Causality Elicitation from Large Language Models
This work provides a framework for inspecting the causal hypotheses implicitly assumed by LLMs, which is significant for researchers interested in understanding the internal knowledge representation of these models.
This paper proposes a pipeline to extract causal relationships from large language models (LLMs) by sampling documents, extracting events, grouping them into canonical events, constructing indicator vectors, and applying causal discovery methods to estimate candidate causal graphs.
Large language models (LLMs) are trained on enormous amounts of data and encode knowledge in their parameters. We propose a pipeline to elicit causal relationships from LLMs. Specifically, (i) we sample many documents from LLMs on a given topic, (ii) we extract an event list from from each document, (iii) we group events that appear across documents into canonical events, (iv) we construct a binary indicator vector for each document over canonical events, and (v) we estimate candidate causal graphs using causal discovery methods. Our approach does not guarantee real-world causality. Rather, it provides a framework for presenting the set of causal hypotheses that LLMs can plausibly assume, as an inspectable set of variables and candidate graphs.