The Curious Case of Control
This research addresses the problem of understanding linguistic biases in AI models for researchers in computational linguistics and AI, but it is incremental as it builds on prior work on heuristics and semantic roles.
The study investigated whether large language models exhibit systematic errors in subject and object control sentences similar to children, finding that models can be grouped into three categories, with the largest group following positional heuristics that succeed on subject control but fail on object control despite object control being more frequent in training data.
Children acquiring English make systematic errors on subject control sentences even after they have reached near-adult competence (C. Chomsky, 1969), possibly due to heuristics based on semantic roles (Maratsos, 1974). Given the advanced fluency of large generative language models, we ask whether model outputs are consistent with these heuristics, and to what degree different models are consistent with each other. We find that models can be categorized by behavior into three separate groups, with broad differences between the groups. The outputs of models in the largest group are consistent with positional heuristics that succeed on subject control but fail on object control. This result is surprising, given that object control is orders of magnitude more frequent in the text data used to train such models. We examine to what degree the models are sensitive to prompting with agent-patient information, finding that raising the salience of agent and patient relations results in significant changes in the outputs of most models. Based on this observation, we leverage an existing dataset of semantic proto-role annotations (White, et al. 2020) to explore the connections between control and labeling event participants with properties typically associated with agents and patients.