Modeling individual and item differences in atypicality inferences with ACT-R
Organizations: Saarland University · Max Planck Institute for Informatics
Abstract
Humans engage in pragmatic reasoning when encountering mentions of highly predictable events. Such mentions may appear redundant, leading comprehenders to infer that the event is atypical, and accommodate this inference with an explanation for why it is atypical. However, these inferences are not drawn uniformly across individuals, and experimental items vary in inference rates due to differences in event typicality. We show that these patterns can be predicted with a resource-rational cognitive model in which comprehenders select interpretation strategies based on individual exploration tendencies and item-level predispositions. These parameters are estimated independently from participants' Raven's Matrices score, event typicality rating and language model based accommodation likelihood. Taken together with recent findings from reference games, our results motivate a new perspective on the mechanisms that support pragmatic inferences.
Figures & tables
| Context | Mary is a journalist who often goes to restaurants after her interviews. Yesterday she went to a popular Chinese place where she ran into her friend David. Later that day David ran into Sally, a mutual friend of him and Mary. |
| Target event | David said to Sally: “I ran into Mary leaving that Chinese place. She ate there! ” |
| Question 1 | How often do you think Mary usually eats, when going to a restaurant? |
| Answer: 82 | |
| Question 2 | Why did you think it is this often? |
| Answer: Going to a restaurant normally involves eating, although one could go for just a drink. | |
| Inference annotation | atypicality |
| Inference Annotation | Inference Drawn | Example |
| normal | no | Usually when you go to a restaurant, it is to eat. |
| atypicality | yes | Going to a restaurant normally involves eating, although one could go for just a drink . |
| notice reject | unclear | After interviews Mary will be tired so she probably eats. She can’t just go to a restaurant for a drink after a long day. |
| not sure | unclear | I’m not sure. |
| other | unclear | He didn’t tell Sally which restaurant, he said that restaurant, as though they go there often. |
| Model | Indiv | Item | Acc. | F1 | Sens. | Spec. | Atypicality | Notice Reject | No Inference | |
| Persis. | Accom | Literal | ||||||||
| Baseline | 51.81 0.52 | 59.25 | 61.78 | 38.76 | 61.73 | 62.08 | 38.76 | |||
| Indiv | ✓ | 56.38 0.30 | 60.63 | 59.19 | 52.73 | 58.67 | 63.04 | 52.73 | ||
| Item | ✓ | ✓ | 55.34 0.34 | 68.54 | 85.79 | 15.42 | 85.90 | 84.96 | 15.42 | |
| Indiv + Item | ✓ | ✓ | 56.73 0.26 | 62.35 | 63.18 | 48.27 | 62.72 | 66.59 | 48.27 | |
| ✓ | ✓ | 56.81 0.31 | 62.67 | 63.91 | 47.53 | 63.14 | 69.58 | 47.53 | ||