of-stay studies, and renewal theory in operations research.
On average, people's friends have more friends than they do, due to sampling bias toward popular nodes.
Ignoring general statistical base rates in favor of specific individual-case info.
A trend in several groups that disappears or reverses when combined.
Use these tools to detect, analyze, or train this aspect.