United States

It is well known that verbal analogy (VA) performance correlates with IQ; we hypothesize high IQ improves VA performance due to more comprehensive relation representation, and therefore priming with congruent (incongruent) relations should especially benefit (hurt) lower-IQ individuals. We find a significant overall priming effect (p&lt;0.001), IQ effect (p&lt;0.001), and a quadratic IQ effect on priming (p=0.012): mid-IQ shows greater priming effect than high-IQ, but low-IQ participants show no effect – due to lower priming accuracy (p&lt;0.001) and the positive contribution of priming accuracy to priming effect (p=0.04). Separately, LLMs show similar priming and IQ effects (when prompted to be low- or high-IQ), with a change in error pattern between congruent and incongruent priming just like humans, suggesting possibly similar mechanisms. Overall, these findings deepen our understanding of the role of relation representation in VA and the influence of IQ.

CogSci 2025

Priming Effects on Verbal Analogy Performance Across IQ Levels in Humans and AI

intelligent agents

computational modeling

psychology

**Welcome to EMNLP 2023!**

On behalf of the EMNLP 2023 Organizing Committee, I extend a warm and heartfelt welcome to all of you to the 2023 Conference on Empirical Methods in Natural Language Processing (EMNLP). It is with immense pleasure and excitement that we gather here in Singapore, a vibrant hub of innovation and technological advancement.

The conference program is packed with insightful presentations, thought-provoking workshops, and engaging networking opportunities. In addition to the technical sessions, we have also planned several social events that will provide you with the opportunity to connect with your colleagues.

The conference is held in person at the Resorts World Convention Centre in Singapore, and available on-line with the help of Underline.

For those in Singapore, I hope that you will find time to explore the exciting Sentosa Island, Gardens by the Bay, Singapore Botanic Gardens and the many other attractions unique to Singapore.

Of course, we are grateful to our sponsors and partners for their generous support of EMNLP 2023, whose contributions make it possible for us to host this world-class event.

Yuji Matsumoto (RIKEN AIP)<br>
EMNLP 2023 General Chair<br><br><br><br>

To access the **EMNLP 2023** event page on Underline, you need to register for the Conference.<br>
Please follow **[this link](https://2023.emnlp.org/registration/)** for more details. 

EMNLP 2023

EMNLP 2023 took place in Singapore from Dec 6th to Dec 10th, 2023.

It is well known that verbal analogy (VA) performance correlates with IQ; we hypothesize high IQ improves VA performance due to more comprehensive relation representation, and therefore priming with congruent (incongruent) relations should especially benefit (hurt) lower-IQ individuals. We find a significant overall priming effect (p<0.001), IQ effect (p<0.001), and a quadratic IQ effect on priming (p=0.012): mid-IQ shows greater priming effect than high-IQ, but low-IQ participants show no effect – due to lower priming accuracy (p<0.001) and the positive contribution of priming accuracy to priming effect (p=0.04). Separately, LLMs show similar priming and IQ effects (when prompted to be low- or high-IQ), with a change in error pattern between congruent and incongruent priming just like humans, suggesting possibly similar mechanisms. Overall, these findings deepen our understanding of the role of relation representation in VA and the influence of IQ.

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**<br>
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. 
<br><br>
* **Conference Creation**<br>
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.
<br><br>
* **Journal Creation**<br>
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.
<br><br>

#### 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)**.
<br><br>


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.

Symbolic visual sketching is a hallmark of human
creativity, enabling the externalization of abstract concepts
through figurative representations. Yet, creative
expression can be constrained by pervasive conceptual
associations—culturally learned mappings between
abstract ideas and standard visual forms (e.g., a dove
symbolizing peace). Generative AI has the potential to
liberate such fixations due to AI’s access to a broad range
of content and ideas, but it remains unclear whether and
how inspiration from verbal or visual modalities better
mitigates fixations. Here, we hypothesized that the verbal
modality induces greater conceptual divergence than the
visual modality by bypassing perceptual constraints,
whereas the visual modality may reinforce perceptually
familiar mapping of visual representations. Participants
generated sketches of abstract concepts (e.g., "time")
before and after receiving GPT-4-generated verbal or
visual inspiration. Drawings were analyzed using deep
neural networks—by comparing perceptual features
(VGG16-based) and semantic-perceptual content (CLIP-
based)— as well as both human and GPT-4 scoring for
creativity. We found that verbal inspiration significantly
increased semantic distance and uniqueness, whereas
visual inspiration led to minimal semantic divergence from
the initial sketches. Importantly, low-level perceptual
features remained unchanged across conditions, indicating
that verbal prompts primarily influenced high-level
conceptual framing of the sketches rather than their visual
features. These findings demonstrate the effect of modality
on mitigating cognitive fixations, with the verbal modality
enhancing more unconventional visual sketching.

Sketching with generative AI: verbal but not visual inspiration mitigates cognitive fixations

Child-directed speech (CDS) is characterized by its adaptive nature: Caregivers not only talk to children, but engage in dy- namic interactions with them. The adaptive/interactive nature of this type of language is understudied in computational mod- eling research, particularly given the limited availability of nat- uralistic data. While recent advances in large language models (LLMs) have demonstrated potential for generating viable syn- thetic dialogue data in various domains, their ability to capture the dynamics of child-caregiver communication remains un- explored. This paper introduces a systematic framework for evaluating LLMs’ capacity to generate developmentally ap- propriate CDS in interaction, examining both static linguistic features and dynamic conversational patterns. We evaluated state-of-the-art LLMs (GPT-4o and Llama 3) against natural interactions from the CHILDES dataset using both single- and multi-turn testing approaches. In single-turn evaluation, mod- els generated responses to individual child utterances, enabling direct comparison with actual caregiver responses. Multi-turn testing assessed sustained interaction capabilities through sim- ulated child-caregiver dialogues. Our results show that while LLMs can successfully approximate surface-level linguistic patterns after few-shot prompting, they struggle with higher- level communicative aspects, with excessive alignment and re- duced diversity compared to natural interactions. Our bench- marking framework elucidates both the potential and limita- tions of LLMs in generating data that preserves the essential properties of child-caregiver language in interactions.

Benchmarking LLMs for Mimicking Child-Caregiver Language in Interaction

Stereotypes change over time and across cultures, but they are hard to change in experiments. This paper proposes a solution by reconceptualizing stereotypes not as simple associations between groups and traits, but as rich networks of interconnected concepts spanning multiple domains. Building on cross-domain mapping, we used a simple question-answering task (''If X were a Y, what Y would it be?'') to create the expansive networks of 100 social groups to eight domains including animals, jobs, sports, colors, beverages, vehicles, musical instruments, and academic subjects. We found, for instance, the stereotype of women lacking agency is part of a larger network where women are associated with preferences for certain drinks such as wine, colors of pink, musical instruments of harps, or sports such as softball. We further tested whether rewiring these broader networks could change stereotypes more effectively than prior methods. Network-based interventions showed promising results for some groups (women were seen as more competent and Muslims were viewed as more friendly), but effects varied for different groups (minimal changes for criminals). This work suggests that successful stereotype change may require engaging with broader networks of subtle, seemingly unrelated associations rather than targeting individual stereotypical beliefs in isolation.

STEREONET: A Network Approach for Stereotype Change

Adults who learn a new language often report feeling that their first language gets in the way. Systematic effects of the first language on additional languages would have straightforward implications for both theory and pedagogical practice if they could be adequately characterized. Unfortunately, this has been shown to be challenging. Languages are vast and complex, and there are a very large number of them. Thus, most studies focus on a few narrowly-defined phenomena and one or two language pairs. The potential complexity of the phenomenon and the sparsity of the observations conspire to make it difficult to establish clear patterns. We present whole- language analyses of the morphosyntax of 133,659 second- language essays spanning 273 L1-L2 pairs. We find clear, consistent effects of the L1 on the morphosyntax of the L2, independent of the L2. We find that not all aspects of morphosyn- tax are equally informative about the L1, suggesting avenues for more precisely specifying how and why L1 influences L2.

Studying Cross-linguistic Structural Transfer in Second Language Learning

Prior research suggests that infants and children prefer people who speak with the same accent as their parents. Here, we investigated whether 13- to 18-month-old monolingual,  American English-learning infants (N=39) and 5- to 8-year-old children from diverse linguistic backgrounds (N=87) use accent to predict social interactions. Participants were familiarized to novel blob-like, English-speaking characters. The central protagonist and one side character shared an accent, while the other side character did not. We varied whether the central character had an American or Chinese accent. In Study 1, when the protagonist in distress had an American accent, but not when the protagonist had a Chinese accent, infants looked first toward the side character with the different accent. In Study 2, 5- to 8-year-old children with American-accented parents were more likely to say that a Chinese-accented character caused distress in an American-accented character. These preliminary findings suggest that infants and children may use accent as a social cue to make inferences about antisocial intent.

Preliminary Evidence that Infants and Children Use Accents to Inform Relational Expectations

Autonoetic consciousness (AuC), originally conceptualized as the phenomenological marker distinguishing episodic from semantic retrieval, faces significant empirical challenges. In particular, episodic retrieval was demonstrated orthogonal from phenomenal characters such as sense of ownership. It has thus been argued that AuC is not a valid construct and should be abolished. While we agree with the empirical failure of AuC’s original conceptualization, we highlight that the phenomenon pertaining to the construct serves practical significance, particularly in terms of supporting one’s sense of self extended through time. Considering empirical cases of dissociations between specific properties pertaining to AuC, including mental time travel, sense of ownership, and introspection, we offer a scientifically-valid reconceptualization of AuC as a specific type of reflective awareness in which explicit metacognition is leveraged to support the sense of personal identity. Under this reconceptualization, episodic retrieval appears as a contingent feature of autonoetic consciousness but not the other way around.

Reconceptualizing Autonoetic Consciousness

The traditional view of biases being cognitive imperfection has been challenged by several strains of research, such as the PSI cognitive architecture. Here, biases are considered to be engineered by evolution, to prevent dissatisfaction and assist subsequent satisfaction of human needs. PSI’s general assumption of higher skills and reasoning capacities alleviating biases has been recently called into question, as high numeracy was associated with an exacerbated effect of political bias. We conduct two studies, the results of which indicate that the basis for this effect 1) does not represent a general cognitive fallacy caused by modulations of perceptional and attentional processes, 2) nor is rooted in the long-term forming of habituated action patterns, associated with prior beliefs. This strengthens the evidence for it to be specific to group dynamics with strong affiliative bounds. Further, we propose a set of revisions to PSI, necessary to model this expert bias phenomenon.

The Logic of Bias: Using Cognitive Architecture to Explore Interactions Between Cognitive Abilities and Decision Error

Prior work has shown that large language models (LLMs) can predict human attitudes based on other attitudes, but this work has largely focused on predictions from highly similar and interrelated attitudes. In contrast, human attitudes are often strongly associated even across disparate and dissimilar topics. Using a novel dataset of human responses toward diverse attitude statements, we found that a frontier language model (GPT-4o) was able to recreate the pairwise correlations among individual attitudes and to predict individuals’ attitudes from one another. Crucially, in an advance over prior work, we tested GPT-4o’s ability to predict in the absence of surface-similarity between attitudes, finding that while surface similarity improves prediction accuracy, the model was still highly-capable of generating meaningful social inferences between dissimilar attitudes. Altogether, our findings indicate that LLMs capture crucial aspects of the deeper, latent structure of human belief systems.

Can Large Language Models Predict Associations Among Human Attitudes?

Misunderstandings in cross-cultural communication often arise from subtle differences in interpretation, but it is unclear whether these differences arise from the literal meanings assigned to words or from more general pragmatic factors such as norms around politeness and brevity.  
In this paper, we report three experiments examining  how speakers of British and American English interpret intensifiers like ``quite'' and ``very,''  finding support for a combination of semantic and pragmatic factors.
To better understand these differences, we developed a computational cognitive model where listeners recursively reason about speakers who balance informativity, politeness, and utterance cost. 
A series of model comparisons suggest that cross-cultural differences in intensifier interpretation stem from (1) different literal meanings, (2) different weights on utterance cost. 
These findings challenge accounts based purely on semantic variation or politeness norms, demonstrating that cross-cultural differences in interpretation emerge from an intricate interplay between the two.

Minding the Politeness Gap in Cross-cultural Communication

Deliberation over a set of options is sometimes needed to choose the most preferred. An example is comparison shopping, where options are products, each with a variety of features. There is evidence that both dorsal and ventral prefrontal cortex (PFC) contribute to these deliberations. We report the results of mouse-tracking experiments involving deliberative choice, constraining a computational cognitive neuroscience model of the role of PFC, and associated brain areas, in shifting attention between option features, as well as in tracking preferences over time. Options are visually displayed, but option features remain hidden until the mouse is moved to their locations. In this way, mouse movements provide a measure of attention allocation. Results reveal strategic patterns of attention shifting which adapt over the course of deliberation and over the course of practice. Observed adaptation is consistent with the reinforcement learning mechanism hypothesized to update the cognitive control states of PFC.

Downloads

Next from CogSci 2025

Sketching with generative AI: verbal but not visual inspiration mitigates cognitive fixations

Similar lecture

DeSIQ: Towards an Unbiased, Challenging Benchmark for Social Intelligence Understanding | VIDEO