United States

We study human learning &amp; decision-making in tasks with probabilistic rewards. Recent studies in a 2-armed bandit task find that a modification of classical Q-learning algorithms, with outcome-dependent learning rates, better explains behavior compared to constant learning rates. We propose a simple alternative: humans directly track the decision variable underlying choice in the task. Under this policy learning perspective, asymmetric learning can be reinterpreted as an increasing confidence in the preferred choice. We provide specific update rules for incorporating partial feedback (outcomes on chosen arms) and complete feedback (outcome on chosen &amp; unchosen arms), and show that our model consistently outperforms previously proposed models on a range of datasets. Our model and update rules also add nuance to previous findings of perseverative behavior in bandit tasks; we show evidence of outcome-dependent choice perseveration, i.e., that humans persevere in their choices unless contradictory evidence is presented.

CogSci 2021

Tracking what matters: A decision-variable account of human behavior in bandit tasks

### 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 study human learning & decision-making in tasks with probabilistic rewards. Recent studies in a 2-armed bandit task find that a modification of classical Q-learning algorithms, with outcome-dependent learning rates, better explains behavior compared to constant learning rates. We propose a simple alternative: humans directly track the decision variable underlying choice in the task. Under this policy learning perspective, asymmetric learning can be reinterpreted as an increasing confidence in the preferred choice. We provide specific update rules for incorporating partial feedback (outcomes on chosen arms) and complete feedback (outcome on chosen & unchosen arms), and show that our model consistently outperforms previously proposed models on a range of datasets. Our model and update rules also add nuance to previous findings of perseverative behavior in bandit tasks; we show evidence of outcome-dependent choice perseveration, i.e., that humans persevere in their choices unless contradictory evidence is presented.

poster

Executive function and metacognition are two core concepts in cognitive psychology, yet research linking EF and metacognition directly remains rare. It has been suggested that error monitoring can be used as a window to study the association between EF and metacognition (Roebers, 2018). Error monitoring is indexed by post-error slowing, a phenomenon describing a delayed response in actions after error commissions. The present study uses hierarchical multiple regression to investigate whether EF and metacognition can both predict PES. Individual constructs of EF are measured by four computerised-based tasks. Metacognition is assessed by a simplified version of the Self-Regulated Learning Questionnaire (Dowson & McInerney, 2004). A total of 456 participants (M age of 11.9, SD = 0.92) are included in the final analysis. Results indicate that two EF constructs: inhibition and planning, predict corresponding PES in inhibition and planning tasks. Only the regulation component of metacognition predicts PES in the planning task.

Bridging Executive Function and Metacognition through Post-Error Slowing

Preparing the next generation for an ever-changing environment is of utmost importance in Higher Education. Though cultural diversity is highly prevalent in modern societies, (implicit) biases, stereotypes and racism maintain a persistent yet mostly overlooked part of everyday life. Many students are not aware of their biases. Hence, it is imperative for them to undergo a critical self-reflection process about their beliefs and (unconscious) biases. To induce this process, raise awareness and confront them with potential (unconscious) biases and stereotypes, 404 university students in teacher training and vocational education completed an Implicit Association Test (IAT) on skin color before watching a lecture discussing biases and stereotypes. The preliminary results show a variety of reactions after taking the test, ranging from denial and anger to discomfort and surprise. We will discuss possibilities to support students in leveraging the resulting cognitive dissonance as an opportunity to begin their individual self-reflection process.

Unveiling unconscious biases and stereotypes in students: The necessity of self-reflection in Higher Education

Numbers and variables often follow the same principles of arithmetic operations, yet numbers can be computed to a value whereas variables cannot. We examined the effect of symbols—numbers versus variables—on middle school students’ problem-solving behaviors in a dynamic algebra notation system by presenting problems in numbers (e.g., 3+5−3) or variables (e.g., x+y−x). We found that compared to problems presented in numbers, students attempted the problems more times and took more total steps when the problems were presented in variables. We did not find differences in pre-solving pause time or strategy efficiency on the two types of problems, indicating that students might notice problem structure in both types of problems. The results have implications for research on cognitive processes of symbols as well as the design of educational technologies.

Numbers vs. Variables: The Effect of Symbols on Students’ Math Problem-Solving

The present study investigated whether judgments of learning (JOLs) and judgments of inference (JOIs) enhance the retention and transfer of previously studied text (backward effect) and newly studied text (forward effect). Participants read two different passages through Section A and B. After studying Section A, participants made either JOLs or JOIs on Section A, while the control group read texts without any interim judgments. Then all groups studied Section B and took a final test for Sections A and B. For Section A, there were no significant performance differences among the groups, showing no backward effect of metacognitive judgments on text learning (neither JOLs nor JOIs). In contrast, for Section B, the JOI group outperformed both the JOL and control groups on the transfer test, indicating a forward effect of JOIs on text learning. Metacognitive judgments with a higher-order learning goal appear to help subsequent learning of new material.

Do Judgments of Learning and Judgments of Inference Enhance Text Learning?

A mental model is an internal representation that explains how something works. Mental model construction is facilitated by self-explanation, the active generation of explanations for oneself. The overarching goal of this research is to empirically investigate the utility of self-explanation for developing mental models when learning to program. Programming is notoriously challenging and, despite evidence of the importance of mental models for learning, little work has focused on mental models of students learning how to program. They need correct mental models of the notional machine, an abstraction of the steps taken by a computer as it processes a program. Because students do not spontaneously self-explain, we are using a user-centered approach to design a computer tutor to prompt for self-explanation about the notional machine. Here, we present qualitative results on students’ interactions with an initial version of the tutor, including the form of their self-explanations and corresponding mental models.

Investigating the Utility of Prompting Novice Programmers for Self-Explanations to Improve Mental Models

Mapping words to referents is important for language acquisition. Learning a word relies heavily on memory because it requires forming associations between labels and referents, integrating examples across time, and retrieving words. Memory supports, such as variability, can be added to word learning events to help children learn words. This study examined the effect of emotional variability on word learning. Four-year-olds learned eight novel words presented four times in a storybook organized in one of three conditions: no variability, low variability, or high variability. Words with no variability were presented in the same emotional context (i.e., happy, sad, afraid, or angry). Words with low variability were presented with two emotions (e.g., happy and sad), and words with high variability were presented with all four emotions. After hearing the book, preschoolers participated in a generalization test. This study informs our understanding of the role of social contextual variability in word learning.

Preschoolers' Learning of Words with Emotional Variability in Shared Book Reading

There are numerous studies that show that the more learner actively participate in the learning process, the more they learn. Although the use of active learning to increase learning outcomes has been recently introduced in a variety of methods, empirical experiments are lacking. In this study, we introduce two frameworks of human active learning and then conducted two experiments to determine how these frameworks can be used as leaning tools. In experiment 1, we compared three types of active learning and passive learning in order to empirically confirm the effect of active learning. In experiment 2, based on the results of experiment 1, we explored through simulation on machine learning with the frameworks that the more active the learners are, the better outcomes can be obtained. Both experiments showed that active learning both also effective in human and machine learning. Therefore, our analyses of the two experiments fit within the taxonomy and classification of the frameworks of active learning. This result is further significant in that it gives practical implications on human and machine learning methods.

Passive Versus Active: Frameworks of Active Learning for Linking Humans to Machines

Distinguishing spurious correlations from unconfounded causal evidence is a challenge in every day reasoning. Adapting the classic “blicket detector” paradigm (Sobel et al., 2014), we investigate whether 15-17-month-old infants’ neural activity (theta oscillations), indicates that infants recognize when the information they are expecting or observing is confounded or not. By concurrently tracking infants’ eye-movements, we investigate also whether infants correctly infer the functionality of unknown objects (when such an inference is possible, i.e. when data is unconfounded) and predict future events based on these causal inferences. Data collection is on-going (current N = 29) and preliminary analysis suggests infants show increased theta activity in anticipation of receiving un-confounded (as opposed to confounded) information. By relating the neural measures of information expectation to the accuracy of infants’ subsequent predictions, we hope to offer a unique insight into the mechanisms, development, and individual differences in early causal reasoning.

Mechanisms of early causal reasoning: Investigating infants' sensitivity to confounded information in a causal reasoning task, using EEG and eyetracking

`Embodied cognition' suggests that our motor experiences shape our cognitive and perceptual capabilities broadly, but often considers tasks that directly relate to or manipulate the body. Here we study how a history of natural embodied experience affects abstract physical problem-solving in a virtual, disembodied physical reasoning task. 
We compare how groups with different embodied experience -- congenitally limb-different versus two-handed children and adults -- perform on this task, and find that while there is no difference in overall performance, limb-different participants solved problems using fewer actions, and spent a longer time thinking before acting. This suggests that differences in embodied experience drive the acquisition of different meta-strategies for balancing acting with thinking, even on tasks that are designed to equalize differences in embodiment.

Meta-strategy learning in physical problem-solving: the effect of embodied experience

Our social media activity data is a valuable source of information about our preferences, psychological and social processes. However, collecting such private data, including messages, for scientific research is at an early stage (Ueberwasser & Stark, 2017), which is natural given privacy issues involved (Bemmann & Buschek, 2020). Our study is geared towards: i) making the process of sharing personal data more ethical, consensual, informed and comfortable; ii) identifying profiles of participants willing to share these data. 293 students of both technical and non-technical background completed an online questionnaire designed to identify the relationship between willingness to share the data and factors such as: 1) kinds of data; 2) method of data processing; 3) purpose of data gathering and use; 4) demographics of participants. Qualitative and quantitative analyses revealed the categories of participants’ concerns and preferences regarding the form of anonymization conditional on the subjects’ profile and their technical skills.

Tracking what matters: A decision-variable account of human behavior in bandit tasks

Downloads

Next from CogSci 2021

Bridging Executive Function and Metacognition through Post-Error Slowing

Similar lecture

Learning Approximate and Exact Numeral Systems via Reinforcement Learning

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES