Canada

We introduce purely visual paradigms that convey the logical structure of illusory inferences from disjunction: (a AND b) OR c, a |- b. Although the logical information was conveyed entirely via non-linguistic means, we found that the visual paradigms induce reasoning fallacies, though less attractive than their linguistic counterparts. The visual paradigms highlight the role of alternative-based reasoning, or question-answer dynamics, as they control for narrowly interpretive processes that confound the study of their linguistic counterparts. To our knowledge, this is the first work to develop visual paradigms that represent reasoning fallacies committed by adults and involve multiple logical operators non-trivially embedded. Previous studies focused on pre-verbal children or non-human animals, and for this reason limited the scope of research to visually representing logically simple, valid inferences.

CogSci 2022

Question-answer dynamics in deductive fallacies without language

## Welcome to EMNLP 2022!
I am delighted to welcome you to EMNLP 2022! I believe this conference has been complicated beyond any precedent. Over the past year, it’s been thrilling to see the organization team approach each new puzzle with creativity and enthusiasm. We hope that those participating in Abu Dhabi as well as those joining remotely will leave the conference feeling newly inspired by the program and newly connected to our ever-growing community. Following EMNLP 2021 and major NLP conferences since, EMNLP 2022 is “hybrid,” serving both virtual and in-person participants.

Our key innovations for EMNLP 2022 include:

* EMNLP 2022 is “hybrid” in a second sense, as well: we allowed both direct and rolling review paper submissions, building on the pilot experiment of EMNLP 2021, which considered a small number of ARR submissions. 
* Familiar from NAACL but new to EMNLP, we’ve added an industry track.
* During the conference, “portals” will link virtual poster sessions to in-person conference participants during poster sessions each day.
* The first ACL-family conference in the United Arab Emirates.

 *Message from Noah A. Smith, University of Washington and Allen Institute for AI, Seattle, Washington, USA* 
***EMNLP 2022 General Chair***
 
[![](https://assets.underline.io/uploads/markdown_image/1/image/9eec7d4a287ee18c278b08229290aa83.png)](https://drive.google.com/file/d/1OlPv6QBeo62VVTughj2jkiLeyHd1WnUt/view)
 
[![](https://assets.underline.io/uploads/markdown_image/1/image/a3db7a768409f05192210d98601edb25.png)](https://emnlp2022.rocket.chat/)

To access this site you need to register. Please register [here](https://2022.emnlp.org/registration/).

Register here

EMNLP 2022

Welcome!
EMNLP 2022 will take place in Abu Dhabi from December 7th to December 11th, 2022. And it will be held in hybrid mode, both online and offline.

technical paper

Join us for CogSci2022! Our 44th Annual Conference will be centered around the theme of diversity, with an accompanying range of keynotes and invited speakers who themselves highlight the diversity of our field. Our three symposia are Cognition across Diverse Populations, Cognition within Diverse Societies, and Global Challenges for Cognitive Science. The conference is being offered in a fully hybrid environment for onsite and online attendees.
 
 [**The Program at a Glance can be found here**](https://cognitivesciencesociety.org/wp-content/uploads/2022/07/Cogsci-2022-PAAG-Detailed-July-12final-1.pdf)
 [**The Detailed Program can be found here**](https://cognitivesciencesociety.org/wp-content/uploads/2022/08/DeArmond-CogSci2022-program-v113-final.pdf)

This conference requires a registration to attend. Please use the link below to secure your ticket. The website will be available to all registered attendees on July 25th. 

[Register to Attend](https://cognitivesciencesociety.org/registration/)

Registration Is Required

The 44th Annual Meeting of the Cognitive Science Society highlighted research on the theme Cognitive Diversity in addition to the full breadth and diversity of research topics offered by the society’s membership. 





Investigation of the relationship between cognitive function and academic performance has recently pivoted from differences in intelligence to executive function. To date, these studies have focused disproportionately on samples recruited from Western countries, despite evidence in support of cultural differences in these putative relationships. To address this gap, the present study investigated whether differences in inhibitory and/or attentional control could predict academic performance in a sample of Chinese adolescents (n=42). Participants reported on demographic details and completed both the Simon task and Attention Network Test. Data were analyzed using multiple linear regression controlling for gender, age, SES, English language proficiency, processing speed, and fluid intelligence. Results showed that one index of inhibitory control derived from flanker task performance explained a significant amount of unique variance in academic performance. Our findings provide evidence that executive function, specifically inhibitory control, plays a significant role in academic performance.

Inhibitory Control Predicts Academic Performance Beyond Fluid Intelligence and Processing Speed in English-Immersed Chinese High Schoolers

Parents and teachers often encourage students (e.g., "You can do it!") when they encounter challenges, but these messages are not always effective. Whose encouragement motivates students the most, and why? Here we tested the hypothesis that others’ domain knowledge (e.g., knowledge about course materials) and ability knowledge (e.g., knowledge about students’ abilities in the course) each inform how students evaluate their encouragement. In a large-scale survey, we find that middle school students (n=288) and high school students (n=425) are most likely to seek out and be motivated by encouragement from someone with both domain and ability knowledge, rather than only one or the other. This effect emerged both when students reasoned about hypothetical classmates (Study 1a) and real people in their lives (Study 1b). Moreover, we find that confidence in others' performance estimates linearly increases when they have greater ability and domain knowledge (Study 1c). Collectively, this work suggests that students do not find all encouragement equally motivating. Rather, students find encouragement most motivating when the speaker has knowledge of their abilities and the domain.

Adolescents are most motivated by encouragement from someone who knows their abilities and the domain

Research in cognitive science indicates that the effects of corrective feedback are powerful, but quite variable. Stemming from the Feedback Intervention Theory, we tested two features of feedback that may influence the learner’s attention to the self and ultimately their performance. In this zoom-based experiment, 6 to 8 year old children (N = 102) completed an online learning activity focused on solving mathematical equivalence problems. During the learning activity, children were assigned to different conditions which varied both feedback source (person vs. computer) and feedback type (correct-answer vs. correct-answer+verification). Feedback source was found to affect accuracy, where computer-based feedback resulted in higher accuracy compared to person-based feedback. Additionally, feedback type was found to affect strategy variability. Low knowledge children were more likely to use a variety of different strategies when they received correct-answer only feedback, while high knowledge children were less affected by the presence or absence of verification cues.

The source and type of feedback influence children's mathematics performance

We describe UMA (Unified Model of Arithmetic), a theory of children’s arithmetic implemented as a computational model. UMA extends a theory of fraction arithmetic (Braithwaite et al., 2017) to include arithmetic with whole numbers and decimals. We evaluated UMA in the domain of decimal arithmetic by training the model on problems from a math textbook series, then testing it on decimal arithmetic problems that were solved by 6th and 8th graders in a previous study. UMA’s test performance closely matched that of children, supporting three assumptions of the theory: (1) most errors reflect small deviations from standard procedures, (2) between-problem variations in error rates reflect the distribution of input that learners receive, and (3) individual differences in strategy use reflect underlying variation in learning parameters.

Testing a Unified Model of Arithmetic

Common critiques of natural language processing (NLP)
methods cite their lack of multimodal sensory information,
claiming an inability to learn situated, action-oriented relations
through language alone. Barsalou’s (1983) theory of
ad hoc categories, which are formed from to achieve goals in
real-world scenarios, correspond theoretically to those types
of relations with which language models ought to have great
difficulty. Recent NLP models have developed dynamic approaches
to word representations, where the same word can
have different encodings depending on the context in which it
appears. Testing these models using categorization tasks with
human response data demonstrates that situated properties may
be partially captured through semantic analysis. We discuss
possible ways in which different notions of situatedness may
be distinguished for future development and testing of NLP
models.

Getting Situated: Comparative Analysis of Language Models With Experimental Categorization Tasks

Semantic features have been playing a central role in investigating the nature of our conceptual representations. Yet the time and effort required to sample features from human raters has restricted their use to a limited set of manually curated concepts. Given recent success of transformer-based language models, we asked whether it was possible to use such models to automatically generate meaningful lists of properties for arbitrary object concepts and whether these models would produce features similar to those found in humans. We probed a GPT-3 model to generate semantic features for 1,854 objects and compared them to existing human feature norms. GPT-3 showed a similar distribution in the types of features and similar performance in predicting similarity, relatedness, and category membership. Together these results highlight the potential of large language models to capture important facets of human knowledge and yield a new approach for automatically generating interpretable feature sets.

Semantic features of object concepts generated with GPT-3

We present a neural dynamic model that perceptually grounds nested noun phrases, i.e., noun phrases that contain further (possibly also nested) noun phrases as parts. The model receives input from the visual array and a representation of a noun phrase from language processing. It organizes a search for the denoted object in the visual scene. The model is a neural dynamic architecture of interacting neural populations which has clear interfaces with perceptual processes. It solves a set of theoretical challenges, including the problem of keeping a nested structure in short-term memory in a way that solves the problem of 2 and massive binding problem emphasized by Jackendoff (2002). The model organizes a search for the objects that are referenced in that structure. We motivate the model, demonstrate simulation results, and discuss how it differs from related models.

A Neural Dynamic Model Perceptually Grounds Nested Noun Phrases

Human language offers a powerful window into our thoughts -- we tell stories, give explanations, and express our beliefs and goals through words. Abundant evidence also suggests that language plays a developmental role in structuring our learning. Here, we ask: how much of human-like thinking can be captured by learning statistical patterns in language alone? We first contribute a new challenge benchmark for comparing humans and distributional large language models (LLMs). Our benchmark contains two problem-solving domains (planning and explanation generation) and is designed to require generalization to new, out-of-distribution problems expressed in language. We find that humans are far more robust than LLMs on this benchmark. Next, we propose a hybrid Parse-and-Solve model, which augments distributional LLMs with a structured symbolic reasoning module. We find that this model shows more robust adaptation to out-of-distribution planning problems, demonstrating the promise of hybrid AI models for more human-like reasoning.

Structured, flexible, and robust: benchmarking and improving large language models towards more human-like behavior in out-of-distribution reasoning tasks

Limited evidence shows that visual input can facilitate learning novel sound-to-meaning mappings that are crucial to learning a second language. However, the mechanisms by which visual information influences auditory learning are still unclear. Here, we investigate to what extent visual input can lead to effective learning in another domain. We trained atonal speakers with Mandarin tones in 4 conditions: Auditory Only (AO) where only auditory tones were given as input; Animated Contour (AC) where moving visual pitch contours indicating the dynamic changes of tones were given in addition to auditory tones; Static Contour (SC) where static visual pitch contours were given in addition to auditory tones; Incongruent Contour (IC) where mismatched pitch contours were given in addition to auditory tones. The results show the advantage of AC and SC over AO in learning tonal categories and that IC inhibits learning, suggesting that extracting ‘compatible’ properties cross modalities benefits learning most.

Cross-modal perceptual learning in learning a tonal language

Some linguistic features are more readily learned than others, and are thereby more likely to be maintained in diachronic language change, giving rise to typological universals. Less readily learned features may give rise to typological gaps. We consider an apparent typological gap—that a morphologically superlative determiner (e.g., gleebest in "gleebest of the cows") with a negative meaning is cross-linguistically unattested—and ask whether it reflects an underlying learning bias. We find 4-year-olds know that such determiners indicate quantity (replicating Wellwood, Gagliardi, & Lidz, 2016), but only when positive (‘most’), but not negative (‘least’). Importantly, the observed bias is not specific to the apparent typological gap: same-age children showed difficulty learning the negative meaning of a non-superlative determiner, though such meanings are attested. The data thus suggest that children are generally biased against negativity, consistent with much prior work on conceptual bias and language learning/processing.

Premium content

Downloads

Next from CogSci 2022

Inhibitory Control Predicts Academic Performance Beyond Fluid Intelligence and Processing Speed in English-Immersed Chinese High Schoolers

Similar lecture

Logical Fallacy Detection