Canada

Neural networks struggle in continual learning settings from catastrophic forgetting: when trials are blocked, new learning can overwrite the learning from previous blocks. Humans learn effectively in these settings, in some cases even showing an advantage of blocking, suggesting the brain contains mechanisms to overcome this problem. Here, we build on previous work and show that neural networks equipped with a mechanism for cognitive control do not exhibit catastrophic forgetting when trials are blocked. We further show an advantage of blocking over interleaving when there is a bias for active maintenance in the control signal, implying a tradeoff between maintenance and the strength of control. Analyses of map-like representations learned by the networks provided additional insights into these mechanisms. Our work highlights the potential of cognitive control to aid continual learning in neural networks, and offers an explanation for the advantage of blocking that has been observed in humans.

CogSci 2022

A Neural Network Model of Continual Learning with Cognitive Control

# Welcome everyone to ACL 2022!

The 60th Annual Meeting of the Association for Computational Linguistics is taking place May 22-27, 2022 as a hybrid event, in Dublin and online. We are happy to welcome all of you to this anniversary edition with an almost 50-50 in-person and virtual participation. 
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, three amazing keynote events and two new initiatives of invited talks: Spotlight Talks for Young Rising Stars (STIRS) and The Next Big Idea Talks. Posters (including Findings of ACL 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 will have an oral session and a poster session as part of Poster Session 1. The program also features eight Tutorials and 28 Workshops. 

 
We wish you a wonderful conference! 
[**The ACL 2022 Organizing Committee**](https://www.2022.aclweb.org/organisers)
 
[**Conference Handbook**](https://drive.google.com/file/d/1_BUCMfhMVrjG9E2e71aHdHeE28KSje0l/view?usp=sharing) 
[**Mini Handbook**](https://drive.google.com/file/d/1qlBKl0wzmlVF1oCeMQl3BahLd9nLP5Ce/view?usp=sharing) 
[**Posters and Demo guides**](https://drive.google.com/file/d/1UucMAoCNncIOaH1rMMDa0owuG9qgvJTG/view?usp=sharing)

ACL 2022

The Association for Computational Linguistics (ACL) is the premier international scientific and professional society for people working on computational problems involving human language, a field often referred to as either computational linguistics or natural language processing (NLP). 

technical paper

Join us for CogSci2022! Our 44th Annual Conference will be centered around the theme of diversity, with an accompanying range of keynotes and invited speakers who themselves highlight the diversity of our field. Our three symposia are Cognition across Diverse Populations, Cognition within Diverse Societies, and Global Challenges for Cognitive Science. The conference is being offered in a fully hybrid environment for onsite and online attendees.
 
 [**The Program at a Glance can be found here**](https://cognitivesciencesociety.org/wp-content/uploads/2022/07/Cogsci-2022-PAAG-Detailed-July-12final-1.pdf)
 [**The Detailed Program can be found here**](https://cognitivesciencesociety.org/wp-content/uploads/2022/08/DeArmond-CogSci2022-program-v113-final.pdf)

This conference requires a registration to attend. Please use the link below to secure your ticket. The website will be available to all registered attendees on July 25th. 

[Register to Attend](https://cognitivesciencesociety.org/registration/)

Registration Is Required

The 44th Annual Meeting of the Cognitive Science Society highlighted research on the theme Cognitive Diversity in addition to the full breadth and diversity of research topics offered by the society’s membership. 





Recent work has shown that the brain organizes abstract, non-spatial relationships between entities into map-like representations. However, an animal’s objectives often depend on only a subset of the features of the environment. Under these circumstances, cognitive control – the capacity to flexibly select the features most relevant in the current context – becomes paramount. Here, we explore the relationship between cognitive control and the geometry of map-like representations by combining fMRI with neural network modeling. We find that brain areas including hippocampus and entorhinal cortex spontaneously organize pairwise relationships into 2D map-like representations, and that this 2D structure was controlled by compressing task-irrelevant dimensions in areas of prefrontal and parietal cortex. Our neural network model reproduced these findings and additionally predicted warping in the geometry along a context-invariant axis. This prediction was confirmed with fMRI, which showed that the degree of warping was correlated with individual differences in cognitive control.

The Geometry of Map-Like Representations under Dynamic Cognitive Control

Distributed vector representations are a key bridging point between connectionist and symbolic representations of cognition. It is unclear how uncertainty should be modelled in systems using such representations. One may place vector-valued distributions over vector representations, although that may assign non-zero probabilities to vector symbols that cannot occur. In this paper we discuss how bundles of symbols in Vector Symbolic Architectures (VSAs) can be understood as defining an object that has a relationship to a probability distribution, and how statements in VSAs can be understood as being analogous to probabilistic statements. We sketch novel designs for networks that compute entropy and mutual information. In this paper we restrict ourselves to operators proposed for Holographic Reduced Representations, and representing real-valued data. However, we suggest that the methods presented in this paper should translate to any VSA where the dot product between fractionally bound symbols induces a valid kernel.

Fractional Binding in Vector Symbolic Architectures as Quasi-Probability Statements

Models in cognitive science are often restricted for the sake of interpretability, and as a result may miss patterns in the data that are instead classified as noise. In contrast, deep neural networks can detect almost any pattern given sufficient data, but have only recently been applied to large-scale data sets and tasks for which there already exist process-level models to compare against. Here, we train deep neural networks to predict human play in 4-in-a-row, a combinatorial game of intermediate complexity, using a data set of 10,874,547 games. We compare these networks to a planning model based on a heuristic function and tree search, and make suggestions for model improvements based on this analysis. This work provides the foundation for estimating a noise ceiling on massive data sets as well as systematically investigating the processes underlying human sequential decision-making.

Improving a model of human planning via large-scale data and deep neural networks

“Slavery ended in the United States because slavery is morally wrong.” This explanation does not seem to fit the typical criteria for explaining an event, since it appeals to ethics rather than causal factors as the reason for this social change. But do people perceive these ethical claims as explanatory, and if so, why? In Study 1, we find that people accept ethical explanations for social change and that this is predicted by their meta-ethical beliefs in moral progress and moral objectivism, suggesting that they treat morality somewhat akin to a causal force. In Study 2, we find that people recognize this relationship between ethical explanations and meta-ethical commitments, using the former to make inferences about individuals’ beliefs in moral progress and objectivism. Together these studies demonstrate that our moral commitments shape our judgments of explanations and that explanations shape our moral inferences about others.

Ethical Explanations

To defend or burnish our moral reputation, we often tell moral narratives. Moral narratives describe morally relevant actions and explanations of those actions, detailing how people acted and why they did so. When and why does communication about moral events include descriptions of peoples’ actions and explanations? We hypothesize that informational, reputational, and presentational goals of narrators shape whether their communication contains clear actions and explanations. We asked a group of “narrators” to communicate with other people following a moral decision. Another group of “audience” members judged them based on their chosen statement. We find that the informational and reputational goals of narrators can explain what information they decide to reveal. Narrators choose what to say based on inferring the audience’s likely perceptions but underestimate how much audiences in fact expect answers to “what” and “why”. Audiences, however, do not always perceive the lack of expected information as indicative of deceit.

Narrating the "what" and "why" of our moral actions

One of the most influential modern theories of morality, Moral Foundations Theory, proposes that morality is formed on innate and shared modular foundations. Psychologists have studied the conceptual development of these moral foundations in childhood, but there exists no comprehensive effort on characterizing the early emergence of moral foundations in naturalistic settings. We explore the emerging order of moral foundations through child and caretaker speech. Using computational methods, we contribute an annotated dataset of moral utterances and find that the individualizing foundations emerge earlier than the binding foundations. Furthermore, caretakers tend to talk more about fairness and degradation, while children talk more about cheating. These results are robust across child gender, family's social class, and race.

The emergence of moral foundations in child language development

Rules undoubtedly guide our moral lives. Simple moral rules prohibit lying, cheating, and stealing, for instance. But the moral mind is more flexible than a theory based only on rule-adherence can account for. In this paper, we look at one particular kind of flexibility: the ability to figure out when it is okay to break a moral rule. We elicit judgments of the moral acceptability of breaking a simple rule: it's wrong to cut in line. We created a video game environment in which agents attempt to gather water -- and sometimes must stand in line behind others to do so. Subjects watch clips of the game being played and make judgments of the moral acceptability of cutting in line across a wide range of spatio-temporally varied and dynamic scenarios. Our data suggests that subjects make judgments by using a generative understanding of the underlying function of the rule about waiting in line. We further show that our data cannot be accounted for by either 1) simple rule adherence or 2) utility maximization.

Flexibility in Moral Cognition: When is it okay to break the rules?

Beyond seemingly lower-level features such as color and motion, visual perception also recovers properties that are more commonly associated with higher-level thought — as when an upwardly accelerating object is seen as self-propelled, and resisting the force of gravity. Past work has explored how speed changes drive the perception of physical forces, but might the reverse also be true? Does seeing a speed change as resisting the force of gravity make us more likely to notice it in the first place? In four experiments, observers were more sensitive to objects’ accelerations when they moved upward (i.e. when those accelerations were opposite to gravity), and they were more sensitive to objects’ decelerations when they moved downward (i.e. when those decelerations appeared as ‘braking’ against gravity). We conclude that the perception of physical forces is not merely an outcome of visual processing, but also determines the perception of other, seemingly lower-level, features of how objects move.

May the Force be against you: Better sensitivity to speed changes opposite to gravity

According to the embodied cognition view, retrieving the meaning of action-related language requires the participation of sensorimotor processes. In consequence, an increasing number of neurostimulation (TMS and tDCS) studies have tried to test this idea. In the present study, we aim to evaluate the evidential value of this body of research (N = 43) by means of p-curve analyses. Our results suggest that the published studies so far do not yet allow to establish if they explore real effects beyond a reasonable doubt. We also found that these studies are quite underpowered (estimated power < 30%), which suggests that a large percentage of these findings are, in fact, false-positive results. In sum, our study suggests that the results derived from brain stimulation studies of embodied semantics are not as reliable as would be desirable. We give some recommendations that will be important for future research on this topic.

Grounding meaning in the motor system: A p-curve analysis of the TMS and tDCS evidence

Using space to think about time appears to be a human universal, but the specifics of people’s space-time mappings vary across individuals and groups. What determines how people spatialize time in their minds? Here we hypothesized that motor experience contributes to individuals’ space-time mappings, independent of linguistic and cultural factors. Many everyday motor actions require using the hands in a particular sequence (e.g., positioning a nail with the nondominant hand before swinging the hammer with the dominant hand), establishing a correlation between space (left hand, right hand) and time (earlier action, later action) that reverses with handedness. These action sequences reinforce Westerners’ typical left-to-right mental timeline for right-handers, but contradict this timeline for left-handers. Accordingly, we find that right-handers associate leftward space with earlier times and rightward space with later times more strongly than left-handers, for whom cultural and bodily experiences present contrasting relationships between space and time.

Premium content

Downloads

Next from CogSci 2022

The Geometry of Map-Like Representations under Dynamic Cognitive Control

Similar lecture

Overcoming Catastrophic Forgetting beyond Continual Learning: Balanced Training for Neural Machine Translation