United States

To what extent do human reward learning and decision-making rely on the ability to represent and generate richly structured relationships between options? We provide evidence that structure learning and the principle of compositionality play crucial roles in human reinforcement learning. In a new multi-armed bandit paradigm, we found evidence that participants are able to learn representations of different reward structures and combine them to make correct generalizations about options in novel contexts. Moreover, we found substantial evidence that participants transferred knowledge of simpler reward structures to make compositional generalizations about rewards in complex contexts. This allowed participants to accumulate more rewards earlier, and to explore less whenever such knowledge transfer was possible. We also provide a computational model which is able to generalize and compose knowledge for complex reward structures. This model describes participant behaviour in the compositional generalization task better than various other models of decision-making and transfer learning.

CogSci 2021

Compositional generalization in multi-armed bandits

It is my great pleasure and honor to welcome you to the 64th Annual Meeting of the Association for Computational Linguistics (ACL 2026), held in San Diego, California, USA, July 2–7, 2026. ACL 2026 continues our fields tradition of excellence in scholarship, innovation, and inclusivity, and I am deeply grateful to the many volunteers who have worked tirelessly to bring this event to life.

It is exciting to see the many developments in the field reflected at our conference, such as the advancement of large language models (born out of our community) into many application directions. This is not without challenges, partly due to unsolved problems, and partly due to misuse of the technology. All this is reflected well at ACL 2026. The conference also reflects that the field is not only about advancing technology, but also the science of fundamentally understanding language and how it is use by many different speaker communities. <br>
[Full Message from General Chair Philipp Koehn](https://drive.google.com/file/d/1QiJFOmMjvDgsDo3i1XzlQnImxe3RU_Cy/view?usp=sharing) <br><br><br>
<html><button style="display: inline-flex; align-items: center; justify-content: center; white-space: nowrap; border-radius: 9999px; font-weight: bold; background: #7c3aed; color: white; font-family: 'Space Grotesk', sans-serif; height: 40px; font-size: 16px; padding: 0 20px; border: none; cursor: pointer" onclick="window.open('https://drive.google.com/file/d/1QiJFOmMjvDgsDo3i1XzlQnImxe3RU_Cy/view?usp=sharing')">Conference Handbook</button></html>
<br><br><br><br>

To access this event page, you need to log in with the **email address you registered with**. <br>Access credentials will be sent to your email from Underline -  subject line "Welcome to ACL 2026". Please be sure to check your spam email folder if you do not see an email confirmation right away.

To access this event page, you are required to register.
Please complete your registration to continue.

We recommend reading [**the registration information**](https://2026.aclweb.org/registration/) first.

**Online Registration Form**: https://acl.swoogo.com/acl2026

Registration Required

ACL 2026

The 64th Annual Meeting of the Association for Computational Linguistics (ACL 2026) will take place in San Diego, California, from July 2 - 7, 2026.

poster

### 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.

Procrastination is ubiquitous, but its underlying processes are poorly understood. Studies of procrastination have used self-report questionnaires and therefore have limited use as the basis for process models. As typical procrastination behavior is characterized by a delay in starting the work and rushing in the end, we argue that studies should emphasize the time course of progress. We designed a reading task to quantify the time course of procrastination. Subjects were given seven days to work online on a boring and lengthy reading task. We tested whether reward delay is a predictor of procrastination. We introduce a metric for quantifying the degree of procrastination from the time course of progress. The degree of procrastination tends to be higher in the delayed reward condition than that in the instantaneous condition. We also observed great individual variation in the progress course. Further work needs to investigate what factors contribute to this variability.

Designing a behavioral experiment to study the factors underlying procrastination

In social psychology, a common finding is that people prefer confirmation-biased information. Although this confirmatory information seeking is commonly treated as an error in judgment, we note that biased sources of information can sometimes be more useful than more accurate sources. Such confirmatory sources will only advise someone to deviate from the policy they think is most useful when these sources are sure taking alternative action is correct. For this reason, these sources can allow people to avoid particularly costly errors. Avoiding such costly errors can sometimes be worth the price of inaccurate beliefs, even though these beliefs lead to more errors in total. In two studies, we find initial support for this idea. Within a Partially Observable Markov Decision process, we show that participants prefer optimistically-biased information when they would otherwise miss out on a particularly large reward, and pessimistically-biased information when they would otherwise face particularly strong punishment.

Deciding to be wrong: Optimism and pessimism in motivated information search

We report a supervised category learning experiment in which the training phase contains both classification and observation learning blocks. To explain the use of different categorization strategies, we propose an account in which use of a stimuli dimension depends on how well the dimension is learned. Our results show that there is an overall preference for a unidimensional categorization based on the perfectly diagnostic dimension. The preference for unidimensional categorization is negatively correlated with how well participants learn the partially diagnostic dimensions. Preference for unidimensional categorization is also negatively correlated with the mean response time. Bayesian modeling results show that participants use a partially diagnostic dimension only when it is learned with a very high level of accuracy. Different strategies are used for categorization depending on how well the perfectly and partially diagnostic dimensions are learned.

Supervised category learning: When do participants use a partially diagnostic feature?

Passability of an aperture, as a perceived affordance, is determined by the fit between the apparent aspects of the environment (e.g. perceived gap) and the perceived body scale. Here, in order to understand the effects of body scaling on the affordance of passability, we conducted a virtual reality study in which in blocked trials, we assigned participants (N = 20) different shoulder widths (narrow, normal, and wide). Participants were instructed to walk naturally to pass through an aperture scaled to their virtual shoulder without colliding and to reach a target on a table. The results showed that participants were closer to the target on the table when assigned narrow rather than normal shoulders. Also reflected in their perceptual judgments, those with narrow virtual shoulders thought they had smaller shoulder width, an effect not seen in the wide shoulder condition, which together demonstrate an asymmetry in the effects of body scaling.

Effects of Scaling Shoulder Width on Passability Affordance in Virtual Reality

Empirical evidence demonstrates the superiority of interleaved over blocked learning schedules in inductive learning, yet many students believe the opposite. We investigated the effect of metacognitive activities on learners’ awareness of interleaved benefits. Participants experienced two learning schedules for different painting styles and took a transfer test. Then we manipulated the provision of post-test feedback for each schedule and metacognitive prompts to reflect on the learning schedules. Finally, participants were asked to choose a more effective schedule between the two. Although there were no significant effects of either the post-test feedback or metacognitive prompts, the performance difference between the two schedules significantly predicted participants’ schedule selection. The greater benefits from the interleaved schedule the participants experienced, the more likely they chose the interleaved schedule. Furthermore, participants’ responses to metacognitive prompts revealed that those who acknowledged spacing benefits from the interleaved schedule were more likely to choose the interleaved schedule.

Appreciating interleaved benefits: The effect of metacognitive activities on the selection of learning strategy

Humans can adapt their knowledge acquired for one problem to solve other problems of different problem domains. To seek for evidence of knowledge transfer, we investigated
the human navigation behavior on the network of Wikipedia. We showed that the performance of human players is between that of the best navigator with full knowledge
on the network structure of Wikipedia and of the poorest navigator with no knowledge on it. This suggests that human players transferred their knowledge on the nominal concepts onto the network structure of Wikipedia. Further, we conducted an analysis of the degree of confidence in decision making based on reinforcement learning. We found that human
players might be very certain even about their first choice, from which we suspect that human players can solve the new problem by transferring their everyday knowledge successfully from the beginning and this can be a true power of knowledge transfer across problem domains.

Transfer of Knowledge in a Semantic Navigation Task Without the Accurate Map: Model-based Analysis of Knowledge Transfer

Long perceptuo-motor sequences underlie skills from walking to language learning, and are often learned gradually and unconsciously in the face of noise. We used a non-parametric Bayesian n-gram model (Teh, 2006) to characterize the multi-day evolution of human subjects’ implicit representation of a serial reaction time task sequence with second-order contingencies. The reaction time for an element in the sequence depended on zero, one and more preceding elements at the same time, predicting frequency, repetition and higher-order learning effects. Our trial-level dynamic model captured these coexistent facilitation effects by seamlessly combining information from shorter and longer windows onto past events. We show how shifting their priors over window lengths allowed subjects to grow and refine their internal sequence representations week by week.

Tracking the Unknown: Modeling Long-Term Implicit Skill Acquisition as Non-Parametric Bayesian Sequence Learning

Although it is documented that vision can impact flavor perception, less known are the interrelations between subsets of cross-modal factors. Within the framework of cross-modal perception three between-group experiments were conducted to determine the impact of color and container type on the flavor perception. Two different containers and 10 colors were tested in two online experiments (n1 = 67, n2 = 63); in the third experiment two non-sweetened colored drinks were tested (n = 32). Hedonic, associative, and emotional measurements were applied.
Our results indicate that color can increase the expected sense of flavor. For instance, red, pink, orange (average values 3.7, 3.7, 3.6 on a 5 pt scale) are the strongest examples of sweetness. For some colors (e.g., red and brown) predicted sweetness is also determined by the type of container (bottle vs. glass). Additionally, the sense of freshness as a cross-modal factor increases the likeability of the drink.

Impacts of colors and container types on predicted and perceived flavor of non-alcoholic beverages

Many cognitive models assume that stimuli can be represented as points in a latent psychological space. However, it has been difficult to provide these spaces with a geometric structure where the distance between items accurately reflects their subjective dissimilarity. In this paper, we propose a new method to give psychological spaces a geometric structure by equating the amount of change undergone by a stimulus with the arc length of a curve in psychological space. We then assess our method with a categorization experiment where participants classified continuously changing visual stimuli according to their rate of change. Our results indicate that individuals’ judgements are well predicted by arc length, suggesting that it may be a promising geometric constraint for psychological spaces in other contexts.

Arc Length as a Geometric Constraint for Psychological Spaces

Probability density estimation is widely-known as an ill-posed statistical problem whose solving depends on extra constraints. We investigated what prior beliefs people might have in their learning of an arbitrary probability distribution, especially whether the distribution is believed to be unimodal or multimodal. In each block of our experiments, participants repeatedly reconstructed a one-dimensional spatial distribution after observing every 60 new samples from the distribution. The probability distribution function (PDF) they reported on each trial was submitted to a spectral analysis, where the powers for 1-cycle, 2-cycle, …, n-cycle components respectively indicate participants’ tendency of reporting unimodal, bimodal, …, n-modal distributions. In two experiments, the reported PDFs showed significant bimodality—that the 2-cycle power was above chance and even larger than the 1-cycle power—not only for bimodal distributions, but also for uniform distribution. Such illusory bimodality for uniform distribution was first found when we used an adaptive procedure analogous to “human MCMC”, updating the generative distribution of samples from trial to trial to reinforce potential biases in PDFs (Experiment 1). However, even when we fixed the generative distribution across trials (Experiment 2), the illusory bimodality did not vanish. The illusory bimodality was even observed before participants experienced any bimodal distributions in the experiment . We considered a few kernel density models and discuss further computational explanations (e.g. prior beliefs following Chinese Restaurant Process) for this new phenomenon.

Downloads

Next from CogSci 2021

Designing a behavioral experiment to study the factors underlying procrastination

Similar lecture

Beyond Monolithic Rewards: Hybrid Multi-Aspect Reward Optimization

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES