Words or numbers? Communicating probability in intelligence analysis.
Bibliographic record
Abstract
Intelligence analysis is fundamentally an exercise in expert judgment made under conditions of uncertainty. These judgments are used to inform consequential decisions. Following the major intelligence failure that led to the 2003 war in Iraq, intelligence organizations implemented policies for communicating probability in their assessments. Virtually all chose to convey probability using standardized linguistic lexicons in which an ordered set of select probability terms (e.g., highly likely) is associated with numeric ranges (e.g., 80-90%). We review the benefits and drawbacks of this approach, drawing on psychological research on probability communication and studies that have examined the effectiveness of standardized lexicons. We further discuss how numeric probabilities can overcome many of the shortcomings of linguistic probabilities. Numeric probabilities are not without drawbacks (e.g., they are more difficult to elicit and may be misunderstood by receivers with poor numeracy). However, these drawbacks can be ameliorated with training and practice, whereas the pitfalls of linguistic probabilities are endemic to the approach. We propose that, on balance, the benefits of using numeric probabilities outweigh their drawbacks. Given the enormous costs associated with intelligence failure, the intelligence community should reconsider its reliance on using linguistic probabilities to convey probability in intelligence assessments. Our discussion also has implications for probability communication in other domains such as climate science. (PsycInfo Database Record (c) 2021 APA, all rights reserved).
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.004 | 0.011 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.004 | 0.005 |
| Science and technology studies | 0.001 | 0.004 |
| Scholarly communication | 0.003 | 0.008 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.003 | 0.003 |
| Insufficient payload (model declined to judge) | 0.003 | 0.002 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".