Does a Neural Model Understand the De Re / De Dicto Distinction?
Bibliographic record
Abstract
Neural network language models (NNLMs) are often casually said to “understand” language, but what linguistic structures do they really learn? We pose this question in the context of de re / de dicto ambiguities. Nouns and determiner phrases in intensional contexts, such as belief, desire, and modality, are subject to referential ambiguities. The phrase “Lilo believes an alien is on the loose,” for example, has two interpretations: one (de re) in which she believes a specific entity which happens to be an alien is on the loose, and another (de dicto) in which she believes some unspecified alien is on the loose. In this paper we confront an NNLM with contexts producing de re / de dicto ambiguities. We use coreference resolution to investigate which interpretive possibilities the model captures. We find that while RoBERTa is sensitive to the fact that intensional predicates and indefinite determiners each change coreference possibilities, it does not grasp how the two interact with each other, and hence misses a deeper level of semantic structure. This inquiry is novel in its cross-disciplinary approach to philosophy, semantics and NLP, bringing formal semantic insight to an active research area testing the nature of NNLMs’ linguistic “understanding.”
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 | 0.001 | 0.006 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.002 |
| Scholarly communication | 0.002 | 0.008 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.002 | 0.002 |
| 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".