AI Glossary / Data science and analytics / Intermediate
Intersection Of Events
Definition
The intersection of events is the event containing all outcomes that belong to each of the specified events. For two events A and B, it is written A intersection B, or A ∩ B, and means A and B both occur. An intersection is a set of outcomes; its probability is the numerical chance assigned to that set.
What it means
Suppose an order can be international, late, both, or neither. The intersection of international and late includes only orders meeting both descriptions. It excludes international orders that arrived on time and late orders that were domestic. This makes intersection useful for defining precise populations and combined conditions. The conditions do not need to describe actions occurring at exactly the same clock time; they must apply to outcomes within the clearly specified experiment or observation unit.
How it works
Define the sample space and each event, then identify their overlap. For equally likely outcomes in a finite sample space, divide the number in the overlap by the total number of outcomes. More generally, the multiplication rule gives P(A and B) = P(B) times P(A given B), when the conditional probability is defined. If the events are independent, this becomes P(A) times P(B). Without independence or other information about their relationship, the separate probabilities usually do not determine the overlap.
Three practical examples
- Customer analysis: a team counts customers who both renewed a subscription and contacted support during a defined period.
- Quality monitoring: an analyst measures the proportion of inspected items that both exceed a size tolerance and have a surface defect.
- Model evaluation: the intersection of predicted positive and actually positive identifies true-positive cases within a labeled evaluation dataset.
Think of it this way
Think of two transparent lists laid over one another, with shared names highlighted. The shared portion represents the intersection. This is only a counting analogy: in probability, outcomes need not be people or equally likely, and duplicate records should not silently change the unit being counted.
Worked example
A fictional dataset contains 200 orders. Of these, 80 are international, 50 are late, and 30 are both international and late. If one order is selected uniformly from these 200, the probability of the intersection is 30/200 = 0.15, or 15%.
The conditional probability of being late among international orders is 30/80 = 0.375, or 37.5%. Multiplying the international proportion by that conditional proportion gives 80/200 times 30/80 = 30/200, recovering the 15% joint probability. Multiplying the two overall proportions instead gives 0.40 times 0.25 = 0.10, which is wrong for this table because the events are not independent.
For comparison, international or late includes 80 + 50 minus 30 = 100 orders. The overlap is subtracted once because it was counted in both totals. Thus the union probability is 50%, while the intersection probability is 15%. Changing and to or changes the population being described.
Illustrative case study
In a fictional service review, a dashboard reports customers who used an AI assistant and later requested human help. One team reads the figure as the proportion of all customers; another reads it as the proportion of assistant users. The analyst separates the intersection count from the two possible denominators and gives each measure a clear label. The team also checks that both events concern the same customer and reporting period. The case shows why a correct overlap count can still support a misleading statement if its denominator or observation unit is unclear.
Common misconceptions and mistakes
Intersection does not mean independence. A and B can overlap strongly, weakly, or not at all. It is also different from conditional probability: P(A and B) uses the full probability space, while P(A given B) restricts attention to B when B has positive probability. For mutually exclusive events, the intersection is empty. In data work, matching records from different periods or joining tables with duplicate keys can produce a count that does not represent the intended intersection of unique outcomes.
Business use
Use intersection language to make business criteria explicit. A campaign eligibility rule may require both an active account and a completed consent step; a quality escalation may require two observed conditions. State whether every condition applies to the same transaction, person, device, or time window. Keep the numerator and denominator visible in reports. When estimating combined-event probabilities from separate sources, do not assume independence merely to obtain a convenient number. Seek joint data or state and assess the dependence assumption.
People also ask
Can an intersection contain many outcomes?
Yes. An event can include any number of outcomes, and the intersection may contain many of them.
Is A and B the same event as B and A?
Yes. The order does not change their overlap.
Can an intersection be more likely than either event alone?
No. It is contained within each event, so its probability cannot exceed either individual probability.
Check your understanding
In the order example, is 30/80 the probability that a randomly selected order is both international and late?
Answer: No. That fraction is the late proportion among international orders. The joint probability among all 200 orders is 30/200.
Related terms
- Union Of Events
- Conditional probability
- Joint probability
- Sample space
- Independent Events
- Mutually Exclusive Events
- Contingency table
- Boolean AND
Learn next
Draw a two-by-two table containing both, A only, B only, and neither. Check that its counts sum to the intended population. Calculate the intersection, union, and two conditional probabilities from the same table. Then practice translating a business rule into explicit event definitions before calculating a percentage.
More in Data science and analytics
- Immutable Data Types - Immutable data types are types whose objects do not support changes to their own value after creation.
- Independent variable - An independent variable is a variable used as an input or explanatory factor when studying or predicting an outcome.
- Interleaving distance - Interleaving distance measures how closely two persistence modules can be related after allowing a shift in their scale parameters.
- Item tree analysis - Item tree analysis, or ITA, is an exploratory method for investigating hierarchical relationships among questionnaire or test items from response patterns.
- k-nearest neighbors - K-nearest neighbors, often written k-NN, is a method that uses the k closest reference examples to make a prediction or retrieve similar cases.
- K-optimal pattern discovery - K-optimal pattern discovery is an exploratory data-mining approach that seeks the k best patterns according to a specified measure and set of constraints.
Sources
OpenStax, multiplication rule and intersections: https://openstax.org/books/introductory-business-statistics/pages/3-3-two-basic-rules-of-probability OpenStax, event terminology and conditional probability: https://openstax.org/books/introductory-statistics/pages/3-key-terms OpenStax, probability and set relationships: https://openstax.org/books/algebra-and-trigonometry-2e/pages/13-7-probability
Previous: Interrupted time series Next: Intersection over union