United States

Word associations are among the most common paradigms for studying the human mental lexicon. While their structure and underlying relation types have been widely studied, surprisingly little attention has been given to the question of why participants produce the observed associations. Answering this question would not only advance understanding of human cognition, but could also aid machines in learning and representing basic commonsense knowledge. Here, we introduce WAX: a large, crowd-sourced data set of 19K English word associations paired with human-generated free-text explanations and relation type labels. We present an efficient framework for eliciting associations together with explanations, and a comprehensive analysis of the emerging types of explanations and relations. We test language models in their ability to predict associations and generate explanations, demonstrating that models struggle to capture the diversity of human-produced associations, and suggesting WAX as a fertile resource for future research.

CogSci 2023

WAX: A Dataset for Word Association eXplanations

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.

A positive disjunction (A or B) in natural language typically receives two logical interpretations: inclusive disjunction vs. exclusive disjunction. Within the Gricean paradigm, the inclusive interpretation of disjunction is considered its basic meaning while the exclusive interpretation is caused by Gricean pragmatic reasoning, called exclusivity implicatures. However, pragmatic reasoning is not the only source of exclusivity. In this paper we study the role of prior beliefs regarding the compatibility of the two propositions (A, B) in a disjunction (A or B). In two experiments, we tested the impact of prior compatibility on exclusivity implication generation using examples from the literature. Our results show that prior beliefs about compatibility contribute to the overall exclusivity of a disjunction, but that significant exclusivity remains to be explained by other sources. We conclude that most exclusivity implications are composite implications that feed from minimally two sources: prior beliefs about compatibility and Gricean pragmatic reasoning.

Prior Beliefs About the Compatibility of Disjuncts Impact the Exclusivity Implication of Disjunction

The present study investigated the degree to which linguistic alignment is affected by communication medium and conversation type. Taking advantage of the changes instigated by the COVID-19 pandemic where nearly everyone was forced to engage over videoconferencing (VC) platforms or to meet face-to-face (FF) with public health constraints, we asked participants to conduct three conversations in one of three communication media. Here we analyze conversations from 23 dyads: 8 dyads who conversed FF, 8 who conversed in a laboratory VC set-up and 7 who conversed in a remote VC set-up. Every dyad had an affiliative, an argumentative, and a task-based cooperative conversation. Results showed differences in lexical and syntactic alignment between conversation types but not communication media. Interestingly, we also found interaction effects. These results point to changes in alignment based on communication constraints and provide support for the interpersonal synergies approach to conversation.


Language in the Time of COVID: Sensitivity of Linguistic Alignment to Conversation Type and Communication Modality

The purpose of this study is to examine Theory of Mind (ToM) in processing the Chinese implicit causality verbs. The Experiment identified the Chinese causal verbs of high tendency with a bias decision task and used the ToM paradigms: Reading the Mind in the Eyes test revised and YONI task, to explore the relationship between English proficiency, Theory of Mind, and causal verb processing.
The main results are: Proficiency in both English and Chinese may affect one’s interpretation of the Chinese implicit causality/consequentiality, while on the other hand, the affective ability of ToM is enhanced as the English proficiency increases due to the understanding of the implicit causes/effects of the same concept under the context of the different languages. The study infers that the process of Chinese and English are different and Bilingualism can uses two mechanism at the same time compared to 
monolingual.

Implicit Causality in Chinese Verbs: The Correlations between Bilingualism and Theory of Mind

Mechanisms underlying children’s attempts to solve referential ambiguity have been described as a process-of-elimination or a novelty bias. We measured 2.5-year-old children’s pointing and eye movements during referent selection in ostensive and non-ostensive naming contexts. Compared to familiar targets, for novel targets, looking proportions were lower and children took longer to point; and looked significantly more at competitors before focusing on targets, indicating more time spent considering competitors. Ostensive naming also predicted better delayed retention, but retention dropped significantly in the non-ostensive condition. Greater novel target looking during referent selection predicted better immediate retention for the ostensive condition, but better delayed retention for the non-ostensive condition. Thus, in contexts with minimal encoding support, a focus on novel targets strengthens word-object associations over time. For toddlers, referent selection is best explained as both a process-of-elimination and a novelty bias; and the latter supports encoding which facilitates long-term retention of novel words.

Do Toddlers Rely On a Process-of-Elimination or a Novelty Bias in Ambiguous Word Learning Contexts?

We conducted two experiments to investigate what triggers sarcasm between close friends and whether the factors prompting sarcastic comments in production are also shared with an external observer. In Experiment 1, participants freely reacted to different types of situations in written form and rated their perception of the given contexts, the level of sarcasm of their responses, and the intentions behind their responses. Results showed that the intentions to say clever things or to mock the addressee in a hilarious or friendly manner triggered a higher number of sarcastic answers. In contrast, the intentions to be direct or be nice to the addressee triggered less sarcastic answers. In Experiment 2, a new group of participants rated the responses collected in Experiment 1 on the same dimensions. Overall, we observed similar patterns in both experiments. However, the intentions to criticize the addressee softly and to say clever things were stronger predictors of sarcasm for the observers than for the producer of the statement.

Intended and Perceived Sarcasm Between Close Friends: What Triggers Sarcasm and What Gets Conveyed?

The meanings of iconic signs are usually not easily accessible to sign-naive people. However, most previous studies asked participants to guess the meaning of iconic signs in isolation and without any contextual cues. We ask whether signs whose form is based on more cross-linguistically common underlying motivations are easier to interpret than signs based on less common underlying motivations. Since recent research suggests that iconicity is a relationship of resemblance between the signifier and the signified that is instantiated contextually, we also provide participants with a contextual cue (in the form of a word). We find that interpretability of iconic signs does correlate with cross-linguistic frequency of the underlying motivation of the sign.

Transparency in Sign Language Animal Signs

How do people prepare for an upcoming distracting stimulus? It has been recently suggested that not only observers do not inhibit distractors before their appearance, but they are rather more alert at those moments. Interestingly, a similar effect was found when the observers were expecting a task-relevant, informative stimulus, supporting a mandatory "attend-all" mechanism. However, thus far, the preparation effect was only demonstrated in speeded dot-probe tasks and it is yet to be determined whether preparing for distractors merely facilitates motor preparation or whether it has other outcomes, such as modifying early perceptual processes. Replacing the dot-probe task with a four-letter difficult memory task revealed that memory encoding was facilitated when the letters appeared when informative stimuli were expected. In contrast, there was no benefit, nor cost, for memory encoding when distracting stimuli were expected. These results indicate that the preparation effect is more flexible than thought.

The consequences of preparing for informative or distracting stimuli

Perceptual-motor tasks offer redundant solutions to achieve a goal. However, not all solutions are equally robust to error-producing variability and thus, skill learning requires the search for behaviors that are error-tolerant. Throwing a ball to hit a target is a skill that has been studied via the “skittles” task, a simplified 2-dimensional throwing environment. This paper investigated whether participants learning the skittles task benefit from task scaffolding via the placements of targets provided by either a human confederate posing as a teacher, or by the participant. These groups were compared to a third sample who received an unstructured curriculum consisting of randomly placed targets. The results demonstrate that participants in the unstructured training condition were better able to perceive the task’s underlying solution geometry, compared to scaffolding provided by a human. Future work will explore methods to design training curricula that are more efficient in facilitating perceptual-motor skill development.

Influence of Curriculum Structure on Early Skill Learning during a Virtual Throwing Task

Response inhibition—our ability to suppress actions—is achieved through coordinated reactive and proactive processes that may recruit different neurophysiological mechanisms. Here we adapted a two-step continuous performance task in which the decision to respond depends on a combination of an initial context cue and a subsequent target probe. Using transcranial magnetic stimulation (TMS), we mapped changes in corticospinal excitability and inhibition, providing indices of reactive and proactive processes that influence the state of the motor system in the lead-up to initiating or suppressing a response. We found distinct changes in corticospinal excitability at critical timepoints when participants were preparing in advance to inhibit a response (during the cue) and while inhibiting a response (during the probe). Motor system activity during early timepoints correlated with behavioural indices of proactive capacity and predicted whether participants would later successfully inhibit their response. Implications for theories of inhibitory control will be discussed.

Distinct patterns of proactive and reactive inhibition in the human motor system

The ability to discriminate between large and small quantities is a core aspect of basic numerical competence in both humans and animals. In this work, we examine the extent to which the state-of-the-art neural networks designed for vision exhibit this basic ability. Motivated by studies in animal and infant numerical cognition, we use the numerical bisection procedure to test number discrimination in different families of neural architectures. Our results suggest that vision-specific inductive biases are helpful in numerosity discrimination, as models with such biases have lowest test errors on the task, and often have psychometric curves that qualitatively resemble those of humans and animals performing the task. However, even the strongest models, as measured on standard metrics of performance, fail to discriminate quantities in transfer experiments with differing training and testing conditions, indicating that such inductive biases might not be sufficient.

Premium content

WAX: A Dataset for Word Association eXplanations

Next from CogSci 2023

Prior Beliefs About the Compatibility of Disjuncts Impact the Exclusivity Implication of Disjunction