Pragmatics and non-literal language understanding are essential to human communication, and present a long-standing challenge for artificial language models. We perform a fine-grained comparison of language models and humans on seven pragmatic phenomena, using zero-shot prompting on an expert-curated set of English materials. We ask whether models (1) select pragmatic interpretations of speaker utterances, (2) make similar error patterns as humans, and (3) use similar linguistic cues as humans to solve the tasks. We find that the largest models achieve high accuracy and match human error patterns: within incorrect responses, models favor literal interpretations over heuristic-based distractors. We also find preliminary evidence that models and humans are sensitive to similar linguistic cues. Our results suggest that pragmatic behaviors can emerge in models without explicitly constructed representations of mental states. However, models tend to struggle with phenomena relying on social expectation violations.

We propose a novel task, G4C, to study teacher-student natural language interactions in a goal-driven and grounded environment. Dungeons and Dragons (D&D), a role-playing game, provides an ideal setting to investigate such interactions. Here, the Dungeon Master (DM), i.e., the teacher, guides the actions of several players—students, each with their own personas and abilities—to achieve shared goals grounded in a fantasy world. Our approach is to decompose and model these interactions into (1) the DM’s intent to guide players toward a given goal; (2) the DM’s guidance utterance to the players expressing this intent; and (3) a theory-of-mind (ToM) model that anticipates the players’ reaction to the guidance one turn into the future. We develop a novel reinforcement learning (RL) method for training a DM that generates guidance for players by rewarding utterances where the intent matches the ToM-anticipated player actions. Human and automated evaluations show that a DM trained to explicitly model intents and incorporate ToM of the players using RL generates better-quality guidance that is 3x more likely to fulfill the DM’s intent than a vanilla natural language generation (NLG) approach.

Prompting is now a dominant method for evaluating the linguistic knowledge of large language models (LLMs). While other methods directly read out models’ probability distributions over strings, prompting requires models to access this internal information by processing linguistic input, thereby implicitly testing a new type of emergent ability: metalinguistic judgment. In this study, we compare metalinguistic prompting and direct probability measurements as ways of measuring models’ linguistic knowledge. Broadly, we find that LLMs’ metalinguistic judgments are inferior to quantities directly derived from representations. Furthermore, consistency gets worse as the prompt query diverges from direct measurements of next-word probabilities. Our findings suggest that negative results relying on metalinguistic prompts cannot be taken as conclusive evidence that an LLM lacks a particular linguistic generalization. Our results also highlight the value that is lost with the move to closed APIs where access to probability distributions is limited.

People rely heavily on context to enrich meaning beyond what is literally said, enabling concise but effective communication. To interact successfully and naturally with people, user-facing artificial intelligence systems will require similar skills in pragmatics: relying on various types of context — from shared linguistic goals and conventions, to the visual and embodied world — to use language effectively. We survey existing grounded settings and pragmatic modeling approaches and analyze how the task goals, environmental contexts, and communicative affordances in each work enrich linguistic meaning. We present recommendations for future grounded task design to naturally elicit pragmatic phenomena, and suggest directions that focus on a broader range of communicative contexts and affordances.

Scalar inferences (SI) are a signature example of how humans interpret language based on unspoken alternatives. While empirical studies have demonstrated that human SI rates are highly variable—both within instances of a single scale, and across different scales—there have been few proposals that quantitatively explain both cross- and within-scale variation. Furthermore, while it is generally assumed that SIs arise through reasoning about unspoken alternatives, it remains debated whether humans reason about alternatives as linguistic forms, or at the level of concepts. Here, we test a shared mechanism explaining SI rates within and across scales: context-driven expectations about the unspoken alternatives. Using neural language models to approximate human predictive distributions, we find that SI rates are captured by the expectedness of the strong scalemate as an alternative. Crucially, however, expectedness robustly predicts cross-scale variation only under a meaning-based view of alternatives. Our results suggest that pragmatic inferences arise from context-driven expectations over alternatives, and these expectations operate at the level of concepts.1

Scalar implicature (SI) arises when a speaker uses an expression (e.g., “some”) that is semantically compatible with a logically stronger alternative on the same scale (e.g., “all”), leading the listener to infer that they did not intend to convey the stronger meaning. Prior work has demonstrated that SI rates are highly variable across scales, raising the question of what factors determine the SI strength for a particular scale. Here, we test the hypothesis that SI rates depend on the listener’s confidence in the underlying scale, which we operationalize as uncertainty over the distribution of possible alternatives conditioned on the context. We use a T5 model fine-tuned on a text infilling task to estimate this distribution. We find that scale uncertainty predicts human SI rates, measured as entropy over the sampled alternatives and over latent classes among alternatives in sentence embedding space. Furthermore, we do not find a significant effect of the surprisal of the strong scalemate. Our results suggest that pragmatic inferences depend on listeners’ context-driven uncertainty over alternatives.

Prior work has shown that structural supervision helps English language models learn generalizations about syntactic phenomena such as subject-verb agreement. However, it remains unclear if such an inductive bias would also improve language models’ ability to learn grammatical dependencies in typologically different languages. Here we investigate this question in Mandarin Chinese, which has a logographic, largely syllable-based writing system; different word order; and sparser morphology than English. We train LSTMs, Recurrent Neural Network Grammars, Transformer language models, and Transformer-parameterized generative parsing models on two Mandarin Chinese datasets of different sizes. We evaluate the models’ ability to learn different aspects of Mandarin grammar that assess syntactic and semantic relationships. We find suggestive evidence that structural supervision helps with representing syntactic state across intervening content and improves performance in low-data settings, suggesting that the benefits of hierarchical inductive biases in acquiring dependency relationships may extend beyond English.

While state-of-the-art neural network models continue to achieve lower perplexity scores on language modeling benchmarks, it remains unknown whether optimizing for broad-coverage predictive performance leads to human-like syntactic knowledge. Furthermore, existing work has not provided a clear picture about the model properties required to produce proper syntactic generalizations. We present a systematic evaluation of the syntactic knowledge of neural language models, testing 20 combinations of model types and data sizes on a set of 34 English-language syntactic test suites. We find substantial differences in syntactic generalization performance by model architecture, with sequential models underperforming other architectures. Factorially manipulating model architecture and training dataset size (1M-40M words), we find that variability in syntactic generalization performance is substantially greater by architecture than by dataset size for the corpora tested in our experiments. Our results also reveal a dissociation between perplexity and syntactic generalization performance.

Targeted syntactic evaluations have yielded insights into the generalizations learned by neural network language models. However, this line of research requires an uncommon confluence of skills: both the theoretical knowledge needed to design controlled psycholinguistic experiments, and the technical proficiency needed to train and deploy large-scale language models. We present SyntaxGym, an online platform designed to make targeted evaluations accessible to both experts in NLP and linguistics, reproducible across computing environments, and standardized following the norms of psycholinguistic experimental design. This paper releases two tools of independent value for the computational linguistics community: 1. A website, syntaxgym.org, which centralizes the process of targeted syntactic evaluation and provides easy tools for analysis and visualization; 2. Two command-line tools, ‘syntaxgym‘ and ‘lm-zoo‘, which allow any user to reproduce targeted syntactic evaluations and general language model inference on their own machine.

Contextual influences on language often exhibit substantial cross-lingual regularities; for example, we are more verbose in situations that require finer distinctions. However, these regularities are sometimes obscured by semantic and syntactic differences. Using a newly-collected dataset of color reference games in Mandarin Chinese (which we release to the public), we confirm that a variety of constructions display the same sensitivity to contextual difficulty in Chinese and English. We then show that a neural speaker agent trained on bilingual data with a simple multitask learning approach displays more human-like patterns of context dependence and is more pragmatically informative than its monolingual Chinese counterpart. Moreover, this is not at the expense of language-specific semantic understanding: the resulting speaker model learns the different basic color term systems of English and Chinese (with noteworthy cross-lingual influences), and it can identify synonyms between the two languages using vector analogy operations on its output layer, despite having no exposure to parallel data.