Fallacy benchmarks detect argument style not true fallacies
Fallacy Benchmarks Measure Scheme Recognition, Not Fallacy Detection
Computation and LanguageMachine Learning
Summary
Benchmarks for detecting fallacies in arguments often group all non-fallacy arguments together, which can confuse machines. The authors show that classifiers trained on these benchmarks learn to recognize the style or type of argument rather than whether it contains a fallacy. When tested with arguments that match the same style but are correct, these systems make many mistakes. This means current benchmarks may overstate how well computers can truly detect fallacies.
What this means in practice
- •For natural language processing engineers: Improve fallacy detection systems by considering argument schemes to reduce false positives when distinguishing true fallacies from correct arguments.
- •For fact-checking tool developers: Use scheme-aware evaluation to better test the accuracy of fallacy detectors integrated into fact-checking software.
Authors
Navyansh Singh, Animesh Pathak, Aarav Singh
Abstract
Fallacy-detection benchmarks pair fallacy classes with a single "valid" or "none" class that takes everything data collection did not label as a fallacy. This construction is misleading: a classifier can learn cues that do well on this class without learning to tell a fallacy from a correct argument. We show that the low false-positive rates benchmarks report are an artifact of how the class is built, not evidence of detection ability. The most informative negative for a fallacy is a correct argument using the same argumentation scheme, and such arguments are at most a few percent of the valid class across the four benchmarks we examined. Evaluated on constructed scheme-matched negatives, false-positive rates rise from 16.6% to 58.9% on CoCoLoFa and from 5.7% to 62.0% on Reddit. That rate depends on how the negatives are written, so we also compare two conditions from the same pipeline that differ only in scheme identity. Classifiers label scheme-matched negatives as the source fallacy type 40.9 points more often than wrong-scheme negatives, which are instead identified as the scheme they actually use 85.9% of the time against 0.4% for the source type. The classifier has learned which scheme an argument uses, not whether it uses it correctly, and on the benchmarks' own test sets the two are indistinguishable. The same dissociation appears in three zero-shot LLM detectors that never saw these benchmarks, and the measurement is far lower on a negative class that was built deliberately. We release the items as Scheme Foils. A reported false-positive rate should not be trusted as a measure of detection until the valid class has been audited for scheme-matched coverage.