Canada

Human communication usually exhibits two fundamental and essential characteristics under environmental pressure, efficiency, i.e., using less communication frequency to achieve comparable performances in cooperation; and robustness, i.e., maintaining a relative performance when communicating in a complicated environment. Since a critical goal of designing artificial agents is making them human-like in many scenarios. How artificial agents could learn a human-like communication mechanisms in terms of efficiency and robustness is a long-existing problem which has not yet been solved. Reinforcement Learning, due to its trail-and-error paradigm, provides a promising framework in solving the above research problem. With reinforcement learning, this paper develops architectures that help agents learn an efficient and robust communication, and carries out extensive experiments which uncover that artificial agents are cognitively capable to learn such human-like communication protocol in various environments (tasks).

CogSci 2022

Achieving Consensus to Learn an Efficient and Robust Communication via Reinforcement Learning

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. 





poster

As virtual reality devices get more widespread, investigating how humans explore omnidirectional panoramic scenes has become a topic of interest. Several computational models now exist for predicting visual saliency in such environments. However, the existing literature largely ignores auditory cues while estimating saliency maps. To pave the road for more capable models, here, we ask how scene attributes and sound categories influence visual exploration. Considering a dataset of omnidirectional videos of dynamic scenes and eye fixations collected with and without sound, we analyzed how different semantic factors affect eye movement behavior; whether there are significant differences across human, animal, or music sounds present in the videos, or for indoor or outdoor scenes. Analyses revealed that the regulatory effects of audio in human fixations significantly increase in indoor scenes or if the audio is of music but decrease in outdoor scenes or if the audio is of animal or human sounds.

How scene attributes and sound influence visual exploration of omnidirectional panoramic scenes

Object location and identity interact with each other asymmetrically in adults: Participants tend to judge two sequentially presented objects as more similar when they appear in the same location compared to in different locations (a phenomenon referred to as Spatial Congruency Bias). Yet no comparable identity congruency bias is found, suggesting that location is special in object-binding. To investigate whether such a special role of location is inherent in our visual system or acquired during development, we conducted experiments in both 5-year-old children and adults performing location-judgment and identity-judgment tasks. The study replicated asymmetric bias in adults by finding that location biased their identity judgments, but not vice versa. However, symmetric bias was found in children, suggesting fused processing and reciprocal influence between location and identity in early childhood. The results indicate that location maintains its influence across development, whereas object identity’s influence on location gets pruned away during development.

Differential patterns of object-location binding in children and adults: Pruning the special role of location

Here, we test whether the perception of liquids is based on abstract, physics-based representations. To do so, we use a novel behavioral task that requires generalization across scene configurations. We created 30 animations of liquids flowing through and interacting with surfaces and objects. These animations came from six qualitatively different scene configurations and five viscosity levels. During the experiment, participants viewed two liquid animations (side-by-side), each from a different scene configuration, and they rated the similarity of underlying viscosities. We found that not only average similarity ratings are linearly predicted based on true viscosity differences, but also the identity of the scene configuration pairs was uncorrelated to similarity ratings. This successful generalization in human behavior suggests an underlying abstract, physics-based representation of liquids. We also present a new probabilistic model of liquid perception that accounts for behavioral ratings, quantitatively supporting our hypothesis.

Perception of liquids relies on generalizable, physics-based representations

Multisensory processing often has facilitation and/or interference effects (Chen & Spence, 2010; Thomas et al., 2017) yet the mechanisms remain unclear. This experiment used a Stroop-like task (Stroop, 1935) to examine how congruent, incongruent, and irrelevant information presented in one sensory modality affects processing and responding in a different modality (see Figure 1 for trial types). Participants were simultaneously presented with pictures and sounds of animals or vehicles. In separate blocks they had to respond to either the visual or auditory stimulus, while ignoring the other modality. Results suggest that visual stimuli have a larger effect on auditory responding than vice versa. Incongruent visual stimuli had a larger effect on accuracy while congruent visual stimuli had a larger impact on response time. We will present additional data at the poster focusing on factors that may affect congruency effects including individual differences and types of auditory stimuli.

Semantic Congruence Across Sensory Modalities

In this paper, we have studied the impact of clickbait headlines on the distribution of visual attention on hyperlinked news articles. Visual attention is a driving factor in ad-based revenue models that support online journalism. Importantly, it is also an indicator of cognitive processes involved in reading and comprehension. We hypothesize that articles with clickbait headlines receive lesser visual attention when controlled for articles’ content. This is based on the premise that a significant proportion of clicks on clickbait headlines are driven by readers’ specific epistemic curiosity rather than knowledge acquisition. An eye-tracker setup was used to infer visual attention from the gaze-fixation analysis conducted on data from 60 participants. Our results suggest that clickbait headlines significantly reduce the visual attention on news articles. Though, article content comprehension measured by a recall test was comparable for clickbait and non-clickbait headlines. Our findings add to the discussions on the cognitive attention and the implications of using clickbait headlines for news publishers, newsreaders, and advertising agencies alike.

Clickbait’s Impact on Visual Attention – An Eye Tracker Study

Having an intention (or not) forms a categorisation system for learning. Correspondingly, incidental learning stands for learning without intention to learn. The current paper examines the influences of incidental and intentional learning. 170 adults participated in three experimental studies. In each experiment, participants examined some animal illustrations and read paragraphs about them. In the first study, animal illustrations were arranged as being common (i.e., low curiosity) or novel to examine the effect of curiosity. Low curiosity level was more effective on incidental learning and vice versa for intentional learning. In the second study, mental time travel manipulation showed focusing on current time increased intentional learning. In the third study, the effect of arousal was examined via the tempo of the music (i.e., faster tempo for high arousal level). Low arousal level increased incidental learning. Ultimately, curiosity, mental time travel, and arousal could be essential in intentional and incidental learning.

Predictors of Incidental and Intentional Learning: Curiosity, Construal Level, and Musical Arousal

Parallel accounts of interference resulting from the generation of related words can be found in the retrieval-induced forgetting (RIF) and the cumulative semantic interference literatures. Recent work on the language production side suggests that the same adaptive learning process may underlie both. However, the literatures remain separate. They use different procedures and dependent measures, and theoretical accounts focus on underlying conceptual representations (memory research) vs. conceptual-lexical links (language research). We propose that the accounts should be reconciled. As an initial step toward this goal we combined a retrieval/generation procedure with a continuous picture-naming test phase to assess their combined effects on interference. We observed both costs and benefits in error data. There were more naming errors (including many time-outs) for non-generated items from activated categories and fewer for previously generated items. Perhaps due to a too-severe cutoff, naming times did not show a RIF influence, only a marginal facilitation effect for generated items. However, naming time showed typical cumulative interference within the picture-naming phase independent of previous retrieval experience. Future work will investigate the locus of interference in conceptual memory representations versus in links to word representations with the goal of producing a unified account of semantic interference.

Aligning Language and Memory Accounts of Semantic Interference

The current information ecosystem has made it difficult to judge the veracity of information. Presumably, evaluating an informational source’s expertise aids in this judgment process. However, whether these judgements influence memory for that information is unclear. In this study, we investigate the extent to which a source's expertise influences recognition memory for explanations given by that source. Participants viewed 20 vignettes describing events and two potential explanations for each event. Each explanation was either said by a source (domain expert vs. non-expert) or given without a source. Participants were exposed to both source and no-source conditions. We investigated whether participants’ recognition of explanations were predicted by condition and source expertise. Our findings indicated that neither condition nor source expertise influenced participants' performance on the recognition task. This suggests that people either do not attend to source in their evaluations of explanations or do not encode these evaluations.

Does Being an Expert Matter? The Influence of Source Expertise on Recognition Memory

Previous demonstrations of false memories for predicted but not presented words used slow encoding and immediate retrieval conditions, potentially exacerbating false memory effects. We present two experiments that investigated whether false memories also occur under self-paced encoding and delayed retrieval conditions, and whether false memories are reduced when the initial prediction was disconfirmed by an implausible word, thought to elicit false memory suppression. Results showed that previous demonstrations of false memories were not contingent on the task conditions: False memories also occur when language processing is self-paced, and they affect longer-term memory structures. Crucially, false memories emerged regardless of whether the prediction-disconfirming word was plausible or not. Results are evaluated against a recent psycho-linguistic account that makes diverging predictions regarding the processing consequences of mild and severe violations of plausibility.

Uninvited and unwanted: False memories for words predicted but not seen

We investigated whether schema congruency differentially affects low level sensory processing (eye gaze) compared to higher-level cognition (memory). Participants performed a two-phase eye tracking task; first a baseline phase with only congruent cartoon events, and subsequently an experimental phase in which the same events were adapted to remain congruent or become incongruent to a theme. Results revealed that participants became quicker in recognizing the congruent cartoon events compared to incongruent in the experimental phase, indicating improved memory for congruent cartoon events. No mean difference in gaze towards congruent versus incongruent events was observed. Surprisingly, a slight bias towards gazing to the left side of the screen in the baseline phase diminished during the experimental phase, indicating that the schema congruency manipulation might affect gaze behavior. Taken together, our results suggest that our schema congruency manipulation affects gaze behavior and memory, but further eye tracking analysis could reveal the dynamic nature of this effect.

Premium content

Downloads

Next from CogSci 2022

How scene attributes and sound influence visual exploration of omnidirectional panoramic scenes

Similar lecture

Towards Situated Communication in Multi-Step Interactions: Time is a Key Pressure in Communication Emergence