United States

Domain-general thought requires an ability to combine information from different cognitive domains within a single judgement. We presented kea parrots (Nestor notabilis) with probabilistic choice tasks that required them to make predictions about which of two hidden samples was most likely to contain a rewarding token. Over the course of three experiments, we found that kea used relative rather than absolute quantities to make their sampling predictions and could integrate either knowledge about a physical barrier or demonstrators’ sampling biases to adjust their judgements. This work provides the first evidence for domain-general statistical inference outside of humans and the great apes. This suggests that at least two structurally distinct brain models have independently evolved an ability to integrate information, highlighting how comparative cognition may inspire the development of novel artificial general intelligence systems.

CogSci 2021

Kea show three signatures of domain-general inference

## Welcome!
"I am excited to welcome you to this year’s edition of the Conference on Empirical Methods in Natural Language Processing! Importantly, it marks the 30th edition of EMNLP. With over 8,000 submissions, more than 3,000 accepted papers, and thousands of attendees, we have come a long way from that first
workshop, which had 14 accepted papers. As the field looks ahead, Suzhou is the fitting location for celebrating this milestone: rooted in a long literary tradition, yet modern and forward-looking, and home to a large share of the NLP community."<br>

*Message from the General Chair, Dirk Hovy*





<br>

EMNLP 2025

Celebrate 30 Years of EMNLP! 
EMNLP 2025 will be held in Suzhou, China from November 5th to November 9th, 2025.

poster

### Looking for Cogsci 2023? Please [click here](https://underline.io/events/405/reception) to switch to 2023 edition.


**Dear Colleagues,**

Welcome to the 43rd Annual Conference of the Cognitive Science Society! 

This year’s conference brings together the latest research from all Cognitive Science and also highlights this year’s theme **Comparative Cognition: Animal Minds.** 

The program features four plenary speakers: Friederike Range (University of Veterinary Medicine, Vienna), Onur Güntürkün (Ruhr University Bochum), Julia Fischer (University of Göttingen), and Erich Jarvis (The Rockefeller University). The keynotes and invited symposia showcase the broad spectrum of ideas that encompass the conference theme of comparative cognition. 

Further, the program includes the Rumelhart Prize keynote presentation by Susan Goldin-Meadow (University of Chicago), and symposia to celebrate the winner of the Elman Prize: Wei Ji Ma (New York University), the Heineken Award winner Robert Zatorre (McGill University), and the Robert J. Glushko Dissertation Prizes session.

The program committee for CogSci 2021 received 1,025 submissions, including 710 full papers, 275 member abstracts, 8 publication-based talks, as well as 15 proposals for symposia, and 17 for workshops and tutorials. After a rigorous review process, the committee selected 152 papers for oral presentation and inclusion in the conference proceedings (14.8%), 332 papers for poster presentation and inclusion in the proceedings (32.4%), and 175 papers for poster presentation with inclusion of abstracts in the proceedings (17.1%) We also selected 275 submitted member abstracts. In addition, we accepted 6 publication-based talks, 12 symposia, and 9 workshops and tutorials to make for an exciting and inclusive program.

We hope that you enjoy the conference this year!<br><br>

**Your Program Chairs,**
![](https://assets.underline.io/uploads/markdown_image/1/image/99984908caba5e6453e5fc5ba34f3507.jpg)

**Learn more about the Vienna Cognitive Science Hub at the University of Vienna, hosts of CogSci 2021**

<iframe width="560" height="315" src="https://www.youtube.com/embed/464h1m3iU68" title="YouTube video player" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture" allowfullscreen></iframe>

To review the proceedings for the conference or review the searchable program, please visit the [website](https://cognitivesciencesociety.org/program/).


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.

The study of emergent languages in deep multi-agent simulations
has become an important research field. While targeting
different objectives, most studies focus on analyzing properties
of the emergent language—often in relation to the agents’
inputs—ignoring the influence of the agents’ perceptual processes.
In this work, we use communication games to investigate
how differences in perception affect emergent language.
Using a conventional setup, we train two deep reinforcement
learning agents, a sender and a receiver, on a reference game.
However, we systematically manipulate the agents’ perception
by enforcing similar representations for objects with specific
shared features. We find that perceptual biases of both sender
and receiver influence which object features the agents’ messages
are grounded in. When uniformly enforcing the similarity
of all features that are relevant for the reference game,
agents perform better and the emergent protocol better captures
conceptual input properties.

Why and how to study the impact of perception on language emergence in artificial agents

Understanding a sequence of images as a visual narrative is challenging because it requires not only the understanding of what is shown at a particular moment but also what has changed, been omitted or is out of frame. The human cognitive system makes inferences about the state of the world based on transitions between sequential frames. In this paper, we present a principled analysis of the stylistic differences between two dominant styles of multi-modal narratives, western comics and manga. These two styles differ in terms of screening, ballooning, layout, language, and reading order. We first provide a systematic account of these differences based on an annotated dataset consisting of both comics and manga. We then annotate these datasets with a new feature set and evaluate the contributions of these features through development of a computational model of multi-modal comprehension. The model evaluation is presented through the cloze test that measures the accuracy of the model in predicting unseen next frames given the prior frames in a sequence. Our results provide initial benchmarks and insight into the fundamental challenges that the multi-modal narrative understanding task presents for computational models both for language and vision.

A Computational Model of Comprehension in Manga Style Visual Narratives

Successful teaching requires an assumption of how the learner learns - how the learner uses experiences from the world to update their internal states. We investigate what expectations people have about a learner with a behavioral experiment: Hu- man teachers were asked to teach a sequential decision-making task to an artificial dog in an online manner using rewards and punishments. The artificial dogs were implemented with either an Action Signaling agent or a Q-learner with different dis- count factors. Our findings are threefold: First, we used ma- chine teaching to prove that the optimal teaching complexity across all the learners is the same, and thus the differences in human performance was solely due to the discrepancy between human teacher’s theory of mind and the actual student model. Second, we found that Q-learners with small discount factors were easier to teach than action signaling agents, challenging the established conclusion from prior work. Third, we showed that the efficiency of teaching was monotonically increasing as the discount factors decreased, suggesting that humans’ theory of mind bias towards myopic learners.

Using Machine Teaching to Investigate Human Assumptions when Teaching Reinforcement Learners

One of the consequences of the pandemic is that throughout 2020 virtual interactions largely replaced face-to-face interactions. Though there are few studies of how social media impacts body image perception across genders, research suggests that socializing through a virtual self-body image might have distinct implications for men and women. On an online study, we examined whether type of social pressure and body-ideal exert distinct pressures on members of the X, Y, and Z generations. Results showed media pressure affected body image satisfaction significantly more than other kinds of social pressure across genders and generations, with young males reporting a higher impact compared to older males. Males experienced more pressure to be muscular and women to be thin, especially for the younger generation.Future research should focus on social media as a potential intervention tool for the detection and prevention of body image disorders in both young female and male adults.

Body Image During Quarantine; Generational Effects of Social Media Pressure on Body Appearance Perception

There has been a widely held view that visual representations (e.g., photographs and illustrations) do not depict negation, for example, one that can be expressed by a sentence "the train is not coming". This view is empirically challenged by analyzing the real-world visual representations of comic (manga) illustrations. In the experiment using image captioning tasks, we gave people comic illustrations and asked them to explain what they could read from them. The collected data showed that some comic illustrations could depict negation without any aid of sequences (multiple panels) or conventional devices (special symbols). This type of comic illustrations was subjected to further experiments, classifying images into those containing negation and those not containing negation. While this image classification was easy for humans, it was difficult for data-driven machines, i.e., deep learning models (CNN), to achieve the same high performance. Given the findings, we argue that some comic illustrations evoke background knowledge and thus can depict negation with purely visual elements.

Visual representation of negation: Real world data analysis on comic image design

Deep Neural Network representations correlate very well with neural responses measured in primates' brains and with psychological representations of human similarity judgement tasks, making them possible models for human behavior-related tasks. This study investigates whether DNNs can learn an implicit association (between colors and emotions) for images. An experiment was conducted in which subjects were asked to select a color for a given emotion-inducing image. These human responses (decision probabilities) were modeled on neural networks using representations extracted from pre-trained DNNs for the images and colors (a square of the color). The model presented showed a fuzzy linear relationship with the decision probabilities. Finally, this model was presented as a model for emotion classification tasks, specifically with very few training examples, showing an improvement in accuracy from a standard classification model. This analysis can be of relevance to psychologists studying these associations and AI researchers modelling emotional intelligence in machines.

Emotion-Color Association in Biologically Inspired Deep Neural Networks

It is a long standing challenge to devise a formal model of ACT-R as a basis for formal reasoning on ACT-R. The ACT-R architecture is a composition of components (such as modules) with predefined interfaces between components and predefined interactions on the interfaces. Reasoning over the correctness of a formal model of ACT-R benefits from the separation of abstraction levels i.e. reasoning on the level of interfaces and interactions between components in isolation from the concrete behaviour of each component. 
We propose a formal semantics of ACT-R that preserves the structural properties of architectural components, i.e., the interfaces of modules to the remaining architecture as well as communication between modules within the architecture. We demonstrate how our new formal semantics of ACT-R serves to prove the correctness of the timed automaton based operational semantics for ACT-R (TA-ACT-R) on the level of architectural components.

A Formal Operational Model of ACT-R: Structure and Behaviour

Humans are highly efficient learners, with the ability to grasp the meaning of a new concept from just a few examples. Unlike popular computer vision systems, humans can flexibly leverage the compositional structure of the visual world, understanding new concepts as combinations of existing concepts. In the current paper, we study how people learn different types of visual compositions, using abstract visual forms with rich relational structure. We find that people can make meaningful compositional generalizations from just a few examples in a variety of scenarios, and we develop a Bayesian program induction model that provides a close fit to the behavioral data. Unlike past work examining special cases of compositionality, our work shows how a single computational approach can account for many distinct types of compositional generalization.

Flexible Compositional Learning of Structured Visual Concepts

Children learn word meanings by making use of commonalities across the usages of a word in different situations. However, early word learning experiences have a high level of uncertainty. For a word in an utterance, there are many possible meanings in the environment (referential uncertainty). Similarly, for a meaning, there are multiple possible words in the utterance (linguistic uncertainty). We propose a general framework to investigate the role of mutual exclusivity bias (asserting one-to-one mappings between words and their meanings) in early word learning. Through a set of computational studies, we show that to successfully learn word meanings under uncertainty, a model needs to implement two types of competition: words competing for the association to a meaning reduces linguistic uncertainty, and meanings competing for a word limits referential uncertainty. Our work highlights the importance of an algorithmic-level analysis to shed light on different mechanisms that implement a computational-level theory.

Mutual Exclusivity as Competition in Cross-situational Word Learning

The ability to learn temporal relationships and use that knowledge to simulate future events is among the most remarkable aspects of cognition. Recently introduced behavioral task called Judgment of Imminence (JOI) combined with a well-known Judgment of Recency (JOR) task pointed to a remarkable symmetry between the temporal organization of memory and prediction. The data were consistent with the hypothesis that both memory and prediction can be organized as a compressed mental timeline. This means that the past and future can be remembered or simulated sequentially relative to the present. The compression implies that events closer to the present, regardless of whether they are in the past or in the future, were represented more accurately than those further from the present. Here we used the existing JOR model based on a compressed memory timeline to build an associative representation that can learn the temporal relationships and create a timeline of the future, which mirrors the timeline of the past. We show that this approach can simultaneously account for response times and accuracy in both JOR and JOI. This work provides a time-local neural-level mechanistic account for how the temporal organization of the memory can be used to learn the temporal structure of the world and simulate the future in an efficient manner as a compressed mental timeline.

Downloads

Next from CogSci 2021

Why and how to study the impact of perception on language emergence in artificial agents

Similar lecture

No Black Boxes: Interpretable and Interactable Predictive Healthcare with Knowledge-Enhanced Agentic Causal Discovery