91Ó°ÊÓ

Conditional probability is a crucial concept for understanding how likely an event is to occur given that another event has already happened. In the context of the surveillance system mentioned in the exercise, we are interested in knowing the conditional probability of someone being an actual terrorist given that the surveillance system has identified them as one. This is expressed as \( P(\text{terrorist} | \text{positive}) \), the probability that a person is a terrorist given a positive identification by the system.

The foundation of Bayes' Theorem lies in calculating this conditional probability. We start with prior probabilities, like the probability of picking a terrorist randomly, \( P(\text{terrorist}) \), and the probabilities associated with the accuracy of our test, which include sensitivity and specificity.

• **Sensitivity** is the probability that the test correctly identifies a person as a terrorist when they indeed are.
• **Specificity** is the probability that the test correctly identifies a person as not being a terrorist when they are not.

Conditional probabilities become especially insightful in systems with very low prevalence of the condition being detected, as demonstrated by the dramatic effect of population size in influencing our conditional probability outcome. By incorporating these probabilities, we can make educated judgments about the reliability of our system.

Sensitivity and specificity are measures of a test's performance. They are crucial in evaluating processes like surveillance systems. In the exercise, sensitivity, also called the true positive rate, is the probability that the system correctly identifies a terrorist. Specificity, the true negative rate, is the probability that the system correctly identifies non-terrorists.

To interpret these two values:

• **Sensitivity** of 99% means out of all actual terrorists, the system will correctly identify 99% of them.
• **Specificity** of 99.9% means it will correctly dismiss 99.9% of actual non-terrorists as not being a threat.

However, to truly assess the effectiveness of such a system, we must consider the base rates of the population — in this case, the very low occurrence of terrorists in a large population. These high sensitivity and specificity values can still lead to many false alarms (false positives) when involving a huge population size, where terrorism is rare.

False positives occur when a test incorrectly identifies a non-terrorist as a terrorist. Despite a high specificity rate of 99.9%, false positives can be misleading in the surveillance example, primarily because the non-terrorists vastly outnumber the terrorists in the population.

From the exercise, the chance of a false positive, noted as \( P(\text{positive} | \text{non-terrorist}) \), is 0.1%. Though seemingly tiny, this rate leads to a significant number of false alarms given the massive population size. This is highlighted in the calculation of the overall probability of a positive test and later analyzing the posterior probability using Bayes' Theorem.

• Despite strong initial specificity, the number of false positives rises due to the low base rate of the actual condition.
• A large number of false positive identifications can result, leading to potential misuse of resources and wrongful suspicion.

Understanding false positives is essential in decision-making about whether such a surveillance system is practical and ethical, considering potential implications for individuals wrongly identified.