United States

In this work, we propose an interpretable knowledge tracking model (KVFKT) based on key value and forgetting. This model maintains the students’ knowledge state matrix in each memory unit, and also maintains the corresponding forgetting matrix to track the last review timestamp of students’ knowledge points. In addition, this model introduces the IRT two parameter model from the perspective of educational psychology to further enhance the interpretability of the model. Finally, the model is tested on four real-world datasets, and the experimental results prove the advantages of the method described in this paper.

CogSci 2023

An Interpretable Model With Forgetting Matrix For Deep Knowledge Tracking

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.

The unprecedented spread of smartphone usage and its various boarding sensors have been garnering increasing interest in automatic mental health detection. However, there are two major barriers to reliable mental health detection applications that can be adopted in real-life: (a)The outputs of the complex machine learning model are not explainable, which reduces the trust of users and thus hinders the application in real-life scenarios. (b)The sensor signal distribution discrepancy across individuals is a major barrier to accurate detection since each individual has their own characteristics. We propose an explainable mental health detection model. Spatial and temporal features of multiple sensory sequences are extracted and fused with different weights generated by the attention mechanism so that the discrepancy of contribution to classifiers across different modalities can be considered in the model. Through a series of experiments on real-life datasets, results show the effectiveness of our model compared to the existing approaches.

Personalized multi-task attention for multimodal mental health detection and explanation

Thousands of learners are taking massive Open Online courses (MOOCs) in search of learning and self-improvement opportunities. However, the learning process often involves a struggle with confusion, which can adversely affect the course participation experience, leading to dropout halfway through. At the same time, it is impossible for the instructors to recognize and respond “confusion” posts from all posts accurately and timely, due to the large volume of registrants. This study identifies "confusion" posts that require instructors' attention through the BERT-CNN model and performs an attribution analysis of confusion based on the visualization results of confusion types and language features. We find that posts identified as "confused" actually represent different information needs of learners, which are mainly related to cognitive and social processes. Furthermore, we evaluated the BERT-CNN model with a 13% improvement in F1 value compared to previous studies and successfully applied it across unseen domains.

Feature Visualization and Attribution Analysis of Confusion for Massive Open Online Course

Behavior Trees (BTs) are widely used in many games, however, they are normally designed in a static way which is hard to deal with dynamic environments. One of the factors that restrict the performance of BTs is the optimization of resources used among sub-tasks, which has not been paid much attention. In this paper, we concentrate on the problem of dynamic allocation of agents for BTs, i.e., an important resource in games. We design a novel hierarchical learning framework, including the allocation learning module and the action learning module, to find strategies for agent allocation and action selection. Each agent is associated with a network for encoding observations, which is then assigned to a sub-task by the allocation learning network. Each sub-task uses a network to compute the action for the corresponding agent. We design how the losses are computed for the two modules respectively and present the whole training procedure. Finally, we examine the performance of the proposed framework on a well-known game, i.e., Google Research Football. The results show that each module works effectively and the whole framework improves the performance of BTs greatly. This work will have an important significance for the community of the usage of BTs.

Dynamic Agent Allocation with Reinforcement Learning for Applying Behavior Trees in Games

Despite increased interest in creating more diverse and inclusive organizational environments, scholars still have relatively little understanding of the cognitive mechanisms underlying leadership formation and how biases against marginalized groups are learned and maintained. Drawing from leadership emergence theory, we investigate potential interventions that support diverse leaders. Using agent-based simulations, we model a collective search process on a fitness landscape. Agents combine individual and social learning, and are represented as a feature vector blending relevant (e.g., individual learning characteristics) and irrelevant (e.g., race or gender) traits. Agents learn how to weigh the traits of others on the basis of reward expectations and dynamically adjust the connections of their social network to grant leadership. By initializing the network to privilege agents with irrelevant traits, we can create biases in how agents weigh certain traits. This framework allows us to test interventions to unlearn bias in favor of performance-relevant traits.

Unlearning the bias: An agent-based simulation for increasing diverse representation through leadership emergence

The ability to detect and analyze a human figure is key to our survival and social interactions. Efficient and robust identification of body parts can help to interpret images when bodies are partially occluded. While previous studies emphasized the configural processing of whole bodies using simplified stimuli, it still remains unclear how local contexts about body parts are integrated to resolve ambiguities (e.g., from occlusion) regarding the identities of or spatial relations among body parts. In a series of online experiments, we asked human observers to identify one ambiguous target body part in the presence of another “context” part. Our results showed that humans can use various amounts of spatial context to discount local ambiguities in natural images of pairs of parts, and are sensitive to low- and mid-level cues such as alignment and connectedness. Further simulations using computational models showed that the models with pairwise relational knowledge based on local part appearances failed to capture human sensitivity to relational properties of spatial context between pairs of body parts. Our results provided behavioral evidence in the human visual representations of various spatial relations between pairs of body parts and revealed different representations of spatial context used in humans and models.

Identifying Body Parts in the Spatial Context of Pairwise Relations: Human Psychophysics and Model Simulations

This study describes the experimentally developed technique of sentiment analysis in an automatic semantic parser that processes daily news input blocks. The 2022 time period was used to analyze news in Telegram channels published by twelve official news agencies and five influential non-official news sources. In the advanced analytical system that we present, automatic learning is combined with a frame-based approach to identify patterns that influence judgments based on a frame representation of the text. The parser analyzes the syntactic structure of each sentence input, constructs its semantic representation, and associates it with one of 3800 semantic frames extracted from a large text corpus. The research revealed that the relative frequency and combination of frames varied between the two groups of news sources. The results prove that news is conceptualized and represented by various media sources in very different ways with the potential to affect the recipient's beliefs and emotions.

Automated Semantic Frame Analysis of Telegram News Channels

We present two models of visual location memory developed within the ACT-R cognitive architecture and compare the model’s performance to that of human participants in a pattern reproduction task. The snapshot model has a fovea-peripheral based activation mechanism, which simulates how more attention and processing resources are given to the centre of the visual field for short stimulus expo-sure trials (50ms and 200ms). For long exposure trials (>=1s), a chunking model was developed based on the snap-shot model by adding chunking processes which can en-code geometric patterns. Both models can match the task response accuracy and pause data of human participants. The results of the modelling reveal that for the short stimulus exposure trials the accuracy of recall is affected by the distance between the object location and the fovea vision location. For trials with long stimulus exposure times, participants were likely to use salient geometric patterns to encode the configuration of discs.


Modelling Pattern Reproduction in The ACT-R Cognitive Architecture

Dual processes theories of memory propose two separate processes: recollection (intentional and deliberate) and an automatic component (effortless). Previous studies have established a strong positive correlation between people’s Judgements of Learning i.e. predictions about their memory performance and the recollection component, but the studies have been inconclusive regarding the link of JOLs with the automatic component. A study by Undorf et. al. (2016) using Process Dissociation Procedure has been able to show a correlation between metamemory and the automatic component of memory using JOLs obtained during the learning phase.. Experiment 1 is a conceptual replication of this study, where we measure the dual processes using a different task and take JOLs at the time of encoding to test the relation between JOLs and the two components of memory. Experiment 2 tests this further by taking the retrospective confidence judgements (RCJ) at the time of retrieval, which is a better predictor of memory performance. Individual correlations between metamemory and the two memory processes across participants are also analyzed. While the correlations obtained through binning JOLs showed similar results to Undorf et al (2016), correlations across subjects using mean JOLs mostly did not show significant correlations. RCJs were better predictors of memory performance than JOLs. In Experiment 2, correlations between RCJs and recollection were significant indicating that metamemory obtained during retrieval perhaps cannot predict automatic influences in memory. 
Keywords: memory; metamemory; metacognition; judgments of learning; retrospective confidence judgment; process dissociation

Metamemory based on retrospective judgments only predict recollection but not automatic influences

In this paper, we simulated true and false recognition in the Deese/Roediger/McDermott (DRM; Deese, 1959; Roediger & McDermott, 1995) paradigm by incorporating neural embeddings for words (word2vec vectors) into an instance memory model (MINERVA2). Previously, Arndt and Hirshman (1998) demonstrated that MINERVA2 (Hintzman, 1988) could capture multiple classic false recognition findings with randomly generated word representations. However, as randomized representations deviate systematically from semantic representations for words learned from the natural language environment, there remains uncertainty about whether MINERVA2 can capture the false memory illusion when scaling up to real-life complexity in word representations. To address this uncertainty, we used word2vec embeddings that are derived from large corpora of natural language instead of randomized representations in MINERVA2. Our results showed that MINERVA2 can still capture the standard true and false recognition, and it can also accommodate the true and false recognition effects of various classic manipulations (e.g., associative strength, number of associates, divided attention, retention interval).

Integrating Neural Embedding Models with an Instance Memory Model to Explain False Recognition

Making judgments of learning (JOLs) can not only monitor but also directly modify memory performance, which is termed JOL reactivity. One hypothesis for JOL reactivity is that making JOLs enhances item-specific processing (as opposed to relational processing). Based on this account, JOL reactivity is expected to be stronger with study materials that encourage relational processing over item-specific processing, such as categorized word lists. However, there have been mixed findings on whether making JOLs enhance recall for categorized lists. The current study identified a key factor that contributed to the mixed findings: list organization. Namely, whether exemplars of a category are presented consecutively (blocked) or randomly intermixed with exemplars of other categories (randomized). Moreover, because blocked lists emphasize relational processing more than randomized lists, the item-specific processing account predicts that making JOLs should produce a larger memory enhancement for blocked lists relative to randomized lists. However, our results showed the opposite: Making JOLs enhanced recall for randomized but not for blocked categorized lists. Further, the dual-retrieval model results showed that the effects of JOLs are localized in gist-based familiarity rather than item-specific recollection for randomized lists. Together, our findings present a challenge to the item-specific processing account for JOL reactivity.

Next from CogSci 2023

Personalized multi-task attention for multimodal mental health detection and explanation

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES