AI-Supported Eliminative Argumentation: Practical Experience Generating Defeaters to Increase Confidence in Assurance Cases
Bibliographic record
Abstract
Assurance cases (AC) are structured arguments that justify why a system is acceptably safe. Though ACs can increase confidence that systems will operate safely and reliably, they are also susceptible to problems such as reasoning errors and confirmation bias. Recent work proposed AI-Supported Eliminative Argumentation (AI-EA), a framework leveraging Generative AI (GAI) models to support AC development by identifying potential reasons why the argument may be invalid (a.k.a. defeaters) so that they can be mitigated. However, this framework was not implemented and its effectiveness was not assessed empirically.In this practical experience paper, we implement AI-EA, explain and justify our design choices, and report on our practical experience in empirically evaluating its effectiveness in collaboration with experts in the safety domain. Our evaluation considers 171 AI-generated defeaters across two industrial case studies from the nuclear and automotive domains. Our findings show that GAI can generate informative defeaters with few significant hallucinations and that 25% of the generated defeaters were confirmed by developers of each AC to represent reasonable doubts or errors in the argument. Our implementation and data are made publicly available.
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 | 0.024 | 0.115 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.001 | 0.003 |
| Scholarly communication | 0.004 | 0.004 |
| Open science | 0.004 | 0.005 |
| Research integrity | 0.003 | 0.003 |
| Insufficient payload (model declined to judge) | 0.006 | 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".