United States

We examined whether embodiment effects at a particular word influence subsequent words. We recorded EEG while participants read action sentences that were literal-concrete (LC), literal-abstract (LA), and metaphorical (MET). Prior work showed that at the verbs, both LC and MET elicited more negative N400s than LA, reflecting sensorimotor simulations. We found that at the object nouns, LC elicited a more negative negativity in the 500-700 ms time window   than LA and MET, which may reflect combined influences from embodiment spilled over from the verbs and imagery linked to noun concreteness and imageability. These findings suggested that literal action embodiment can yield extended motor engagement extending into subsequent words, but metaphorical action verbs cannot. Moreover, lexical properties at the noun played a role.

CogSci 2025

What does action do to object? The case of metaphoric action.

electroencephalography (eeg)

cognitive neuroscience

embodied cognition

language comprehension

semantic memory

## Welcome to NAACL 2022!
The Annual Meeting of the North American Chapter of the Association for Computational Linguistics (NAACL) is taking place July 10-15, 2022 as a hybrid event, in Seattle, WA and online. We are happy to welcome all of you to conference! 
 
The main conference program features oral presentations, in-person and virtual posters and demo sessions, a plenary session for our best paper presentations and awards, keynote presentations, a plenary panel on the place of linguistics and symbolic structures in NLP, and an Industrial Track panel on Careers in NLP.
 
Posters (including Findings of NAACL 2022) and demos are grouped by areas for both the in-person and the virtual sessions. For the virtual component, the talks will be on Zoom and the posters and the demos will be in GatherTown. The Student Research Workshop and will have an oral session and a poster session.
 
The program also features 6 Tutorials and 26 Workshops.
 
 We wish you a wonderful conference!
 
[The NAACL 2022 Organizing Committee](https://2022.naacl.org/committees/organization/)

NAACL 2022

2022 Annual Conference of the North American Chapter of the Association for Computational Linguistics

poster

### Welcome to CogSci Conference 2025!

The 47th Annual Meeting of the Cognitive Science Society was a hybrid meeting held in San Francisco. 

<div style="position:relative;padding-top:0;width:900px;height:500px;"><iframe style="position:absolute;border:none;width:100%;height:100%;left:0;top:0;" src="https://online.fliphtml5.com/ebtyf/amvr/" seamless="seamless" scrolling="no" frameborder="0" allowtransparency="true" allowfullscreen="true" ></iframe></div>

#### About

The Cognitive Science Society brings together researchers from around the world who hold a common goal: understanding the nature of the human mind. The mission of the Society is to promote Cognitive Science as a discipline, and to foster scientific interchange among researchers in various areas of study, including Artificial Intelligence, Linguistics, Anthropology, Psychology, Neuroscience, Philosophy, and Education.

The Society is a non-profit professional organization and its activities include sponsoring an annual conference and publishing the journals Cognitive Science and TopiCS.

#### Our History 

* **Society Creation** 
The Society was incorporated as a 501(c)(3) non-profit professional organization in Massachusetts in 1979. The organizing committee included Roger Schank, Allan Collins, Donald Norman, and a number of other scholars from psychology, linguistics, computer science, and philosophy. 
 
* **Conference Creation** 
The first conference on cognitive science was held at La Jolla, California in August, 1979, and has occurred annually since then. The proceedings of each conference are published, and those from most years are available through Lawrence Erlbaum Associates, Inc. The annual proceedings of the Cognitive Science Conference represent a major source of information on new work and new ideas in the scientific study of thinking. In 1990, the Society, with help from an anonymous donor, established the David Marr Prize for the best student paper at each annual meeting.
 
* **Journal Creation** 
The Journal, Cognitive Science, began publication in 1976, and is now published by Wiley-Blackwell. The Executive Editor is currently Richard P. Cooper of Birkbeck, University of London, and there are 18 Associate Editors and a 30-member editorial board. It serves as the premier outlet for research reports that intersect two or more disciplines. Copyrights for articles published in the journal are held by the Society. The Governing Board of the Cognitive Science Society voted in late 2006 to found a new journal, Topics in Cognitive Science (topiCS). The Editor in Chief is Wayne Gray, Cognitive Science Department, Rensselaer Polytechnic Institute. The journal seeks to fill a niche not occupied by Cognitive Science Journal or other cognitive science journals. Membership in the Society includes a subscription to Cognitive Science and TopiCS. Copyrights for articles published in the journal are held by the Society.
 

#### Code of Conduct

By attending the CogSci 2025 Conference, you are required to adhere to the society’s **[Code of Conduct](https://drive.google.com/file/d/1ChPuihLy6jE_BWqfO7J2KKgX35JW2zsM/view?usp=sharing)**.

You need to log in with the email address you registered with. 

Login credentials were sent to you from Underline -  subject line "Welcome to the CogSci 2025 Conference". Please be sure to check your spam/promotional inbox  if you do not see an email confirmation right away.





Please log in to join this event.

To access the site, please register [**here**](https://cognitivesciencesociety.org/registration/).

If you are registered and feel like you are seeing this message by mistake, please make sure you are logged in with the same email that you registered with. 

Please register!

The 47th Annual Meeting of the Cognitive Science Society presents the latest research across cognitive science and highlights the theme of Cognition in Context.

Rather than merely solving math problems, students that self-explain correct and incorrect worked examples increase their procedural knowledge and develop deeper conceptual understandings of key concepts. The present study is a large-scale test of the efficacy of the MathByExample intervention, which targets key math concepts and common misconceptions for students in 5th grade through correct and incorrect worked examples. Our cluster-randomized controlled trial included 42 schools (n = 830 5th grade students) across two cohorts. Preliminary results show effects in the predicted direction, students who received MathByExample exercises outperformed students in the control condition, yet the difference is not significant. Our poster will discuss possible explanations for the findings, discuss exploratory moderators (e.g., dosage received of MathByExample exercises), and include data from a third cohort for which data collection is currently ongoing.

Evaluating the Efficacy of MathByExample: Preliminary Evidence 

We study how children develop a causal curriculum to achieve a challenging goal that is not solvable at first. Adopting the Procgen environments that include various challenging game tasks, we found that 5- to 7-year-old children actively used their current level competence to determine their next step in the curriculum and made improvements to their performance during this process as a result. This suggests that children treat their level competence as an intrinsic reward, and are motivated to master easier levels in order to do better at a more difficult one, even without explicit reward. However, our findings also suggest that children’s self-designed curricula may not always be the most effective design. Rather, repeatedly practicing on the difficult target task may be sufficient. Notably, when constrained to stay on the target task instead of crafting their own curricula, more children actually succeeded and made greater progress in the game, suggesting that children perceive a curriculum as beneficial, even when focusing on a singular difficulty might prove more effective.

Children Spontaneously Design Curricula to Tackle Challenging Tasks

Humans can track objects and predict their motion even when they are temporarily occluded. How does the absence of changing visual evidence alter predictive beliefs about a moving object? In our study, participants were tasked with continuously anticipating the destination of a simulated ball in occluded and un-occluded 2.5D environments. Our findings reveal that humans actively update their judgments throughout the period of occlusion while making predictions grounded in physical realism, even as occlusion impairs accuracy. To model this behavior, we integrate perception with physical reasoning, unifying tracking and prediction. This is implemented via massively parallel probabilistic inference in a hierarchical generative model for the motion of intermittently visible objects, represented using the GenJAX probabilistic programming platform. This model predicts time-varying human judgments more accurately than alternative models, suggesting that humans integrate perception and physics to reason about occluded motion. Paper is available at https://arijit-dasgupta.github.io/jtap/.

Seeing through Occlusion: Uncertainty-aware Joint Physical Tracking and Prediction

Some people like coffee while others prefer tea, but little is known about whether preferences like these are shared among groups and whether they vary systematically across many common semantic categories. 
This study addresses this gap by examining two major sources of variation -- gender and extraversion -- across twelve categories or domains, ranging from fruit and animals to sports and personal qualities. In Study 1, participants rated their own preferences for a set of 300 exemplars. Results showed significant preference differences between men and women for 40% of items spread across all categories, and smaller but reliable differences between introverts and extraverts for 11% of items concentrated in domains like personal qualities. Study 2 used an allocentric categorisation task where the same participants categorised items based on which they thought would be preferred by men vs women or introverts vs extraverts. Using the ratings from Study 1 to score accuracy, the judgments from Study 2 showed that participants were sensitive to even subtle differences in preference, although accuracy varied by the judge's gender and extraversion: women were more accurate than men across many categories and introverts more accurate than extraverts for a few categories.
We also found incorrect but widely shared judgments for about 20% of items, suggestive of inaccurate stereotypes about group preferences. 
Together, these results suggest widespread and systematic variation by gender (and to a lesser extent extraversion) that can be accurately predicted by others, although with systematic biases. Our results have implications for theories of semantic representation and social cognition.

Who Likes What? Comparing Personal Preferences with Group Predictions based on Gender and Extraversion Across Common Semantic Domains.

Human experience is rich with sensory information, from which
physical regularities are internalized to guide behavior. This
paper presents the Patchwork Approach, a method for modeling
sensorimotor predictions based on these regularities, without explicitly
encoding physics. This method leverages environmental
regularities, enabling resource-rational perceptual predictions.
Using data from previous studies (Deeb et al., 2021, 2024) on
human perception, the model outperforms Newtonian-based
models in capturing human prediction errors and interpolates
across untested conditions. One test demonstrates its superiority
in a projectile motion task, while another illustrates its ability
to predict deflection angles in a collision event from untrained
aiming angles. This approach provides valuable insights into
how physical laws are internalized and used to guide perception
and action–arguing that perception provides foundational
input to intuitive physics reasoning, a role that can complement
and be extended to higher-level cognitive processes like those
explained by the Intuitive Physics Engine (IPE).

The Patchwork Approach: Toward a Perceptual Theory of Intuitive Physics

Causal reasoning is a core component of intelligence. Large language models (LLMs) have shown impressive capabilities in generating human-like text, raising questions about whether their responses reflect true understanding or statistical patterns. We compared causal reasoning in humans and four LLMs using tasks based on collider graphs, rating the likelihood of a query variable occurring given evidence from other variables. 
LLMs’ causal inferences ranged from often nonsensical (GPT-3.5) to human-like to often more normatively aligned than those of humans (GPT-4o, Gemini-Pro, and Claude). Computational model fitting showed that one reason for GPT-4o, Gemini-Pro, and Claude's superior performance is they didn't exhibit the "associative bias'' that plagues human causal reasoning. Nevertheless, even these LLMs did not fully capture subtler reasoning patterns associated with collider graphs, such as "explaining away". These findings underscore the need to assess AI biases as they increasingly assist human decision-making.

Do Large Language Models Reason Causally Like Us? Even Better?

Financial decisions such as those involved in stock trading should, at least partly, be based on similar features as detecting trends in time-series data. When presented as a purely perceptual task, people’s accuracy in detecting trends is generally considered good, but real-world individual investors underperform the market globally. In a series of controlled experiments, we contrast financial decisions to perceptual ones, presenting participants with real-time evolving time series whilst manipulating the reward structure and the context. Our results show that participants’ decisions were not affected by trend direction in a classic perceptual decision-making scenario, whilst in a classic trading scenario they performed worse in both speed and accuracy during downward (i.e., bear markets) compared to upward trends (i.e., bull markets). In a final experiment where we carefully controlled the reward structure of both scenarios and the only relevant differentiating factor were the labels of the decisions, we did not find evidence for this difference between scenarios, but participants were slower in the trading scenario. Employing the Drift-Diffusion Model, we found evidence of lower efficiency in the classic trading, compared to the classic perceptual decision-making scenario. Our results provide much-needed insight into the cognitive basis of trading decisions and the general underperformance of real-world individual investors.

Watch out for Bears: Do People Behave Differently in Perceptual and Financial Decisions?

Analysis of adult language shows positive correlations between a word’s affective features (valence, arousal) and those of its surrounding language context, but whether this extends to children’s language remains unclear. To address this, we quantified the emotional context of words within a large corpus of children’s written stories (N > 100,000, ages 7-13). Following Snefjella and Kuperman’s (2016) procedures, we defined a word’s context as the five content words immediately preceding and following it (10 in total). Context valence and arousal were computed for each occurrence and averaged across all occurrences in the corpus, yielding affective values for 24,383 words and their contexts. We found positive correlations between context-level and word-level affective variables (valence: r = 0.46, arousal: r = 0.32), consistent across the 7-13 age range. This study extends adult findings to children’s written language and provides a resource for future research on emotional contexts in language development.

Quantifying the context-level valence and arousal in children’s written language

Understanding both the processes and products of science are
core components of science literacy, but do these types of
knowledge interact during learning? We propose that Nature of
Science (NoS) understanding can act as a schema to facilitate
comprehension of science content. Across two experiments,
we tested whether NoS lessons about theory change improve
students’ comprehension of psychology lessons which are centered
around theory development. In Experiment 1, undergraduates
who watched a NoS lesson showed improved NoS understanding,
but this understanding did not lead to better comprehension
of a matched psychology lesson compared to control.
In Experiment 2, three NoS lessons were experimentally
integrated into a college psychology course, preceding content
lessons involving theory change. While this intervention
did not improve learning, we found several relationships between
science beliefs and academic performance. This work
contributes to our limited understanding of how these distinct
components of science knowledge interact during learning.

Does Connecting the Processes and Products of Science Facilitate Learning? A Schema-Based Approach

What algorithms do LLMs actually learn and use to solve problems? Studies addressing this question are sparse, as research priorities are focused on improving performance through scale. Here we introduce a framework for systematic research into the algorithms that LLMs learn and use (AlgEval). Toward this goal, we conducted a graph navigation study that typically requires multi-step search, and evaluated whether Llama-3.1-8B uses classic search algorithms. We formed top-down hypotheses about candidate algorithms (e.g., breadth first, BFS, or depth first search, DFS), and tested these hypotheses via circuit-level analysis of attention patterns and hidden states or representations. We found that 1) Extracting possible sequences processed by the model's layer-by-layer representations did not support either BFS or DFS. 2) Attention patterns showed a cascading shift toward the correct path as the prompt was processed. 3) Projecting node-token representations across layers to a manifold revealed gradual separation of the goal from its competitor in representation space. Overall, our results don't support the idea that the model relies on forming or using an accurate map of the environment, and instead of a step by step search, it seems to rely on more policy-dependent shifts. Future work can connect these findings to failure modes in multi-step reasoning. A rigorous, algorithmic evaluation of how LLMs solve tasks offers an alternative to resource-intensive scaling, potentially enabling more sample-efficient training, performance, and novel architectures.

Downloads

Next from CogSci 2025

Evaluating the Efficacy of MathByExample: Preliminary Evidence

Similar lecture

Disentangled Action Recognition with Knowledge Bases

.css-70qvj9{display:-webkit-box;display:-webkit-flex;display:-ms-flexbox;display:flex;-webkit-align-items:center;-webkit-box-align:center;-ms-flex-align:center;align-items:center;}Downloads

Next from CogSci 2025

Evaluating the Efficacy of MathByExample: Preliminary Evidence

Similar lecture

Disentangled Action Recognition with Knowledge Bases

Downloads