United States

We investigated how the language background (L1) of bilinguals influences the representation and use of the second language (L2) through computational models. With the essays part from The International Corpus Network of Asian Learners of English (ICNALE), we compared variables indicating syntactic complexity in their L2 production to predict L1. We then trained neural language models based on BERT to predict the L1 of these English learners. Results showed the systematic influence of L1 syntax properties on English learners&#39; L2 production, which further confirmed integrations of syntactic knowledge across languages in bilingual speakers. Results also found neural models can learn to represent and detect such L1 impacts, while multilingually trained models have no advantage in doing so.

CogSci 2021

Language representations in L2 learners: Toward neural models

EMNLP 2021 is planned to be a hybrid event in Punta Cana, Dominican Republic, with both on-site and fully virtual participation possible. The experience for on-site participants would closely approximate a normal pre-COVID *ACL conference, with 5-6 thematically organized parallel sessions and live Q/A and interactive discussion immediately after the talks. Presentations by virtual participants will be equitably interleaved with those of on-site participants, projected on the auditorium screens as if on-site, and also followed immediately by live Q/A and interactive discussion at a time during reasonable waking hours for the virtual presenter. For all participants, on-site and virtual, who are unable to attend a session due to either time-zone issues or because they are participating in another session live, talk recordings and slides will be available online at a minimum after the live presentation (and in many cases before as well), and questions may be submitted in advance on session-specific discussion boards and answered live in session with the usual visual aids if desired.

<iframe style="width:700px;height:400px" src="https://online.fliphtml5.com/ebtyf/ceby/"  seamless="seamless" scrolling="no" frameborder="0" allowtransparency="true" allowfullscreen="true" ></iframe>

Please Note: The EMNLP registration system is not currently connected to the underline site as we are still in the process of building out EMNLP 2021. You will receive access instructions from underline the week of November 1st. 

Access is given only to EMNLP upon registration, if you have not registered please do so [here](https://2021.emnlp.org/registration).

Registered attendees will receive access the week of November 1st.

EMNLP 2021

EMNLP 2021 is planned to be a hybrid event in Punta Cana, Dominican Republic, with both on-site and fully virtual participation possible.

We investigated how the language background (L1) of bilinguals influences the representation and use of the second language (L2) through computational models. With the essays part from The International Corpus Network of Asian Learners of English (ICNALE), we compared variables indicating syntactic complexity in their L2 production to predict L1. We then trained neural language models based on BERT to predict the L1 of these English learners. Results showed the systematic influence of L1 syntax properties on English learners' L2 production, which further confirmed integrations of syntactic knowledge across languages in bilingual speakers. Results also found neural models can learn to represent and detect such L1 impacts, while multilingually trained models have no advantage in doing so.

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.

We use forecasting models for the 2020 US presidential election to test a model of human causal judgment. Across tens of thousands of simulations of possible outcomes of the election, we computed, for each US state, an adjusted measure of the correlation between a democratic victory in that state and a democratic victory at the national level. These scores accurately predicted the extent to which US participants (N=207, pre-registered) viewed victory in a given state as having caused Joe Biden to win the presidency. This supports the theory that people intuitively select as causes of an outcome the factors with the largest average causal effect on that outcome across possible worlds. This is the first evidence that the theory scales to real-world complex settings, and suggests a deep connection between cognitive processes for prediction and causal judgment.

Causal judgment in the wild

As the mental health crisis deepens with the prolonged COVID-19 pandemic, there is an increasing need for understanding individuals’ emotional experiences. We have built a large-scale Korean text corpus with five self-labeled psychological ground-truths: empathy, loneliness, stress, personality, and emotions. We collected 19,025 documents of daily emotional experiences from 3,805 Korean residents from October to December 2020. We collected 42,128 sentences with different levels of theory-of-mind. Each sentence was annotated by trained psychology students and reviewed by experts. Participants varied in their ages from the early 20s to late 80s and had various social and economic statuses. The pandemic impacted the majority of daily lives, and participants often reported negative emotional experiences. We found the most frequent topics: responses to confirmed cases, health concerns of family members, anger towards people without masks, stress-relief strategies, change of the lifestyle, and preventive practices. We then trained the Word2Vec model to observe specific words that match each topic from the topic model. The current dataset will serve as benchmark data for large-scale and computational methods for identifying mental health levels based on text. This dataset is expected to be used and transformed in many creative ways to mitigate COVID-19-related mental health problems

Building a Psychological Ground Truth Dataset with Empathy and Theory-of-Mind During the COVID-19 Pandemic

Egocentric space perception is multimodal, closely tied to action and bodily movements and has an inherent phenomenal dimension. One prominent account, provided by Rick Grush, has postulated posterior parietal cortex as key neural area. Computational model based on Kalman filter has been proposed to account for the operation of this brain region, underscoring the importance of bodily skills for perceiving spatial properties.
The current study provides a first direct simulation of this model in a semi-realistic 3D virtual environment. The goal of the simulation was to develop an agent with a realistic ability for egocentric space perception based on a neural approximation of Kalman filter. To achieve this goal, we use machine learning techniques, with a strong focus on unsupervised methods of reinforcement learning.
Resulting agent is tested behaviorally on ecologically plausible tasks to evaluate its internal, learned representations. Poster presents simulation results and discusses the model.

Modeling "spatial purport of perceptual experience": egocentric space perception in a semi-realistic 3D virtual environment

When exposed to perceptual sequences, we are able to gradually identify patterns within and form a compact internal description of the sequence. One proposal of how disparate sequential items can become one is people's ability to form chunks. We study chunking under the regime of serial reaction time tasks. We propose a rational model of chunking that progressively rearranges and modifies its representation to arrive at one that is beneficial to participants' utility under task demands. Our model predicts that participants should, on average, learn longer chunks when optimizing for speed than optimizing for accuracy. We tested this prediction experimentally by instructing and rewarding one group of participants to act as fast as possible, while the other group was instructed to act as accurately as possible. From several independent sources of evidence, we confirmed our model's predictions that participants in the fast condition chunked more than participants in the accurate condition. These results shed new light on the benefits of chunking and pave the way for future studies on structural and representation learning domains.

Chunking as a Rational Solution to the Speed-Accuracy Trade-off in a Serial Reaction Time Task

Social interactions are composed of coordinated, multimodal behaviors with each individual taking turns and sharing attention. By the second year of life, infants are able to engage in coordinated interactions with their caregivers. Although research has focused on the social behaviors that enable parent-infant dyads to engage in joint attention, little work has been done to understand the sensorimotor mechanisms underlying coordination. Using wireless head-mounted eye trackers and motion sensing, we recorded 31 dyads as they played freely in a home-like laboratory. We identified moments of visual joint attention, when parent and infant were looking at the same object, and then measured the dyad’s head and hand movements during and around joint attention. We found evidence that both parents and infants still their bodies during joint attention. We also compared instances of joint attention that were led by the parent or by the infant and identified different sensorimotor pathways that support the two types of joint attention. These results provide the foundation for continued exploration of the critical role of sensorimotor processes on coordinated social behavior and its development.

The Sensorimotor Dynamics of Joint Attention

Despite the rise in communicative technologies for healthy aging, little research has focused on how effectively older adults process language spoken by artificial agents. We explore whether a robot's redundant (but potentially helpful) descriptions facilitate real-time comprehension in younger and older listeners. Gaze was recorded as participants heard instructions like "Tap on the [purple/closed] umbrella" for a display containing eight unique objects. We manipulated the description (no-adjective, color-adjective, state-adjective) and the visual context, specifically whether there was another object bearing the property denoted by the adjective (purple/closed notebook). Relative to the no-adjective condition, redundant color adjectives speeded comprehension when they uniquely identified targets, whereas (less-salient) state adjectives always impeded comprehension. No age-related differences were observed. Paralleling human-human studies, language processing in human-robot communication is facilitated when salient information narrows visual search. Together, these findings help inform the future design of communicative technologies.

Aging and Social Robots: How Overspecification Affects Real-Time Language Processing

Foraging has been suggested to provide a more ecologically-valid context for studying decision-making. However, the environments used in human foraging tasks fail to capture the structure of real world environments which contain multiple levels of spatio-temporal regularities. We ask if foragers detect these statistical regularities and use them to construct a model of the environment that guides their patch-leaving decisions. We propose a model of how foragers might accomplish this, and test its predictions in a foraging task with a structured environment that includes patches of varying quality and predictable transitions. Here, we show that human foraging decisions reflect ongoing, statistically-optimal structure learning. Participants modulated decisions based on the current and potential future context. From model fits to behavior, we can identify an individual's structure learning ability and relate it to decision strategy. These findings demonstrate the utility of leveraging model-based reinforcement learning to understand foraging behavior.

Model-based foraging using latent-cause inference

Is analogical reasoning a task that must be learned to solve from scratch by applying deep learning models to massive numbers of reasoning problems? Or are analogies solved by computing similarities between structured representations of analogs? We address this question by comparing human performance on visual analogies created using images of familiar three-dimensional objects (cars and their subregions) with the performance of alternative computational models. Human reasoners achieved above-chance accuracy for all problem types, but made more errors in several conditions (e.g., when relevant subregions were occluded). We compared human performance to that of two recent deep learning models (Siamese Network and Relation Network) directly trained to solve these analogy problems, as well as to that of a compositional model that assesses relational similarity between part-based representations. The compositional model based on part representations, but not the deep learning models, generated qualitative performance similar to that of human reasoners.

Visual Analogy: Deep Learning Versus Compositional Models

People learn new categories on a daily basis, and the study of category learning is a major topic of research in cognitive science. However, most prior work has focused on how people learn categories over abstracted, artificial (and usually perceptual) representations. Little is known about how new categories are learnt for natural objects, for which people have extensive prior knowledge. We examine this question in three pre-registered studies involving the learning of new categories for everyday foods. Our models use word vectors derived from large-scale natural language data to proxy mental representations for foods, and apply classical models of categorization over these vectorized representations to predict participant categorization judgments. This approach achieves high predictive accuracy rates, and can be used to identify the real-world settings in which category learning is impaired. In doing so, it shows how existing theories of categorization can be used to predict and improve everyday cognition and behavior.

Learning New Categories for Natural Objects

Multidimensional scaling (MDS) has provided insight into the structure of human perception and conceptual knowledge. However, MDS usually requires participants to produce large numbers of similarity judgments, leading to prohibitively long experiments for most developmental research. Here we propose a method that combines simple grouping tasks with recent neural network models to uncover participants’ psychological spaces. We validate the method on simulated data and find that it can uncover the true structure even when given heterogeneous groupings. We then apply the method to data from the World Color Survey and find that it can uncover language-specific color organization. Finally, we apply the method to a novel developmental experiment and find age-dependent differences in conceptual spaces.

Downloads

Next from CogSci 2021

Causal judgment in the wild

Similar lecture

Are BERTs Sensitive to Native Interference in L2 Production?