“‘Understanding’ and prediction: Disentangling meaning extraction and predictive processes in humans and AI“
The Transdisciplinary Institute in Applied Data Sciences (TRIADS) speaker series is co-hosted by The Center For Empirical Research in the Law and the Department of Psychological & Brain Sciences.
Abstract: The interaction between “understanding” and prediction is a central theme both in fields that study language processing in humans, and in fields that engineer natural language processing (NLP) in artificial intelligence. Evidence indicates that the human brain engages in predictive processing while extracting the meaning of language in real time, while NLP models in AI use training based on prediction in context to learn strategies of language “understanding”. In this talk I will discuss work that tackles key problems in both of these domains by exploring and teasing apart effects of systematic meaning understanding and effects of statistical-associative processes associated with prediction. I will begin with work that diagnoses the linguistic capabilities of NLP models, showing that with properly controlled tests we uncover important limitations in the ability of current NLP models to understand language robustly as humans do — however, these models show signs of alignment with statistical sensitivities seen in human real-time processing. Leveraging this knowledge, I will then turn to work that directly models the mechanisms underlying human real-time language comprehension, where I will show that by combining psycholinguistic theory with targeted use of measures from NLP models, we can strengthen the explanatory power of psycholinguistic models and achieve nuanced accounts of interacting factors underlying a wide range of observed effects in human language comprehension.
For inquiries contact TRIADS@wustl.edu.