United States

The ability to form novel conventions is a key signature of efficient linguistic communication. Reduction of referring expressions, one measure of convention formation, is found robustly in dyadic repeated reference games when the target images are initially difficult to name. In reference games, participants share the explicit goal of establishing joint reference. However, establishing reference is a key subgoal of many conversations where interlocuters have more complex goals. In the current work, we analyze a dataset where reference was embedded in strategic 3- player negotiation and coordination games. In these more complex games, we found that the patterns of reduction and convergence to conventions still held across two different incentive conditions, with some modest differences between the conditions.

CogSci 2023

Communicative reduction in referring expressions within a multi-player negotiation game

It is my great pleasure and honor to welcome you to the 64th Annual Meeting of the Association for Computational Linguistics (ACL 2026), held in San Diego, California, USA, July 2–7, 2026. ACL 2026 continues our fields tradition of excellence in scholarship, innovation, and inclusivity, and I am deeply grateful to the many volunteers who have worked tirelessly to bring this event to life.

It is exciting to see the many developments in the field reflected at our conference, such as the advancement of large language models (born out of our community) into many application directions. This is not without challenges, partly due to unsolved problems, and partly due to misuse of the technology. All this is reflected well at ACL 2026. The conference also reflects that the field is not only about advancing technology, but also the science of fundamentally understanding language and how it is use by many different speaker communities. <br>
[Full Message from General Chair Philipp Koehn](https://drive.google.com/file/d/1QiJFOmMjvDgsDo3i1XzlQnImxe3RU_Cy/view?usp=sharing) <br><br><br>
<html><button style="display: inline-flex; align-items: center; justify-content: center; white-space: nowrap; border-radius: 9999px; font-weight: bold; background: #7c3aed; color: white; font-family: 'Space Grotesk', sans-serif; height: 40px; font-size: 16px; padding: 0 20px; border: none; cursor: pointer" onclick="window.open('https://drive.google.com/file/d/1QiJFOmMjvDgsDo3i1XzlQnImxe3RU_Cy/view?usp=sharing')">Conference Handbook</button></html>
<br><br><br><br>

To access this event page, you need to log in with the **email address you registered with**. <br>Access credentials will be sent to your email from Underline -  subject line "Welcome to ACL 2026". Please be sure to check your spam email folder if you do not see an email confirmation right away.

To access this event page, you are required to register.
Please complete your registration to continue.

We recommend reading [**the registration information**](https://2026.aclweb.org/registration/) first.

**Online Registration Form**: https://acl.swoogo.com/acl2026

Registration Required

ACL 2026

The 64th Annual Meeting of the Association for Computational Linguistics (ACL 2026) will take place in San Diego, California, from July 2 - 7, 2026.

poster

The 45th Annual Meeting of the Cognitive Science Society will take place in Sydney, Australia as a modified hybrid conference. Presentations from the main-session room will be live-streamed every day for all events. This will include all invited talks and award sessions. Recordings of all CogSci 2023 sessions will be made available for on-demand viewing after the conference.<br>
The in-person component will be held at the International Convention Centre Sydney at Darling Harbour. The Conference will take place Wednesday July 26th – Saturday July 29th, 2023 (Wednesday, July 26th is the Workshops Day).

Join us!

<br>[**The Program at a Glance can be found here**](https://cognitivesciencesociety.org/wp-content/uploads/2023/07/COgSci-PAAG-V2.pdf)
<br>[**The Detailed Program can be found here**](https://online.fliphtml5.com/ebtyf/mswz/)
<br><br>

#### **CogSci 2023 Program Book**
<iframe style="width:700px;height:400px" src="https://online.fliphtml5.com/ebtyf/tzgk/"  seamless="seamless" scrolling="no" frameborder="0" allowtransparency="true" allowfullscreen="true" ></iframe>
<br>

The 45th Annual Meeting of the Cognitive Science Society presents the latest research across cognitive science and highlights the theme of Cognition in Context.

While human communication occurs across groups of varying sizes, current cognitive models typically focus on the dynamics between two individuals. However, generalizing to group communication is not trivial as it requires a joint inference over what a signal means and who it is intended for. Prevailing research relies on agents that recursively model each other to infer a signal's meaning which becomes incredibly complex in group settings. Here we demonstrate how a shared agency approach to communication called the Imagined We (IW) can naturally be leveraged to address challenges of group communication in simulation. In a cooperative communication task, we show that IW agents perform better and that their strategies for communicating under ambiguity are robust to the group size, unlike a baseline comparison model based on recursion. These results offer insight into how the cooperative logic of shared agency can constrain the uncertainty introduced by communicating in group settings.

Simplifying Group Communication: A Shared Agency Modeling Approach

Moral injury describes the set of psychological symptoms resulting from traumatic experiences that violate one’s moral presuppositions. Such disruption occurs when an individual encounters information from the environment that cannot be reconciled with the fundamental assumptions underlying their predictive models of the world. Examination of predictive models has been rapidly developing within cognitive science, with the predictive processing framework emerging as a central paradigm. Predictive processing entails estimations of sensory uncertainty scaffolded by previous predictions and modified by attention. This model describes cognition as seeking to minimize sensory prediction error using dynamic interactions between top-down and bottom-up processes. Therefore, the predictive processing framework may be fruitfully used to examine psychological changes related to moral injury. Towards this end, we will consider moral injury as a form of belief updating, dysregulation in precision estimates of predictive models, and a breakdown in what Ramstead et al. (2016) call ‘regimes of shared attention.

Examining Moral Injury using a Predictive Processing Framework

This paper introduces a philosophy of cognition inspired by recent developments in computational neuroscience relative to the hippocampal formation showing us how thinking, remembering, and navigating are nested measurements of a common process. Reframing cognition as 'all a body does to make its way through its encounter' allows us to understand mind and life as continuous and alleviates current obstacles to progress such as those arising from the so-called mind-body problem and the existence and definition of mental representations. Through hippocampal models, waymaking frames a philosophy whereby pathfinding, memory-development, and knowledge-acquisition are flexibly algorithmic measurements of the same process from sensory and developmental positions of statistical regularity—in other words, where patterns of bodily activity scale fractally once sensory regularities are statistically accounted for at all experimental levels. Taking this seriously, however, means statistically measuring the stimulus as much as the response in future neuroscientific experiments.

Waymaking: a nested approach to cognition inspired by cognitive and computational hippocampal models

In this study, we investigate how human choosers adapt their value encoding strategy to the statistics of the choice environment. Specifically, we ask whether human value encoding mechanism exhibits divisive normalization only in Pareto distributed environments in which it is information-maximizing. To test this theory, we conduct a risky choice experiment in which subjects are presented with blocks of choice stimuli drawn from either a Pareto-distributed environment or a uniform-distributed environment. Our results show that subjects exhibit some degree of normalization regardless of whether it is efficient or not, but do adapt the curvature of the encoding function to the environment. These findings suggest that human value coding mechanisms are flexible but biologically constrained to be perfectly efficient only in specific environments. This study provides new insights into the neural mechanism of human decision-making and the role of environmental statistics in shaping it.

Choosers Adapt Value Coding to the Environment, But Do Not Attain Efficiency

In this paper we present a new modelling framework for concepts based on quantum theory, and demonstrate how the conceptual representations can be learned from data. Our approach builds upon Gardenfors’ classical framework of conceptual spaces, in which cognition is modelled geometrically through the use of convex spaces, which in turn factorise in terms of simpler spaces called domains. We show how concepts from the domains of shape, colour, size and position can be learned from images of simple shapes, where individual images are represented as quantum states and concepts as quantum effects. Concepts are learned by a hybrid classical-quantum network trained to perform concept classification. We also use discarding to produce mixed effects, which can then be used to learn concepts which only apply to a subset of the domains, and show how entanglement (together with discarding) can be used to capture interesting correlations across domains.

A Quantum Model of Concepts

A key aspect of understanding how children learn the meanings of words involves understanding how they mine different sources of information (e.g., observational, linguistic) in the service of learning. According to one dominant view, there exists a class of words (i.e., “hard words”; Gleitman et al., 2005) for which learning their meaning requires access to information beyond the observational contexts in which those words occur. By making alterations to the Human Simulation Paradigm, a paradigm commonly used for investigating vocabulary learning, this study revisits the role of observational contexts for the acquisition of one class of hard words: nouns that denote non-basic level object categories (or “hard nouns”; see Kako, 2005). These data reveal that although observational contexts may not be sufficient to yield learning of precise hard noun meanings, they allow learners to extract systematic partial knowledge, knowledge that may lay a critical foundation for meaning acquisition.

Laying the Foundation: Extracting Partial Meanings of Hard Nouns via Observational Contexts

Our knowledge about the world is populated with categories like dog, chair, and cup. Yet much of what we understand about how we acquire this knowledge comes from studies of learning in circumstances that little resemble real-world experience. In the lab, category learning typically involves pursuing an explicit goal to learn categories that prompts a search for just one or a few features diagnostic of category membership. In contrast, everyday experience is full of incidental encounters that allow us to observe how features cluster together in categories, such observing the co-occurrence of four legs, tail, and snout in dogs we happen to pass on the street. Here, we investigated how incidental exposure shapes category learning using a combined behavioral, eye tracking, and computational modeling approach. We found that learners picked up on the way features clustered together in categories just from incidental exposure, with pronounced downstream consequences for category learning.

Without even Trying: How Incidental Exposure Shapes Category Learning

The category variability effect describes assigning objects to a high-variability category rather than to a low-variability category. We show that similarity-based categorization theories can predict the category variability effect by adapting the similarity computation and conduct a rigorous empirical test. In an optimized categorization experiment, participants learned to assign geometrical figures to a high-variability and a low-variability category and then categorized previously unseen transfer stimuli located between the categories. We compared a formal model that ignores the category variability (Euclidean model) to one that considers the category variability (Mahalanobis model) during similarity computation. The data (N = 43) revealed that most participants did not show the category variability effect, in line with the Euclidean model. Nevertheless, the Mahalanobis model, which considers category variability, consistently described some participants. This demonstrates that—contrary to previous claims—similarity can account for the category variability effect; however, most people do not show the effect.

Modeling the Category Variability Effect in an Exemplar-Similarity Framework

Prior knowledge has long been known to influence retention of newly learned information. In particular, known semantic associations across items facilitates subsequent retrieval for these items, and this effect has been shown to increase with measures of semantic relatedness. In the field of concepts and categories, the processing of taxonomic (e.g., cup-fork, dog-bird) versus thematic (e.g., cup-drink, dog-leash) conceptual relations has been a topic of extensive investigation. Recent work indicates that these two types of conceptual relations can be dissociated at the behavioral and neural levels. However, while semantic relatedness has been explored more broadly, the precise effects of these distinct conceptual relations on associative memory remain unresolved. In this study, we demonstrate that these relations might provide a principled approach to understanding the influence of prior semantic knowledge on memory. We used a stimulus set which consists of word pairs that are either taxonomically or thematically related, while controlling a variety of lexical features relevant to memory performance (e.g., written frequency, number of syllables, concreteness, etc.). We additionally generated semantically unrelated word pairs (e.g., cookie-slate, wool-basement) using word-embeddings from a pre-trained autoencoder model. Participants (n=79) were sequentially shown word pairs (taxonomic, thematic, and unrelated) and subsequently given a cued memory test, in which they were prompted to type the word previously paired with a probe word. We found that word pairs with thematic relations led to better memory performance and shorter reaction times, followed by taxonomic relations, and finally unrelated pairs. These findings may indicate that thematic relations are better remembered due to their known functional or causal properties. This study showed that different conceptual relations differ in the extent to which they facilitate memory performance, providing new insights regarding the influence of prior semantic knowledge on memory.


Thematic relations outperform taxonomic relations in a cued recall task

Key questions in the study of categorization are how individuals form categories from experience and extend that knowledge to assess membership of novel examples. Popular accounts predict generalization to be based on either similarity to reference points or the application of rules or bounds. However, recent data show that some categorization behavior defies the predictions of leading accounts. Expanding on these findings, in the present experiment participants learned a one or two dimensional alternating category structure and were then tested on near and far transfer tasks. Findings reveal that individuals can extend a learned alternating category structure across multidimensional spaces and increasingly distant generalization regions. A new far transfer task suggests that alternation is learned in the form of a relational category at the level of the domain space within which subjects have learned an attribute-based categorization scheme operating at the item level.

Premium content

Downloads

Next from CogSci 2023

Simplifying Group Communication: A Shared Agency Modeling Approach

Similar lecture

Success and Cost Elicit Convention Formation for Efficient Communication