Terrorism PP Briefing

UNCLASSIFIED//FOR OFFICIAL USE ONLY OFFICE OF INTELLIGENCE & ANALYSIS Tradecraft Note from the Research Director _________________________________ “A colleague who spent half his career in the [CIA’s] Directorate of Intelligence and half in the National Clandestine Service once said only half jokingly, ‘You know what the DI’s problem is? You guys write things down.’…He was right in the sense that the written word is forever. Once it is printed, there i s no taking it back or modifying it. Every time we publish, we go ‘on the record’ and the record is there forever, for the second guessers, the hindsight experts, and anyone with an agenda. Thus, our judgments need to be as precise as we can make them, sup ported by evidence and argument.” —Martin Peterson, “What I Learned in 40 Years of Doing Intelligence Analysis” _________________________________ TN-4 February 2014 (U) Why Precise Language in Intelligence is So Important (U//FOUO) Precise language ensures that intelligence customers receive an analytic message as intended and that analysts can subsequently evaluate their predictive judgments for accuracy and learn from them. A recent DNI study on accuracy in finished intelligence honed in on vague statements of probability and predictive judgments . T he m ost common problems included:  Probabilistic qualifiers, such as “may,” “might,” or “could ,” without reference to the circumstances in which an event would become likely or explaining the reasoning behind the probability statement .  Relative probability statements, such as “inc reases the risk” or “greater likelihood” without reference to a baseline for comparison , as well as similarly vague statements of causality.  Judgments —often unverifiable —about a person’s beliefs, plans, or mental status without addressing the prospects of the individual taking a particular action . (U) Probabilistic Qualifiers (U//FOUO) Policymakers look to intelligence analysts to render judgments that go beyond an “ anything might happen” assessment to gain insight into the absolute likelihood of a development or the circumstances under which a possible outcome would become a probable or almost certain o ne. Vague probability language can be misinterpreted in a way that was not intended and provides little insight into the likelihood of a development occurring or the reasoning behind the judgment . Probability statements should always be accompanied by an explanation of why the probability is assigned . A good rule -of -thumb is to include a clause starting with “if” or “because” either before or after the judgment :  “___ suggests___ will occur because/if _____. ” UNCLASSIFIED//FOR OFFICIAL USE ONLY 2  “____ risks ____ if/because____.”  “If/because ____, ____is possible .” (U//FOUO) Although you can strengthen a probability judgment by outlining the conditions under which the outcome is more probable, be careful to guard against tautologi es (e.g., “If necessary, X will…” or “ …will do X if it is deemed a threat to national security”). Also, n ever strengthen a probability judgment in the absence of supporting evidence simply to avoid the “may/might/could/suggests/poses risks” formulation (s). Instead , clarify for the consumer ho w risky the development is or the plausibility of a given outcome . Identify the key sources of uncertainty or the intelligence gaps that prevent a more confident assessment. (U) Relative vs. Absolute Probability (U//FOUO) Relative probability statements —e.g ., “has raised the risk that” —should refer back to a baseline for comparison or be re- crafted in absolute terms . There is a big difference between saying the likelihood of war has increased from five to 10 percent and saying that it has increased from 40 to 80 percent. In saying “ ____has raised the risk of war” w ithout a baseline reference, one reader could conclude that war is imminent (in the first case ) and another reader that it is a longshot ( in the second). Since it is usually not possible to assign a numeric probability in predictive intelligence, it is better to reframe your judgment in absolute terms, “…____ has raised the risk of war so that now war is very likely.” The use of some verbs, such as “aggravate” or “reinforce,” also often imply a situation is becoming more or less likely without explicitly saying so. Relative or vague probability language such as the following s hould be clarified or explained:  “____ reduces the likelihood that ____. ”  “_____ will lead to greater pressure to ____.”  “_____ would reduce the ability of _____to ____.”  “____is putting pressure on____ to.”  “____would create an opening for____.”  “____make___vulnerable to____.”  “____probably will struggle to____.” (U) Decision -making vs. State of Mind (U//FOUO) When making a statement about a person’s plans, beliefs, or attitudes, try to supplement it with a judgment about what observable behavior these mental dispositions are likely to induce. Although it is helpful to be told an individual plans to undertake a particular action, it is more helpful to know why or under what conditions he or she will take such action, what the action’s impact would be, and whether the actio n would succeed . Similarly, in addition to noting an individual is becoming “increasingly concerned” about something, note what specific behavior(s) that concern is likely to induce. Additional context and specificity is required when using p hrases such as: UNCLASSIFIED//FOR OFFICIAL USE ONLY 3  “____is likely to be distrustful of ____”  “_____would weigh these factors in deciding whether to_____”  “______may consider doing_____.” Say This: Rather Tha n: Here’s Why: Avoid Vague Probabilistic Qualifiers In light of recent regime setbacks in Sparta , we are monitoring indications that the allied Etruscan Government intends to blame the U nited States — one of several indicators that would suggest it is preparing to attack the Homeland (see Table for status of all indicators). An Etruscan -launched attack against the Homeland is possible.

Anything is possible. The reader needs to understand the factors driving a decision to attack. The preferred language calls out the development that has changed—regime setbacks in a country ; the implication— possible attack on the Homeland; and an indicator that enables a probability judgment . Try to Craft Probability in Absolute Terms Criminal networks have long been capable of launching the kind of attack that recently impact ed Floor -M art consumers, but the attack’s scale will almost certainly encourage copycat attacks. The widely reported theft of Floor -Mart customers’ credit card data raises the risk that other stores will be similarly hacked. The “raises the risk” formulation pro vides no baseline for comparison. The preferred language says clearly that more attacks are coming. Eschew Vague Statements that Hint at Probability The publicity about US surveillance on Spartan leaders makes it unlikely the Spartans will agree to broad data -sharing on refugee flows , although bilateral cooperation on specific cases will continue behind the scenes because of fears that Bo hemia will plant subversives among the migrants .

Recent leaks of US surveillance capabilities are putting pressure on Spartan leaders to reduce cooperation with the United States. The vague “putting pressure on” formulation falls short of judging whether the hinted at outcome will occur. The preferred language clearly identifies which cooperation will be affected. (U) Focus on Decision -making, Rum- running organizations are unlikely to attack US officials because they probably When considering targeting US officials, the rum -runners will weigh the impact of a The latter formulation does not explain how consideration of a US response would impact the decision to attack UNCLASSIFIED//FOR OFFICIAL USE ONLY 4 Not State of Mind think doing so would invite retaliation damaging to their business interests. US response on its business interests. US officials.