Probabilistische Modellierung pragmatischer Verarbeitung und Produktion
Zusammenfassung der Projektergebnisse
The project Pro3 (short for: “Probabilistic modeling of pragmatic processing and production”) investigates how people choose expressions to communicate about complex situations and, more importantly, how people interpret descriptions of complex situations incrementally, i.e., while the speech signal or text is coming in. This much is not particularly new. But an innovative twist of this project was to connect experimental data with probabilistic modeling. It constructed formal probabilistic models of a speaker’s utterance choice, and of a listener’s incremental interpretation process (as each word come in). The latter is the key innovation. The project used a so-called rational analysis of behavior, i.e., asking essentially: “What is the most obvious efficient solution to the general computational problem of incremental implementation; or: how would an idealized information-processing agent do this?” The project then compared the predictions of such a model to human data. The findings are mixed. Some of our results support the idea that human processing resembles the idealized model’s behavior. Yet other results suggest quite clearly that, at least in some contexts or for some tasks, measures of human processing difficulty deviate systematically and strongly from what we would expect under the model’s predictions. Discrepancies between model predictions and experimental data arose in cases where participants also had a rather difficult task of remembering complex visual scenes. On the other hand, the model’s predictions are supported under other, arguably easier, experimental tasks.
Projektbezogene Publikationen (Auswahl)
-
Dynamic speech adaptation to unreliable cues during intonational processing. American Geophysical Union (AGU).
Roettger, Timo Benjamin & Franke, Michael
-
Evidential strength of intonational cues and rational adaptation to (un-)reliable intonation. American Geophysical Union (AGU).
Roettger, Timo Benjamin & Franke, Michael
-
Gricean Expectations in Online Sentence Comprehension: An ERP Study on the Processing of Scalar Inferences. Cognitive Science, 43(8).
Augurzky, Petra; Franke, Michael & Ulrich, Rolf
-
Positional biases in predictive processing of intonation. American Geophysical Union (AGU).
Roettger, Timo Benjamin; Franke, Michael & Cole, Jennifer
-
Pragmatic Prediction in the Processing of Referring Expressions Containing Scalar Quantifiers. Frontiers in Psychology, 12(c(2021, 8, 31)).
Macuch Silva, Vinicius & Franke, Michael
-
This Better Be Interesting: A Speaker’s Decision to Speak Cues Listeners to Expect Informative Content. Open Mind, 6(2022), 118-131.
Rohde, Hannah; Hoek, Jet; Keshev, Maayan & Franke, Michael