On measuring agreement with numerically bounded linguistic probability schemes: A re-analysis of data from Wintle, Fraser, Wills, Nicholson, and Fidler (2019)
Bibliographic record
Abstract
Across a wide range of domains, experts make probabilistic judgments under conditions of uncertainty to support decision-making. These judgments are often conveyed using linguistic expressions (e.g., x is likely). Seeking to foster shared understanding of these expressions between senders and receivers, the US intelligence community implemented a communication standard that prescribes a set of probability terms and assigns each term an equivalent numerical probability range. In an earlier PLOS ONE article, [1] tested whether access to the standard improves shared understanding and also explored the efficacy of various enhanced presentation formats. Notably, they found that embedding numeric equivalents in text (e.g., x is likely [55-80%]) substantially outperformed the status-quo approach in terms of the percentage overlap between participants' interpretations of linguistic probabilities (defined in terms of the numeric range equivalents they provided for each term) and the numeric ranges in the standard. These results have important prescriptive implications, yet Wintle et al.'s percentage overlap measure of agreement may be viewed as unfairly punitive because it penalizes individuals for being more precise than the stipulated guidelines even when the individuals' interpretations fall perfectly within the stipulated ranges. Arguably, subjects' within-range precision is a positive attribute and should not be penalized in scoring interpretive agreement. Accordingly, in the present article, we reanalyzed Wintle et al.'s data using an alternative measure of percentage overlap that does not penalize in-range precision. Using the alternative measure, we find that percentage overlap is substantially elevated across conditions. More importantly, however, the effects of presentation format and probability level are highly consistent with the original study. By removing the ambiguity caused by Wintle et al.'s unduly punitive measure of agreement, these findings buttress Wintle et al.'s original claim that the methods currently used by intelligence organizations are ineffective at coordinating the meaning of uncertainty expressions between intelligence producers and intelligence consumers. Future studies examining agreement between senders and receivers are also encouraged to reflect carefully on the most appropriate measures of agreement to employ in their experiments and to explicate the bases for their methodological choices.
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 | 0.092 | 0.350 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.008 | 0.007 |
| Science and technology studies | 0.003 | 0.005 |
| Scholarly communication | 0.003 | 0.006 |
| Open science | 0.003 | 0.005 |
| Research integrity | 0.002 | 0.005 |
| Insufficient payload (model declined to judge) | 0.004 | 0.001 |
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".