ATP 2-33.4, page 40

Chapter 3 3-39. The expression of numerical probabilities can mitigate the imprecision of probability phrases (“very likely” or “improbable”). Moreover, numerical probabilities mitigate the potential for analysts to exploit imprecision in favor of their position. Using numerical probabilities ensures mathematical rules are followed and forces consideration of a complete set of alternatives. This in turn gives the analyst a rational basis to judge whether the probability distribution is an accurate reflection of the analyst’s beliefs. 3-40. Assignments of probability require a complete set of non-overlapping (mutually exclusive) answers, events, scenarios, or COAs. In addition, misuse can feed availability and anchoring biases. 3-41. When using subjective probability, it is essential in defining a numerical score and range to ensure all personnel involved understand the meaning of the terms. Table 3-1 shows an example of subjective probability and the language associated with each score or range. Table 3-1. Subjective probability table Subjective Probability Table | torm TScore Range (percent) Highly probable 91 to 100 Probable [3s 84 t090 Highly kely re 71 t080 Likely 61 to 70 41 10 60 Unlikely 31 0.40 Highly untikely 21 to 30 Improbable 1410 20 Highly improbable 11010 3-42. While subjective probability looks similar to event mapping, the differences are with subjective probability you are not determining a timeline for any particular events; you are simply attempting to predict an outcome and applying a percentage of probability to each outcome. THE METHOD 3-43. Subjective probability rules must be followed: e The probability assigned to a given hypothesis must be within the range of 0.0 (or 0 percent) to 1.0 (100 percent). A probability of 0.0 means the hypothesis is certainly wrong; whereas a probability of 1.0 means that the hypothesis is certainly correct. e@ = The total probability distributed among all hypotheses is a complete, non-overlapping set must add to 1.0 (100 percent). 3-44. The following are steps for this technique: e@ Step 1. Identify a complete set of high-level, non-overlapping hypotheses that seek to answer a clearly defined question. Use the technique of defining the issue to ensure that the question is clear. e@ Step 2. Generate simple chains of events or facts for each hypothesis. Event trees and event mapping are two techniques that aid in this step. The number of scenarios that can be constructed for a given hypothesis depends on the detail desired. Each scenario describes one instance of how the associated hypothesis may come to pass. © Step 3. The probability of a given hypothesis is a function of the probabilities of all the scenarios that would support a hypothesis as being true. The probability of a given scenario is a function of all the events within that scenario occurring. That is, the probabilities (percentages) for each option are multiplied throughout the scenario to determine the probability for the scenario. There are two types of probability events that need to be analyzed: 3-10 ATP 2-33.4 18 August 2014 Chapter 3 3-10 ATP 2-33.4 18 August 2014 3-39. The expression of numerical probabilities can mitigate the imprecision of probability phrases (“very likely” or “improbable”). Moreover, numerical probabilities mitigate the potential for analysts to exploit imprecision in favor of their position. Using numerical probabilities ensures mathematical rules are followed and forces consideration of a complete set of alternatives. This in turn gives the analyst a rational basis to judge whether the probability distribution is an accurate reflection of the analyst’s beliefs. 3-40. Assignments of probability require a complete set of non-overlapping (mutually exclusive) answers, events, scenarios, or COAs. In addition, misuse can feed availability and anchoring biases. 3-41. When using subjective probability, it is essential in defining a numerical score and range to ensure all personnel involved understand the meaning of the terms. Table 3-1 shows an example of subjective probability and the language associated with each score or range. Table 3-1. Subjective probability table Subjective Probability Table Term Score Range (percent) Highly probable 10 91 to 100 Probable 9 81 to 90 Highly likely 8 71 to 80 Likely 7 61 to 70 Possible 5 to 6 41 to 60 Unlikely 4 31 to 40 Highly unlikely 3 21 to 30 Improbable 2 11 to 20 Highly improbable 1 1 to 10 3-42. While subjective probability looks similar to event mapping, the differences are with subjective probability you are not determining a timeline for any particular events; you are simply attempting to predict an outcome and applying a percentage of probability to each outcome. THE METHOD 3-43. Subjective probability rules must be followed:  The probability assigned to a given hypothesis must be within the range of 0.0 (or 0 percent) to 1.0 (100 percent). A probability of 0.0 means the hypothesis is certainly wrong; whereas a probability of 1.0 means that the hypothesis is certainly correct.  The total probability distributed among all hypotheses is a complete, non-overlapping set must add to 1.0 (100 percent). 3-44. The following are steps for this technique:  Step 1. Identify a complete set of high-level, non-overlapping hypotheses that seek to answer a clearly defined question. Use the technique of defining the issue to ensure that the question is clear.  Step 2. Generate simple chains of events or facts for each hypothesis. Event trees and event mapping are two techniques that aid in this step. The number of scenarios that can be constructed for a given hypothesis depends on the detail desired. Each scenario describes one instance of how the associated hypothesis may come to pass.  Step 3. The probability of a given hypothesis is a function of the probabilities of all the scenarios that would support a hypothesis as being true. The probability of a given scenario is a function of all the events within that scenario occurring. That is, the probabilities (percentages) for each option are multiplied throughout the scenario to determine the probability for the scenario. There are two types of probability events that need to be analyzed: