United States

As AI and robots are used in society, the relationship between people and AI/robots is becoming increasingly important; when AI/robots assist people, there is a need to focus on where responsibility for the results of the assistance goes. This study focused on the importance of prior knowledge and topics about AI/robots to investigate who is held accountable when AI/robot assistance fails. The experiment was conducted in a three-factor mixed design, and the ANOVA analysis was based on data from 588 participants. Results showed that prior knowledge about AI/robots increased their pursuit of responsibility toward them and their developer/provider. Additionally, shared responsibility for the AI/robot developer or provider increased when the topic was very important to that person. This study helps to reduce the risk for AI/robots developers or providers when they assist people by clarifying who is responsible for failures that occur when AI/robots assist people in different situations.

CogSci 2025

Where Responsibility Lies in Human-AI decision making: The Role of Knowledge and Importance

computer science

statistics

human-computer interaction

psychology

### Welcome to CogSci Conference 2025!

The 47th Annual Meeting of the Cognitive Science Society was a hybrid meeting held in San Francisco. 

<div style="position:relative;padding-top:0;width:900px;height:500px;"><iframe style="position:absolute;border:none;width:100%;height:100%;left:0;top:0;" src="https://online.fliphtml5.com/ebtyf/amvr/"  seamless="seamless" scrolling="no" frameborder="0" allowtransparency="true" allowfullscreen="true" ></iframe></div>

#### About

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.

#### Our History 

* **Society Creation**<br>
The Society was incorporated as a 501(c)(3) non-profit professional organization in Massachusetts in 1979. The organizing committee included Roger Schank, Allan Collins, Donald Norman, and a number of other scholars from psychology, linguistics, computer science, and philosophy. 
<br><br>
* **Conference Creation**<br>
The first conference on cognitive science was held at La Jolla, California in August, 1979, and has occurred annually since then. The proceedings of each conference are published, and those from most years are available through Lawrence Erlbaum Associates, Inc. The annual proceedings of the Cognitive Science Conference represent a major source of information on new work and new ideas in the scientific study of thinking. In 1990, the Society, with help from an anonymous donor, established the David Marr Prize for the best student paper at each annual meeting.
<br><br>
* **Journal Creation**<br>
The Journal, Cognitive Science, began publication in 1976, and is now published by Wiley-Blackwell. The Executive Editor is currently Richard P. Cooper of Birkbeck, University of London, and there are 18 Associate Editors and a 30-member editorial board. It serves as the premier outlet for research reports that intersect two or more disciplines. Copyrights for articles published in the journal are held by the Society. The Governing Board of the Cognitive Science Society voted in late 2006 to found a new journal, Topics in Cognitive Science (topiCS). The Editor in Chief is Wayne Gray, Cognitive Science Department, Rensselaer Polytechnic Institute. The journal seeks to fill a niche not occupied by Cognitive Science Journal or other cognitive science journals. Membership in the Society includes a subscription to Cognitive Science and TopiCS. Copyrights for articles published in the journal are held by the Society.
<br><br>

#### Code of Conduct

By attending the CogSci 2025 Conference, you are required to adhere to the society’s **[Code of Conduct](https://drive.google.com/file/d/1ChPuihLy6jE_BWqfO7J2KKgX35JW2zsM/view?usp=sharing)**.
<br><br>


You need to log in with the email address you registered with. 

Login credentials were sent to you from Underline -  subject line "Welcome to the CogSci 2025 Conference". Please be sure to check your spam/promotional inbox  if you do not see an email confirmation right away.





Please log in to join this event.

To access the site, please register [**here**](https://cognitivesciencesociety.org/registration/).

If you are registered and feel like you are seeing this message by mistake, please make sure you are logged in with the same email that you registered with. 

Please register!

The 47th Annual Meeting of the Cognitive Science Society presents the latest research across cognitive science and highlights the theme of Cognition in Context.

poster

This presentation explores the use of Large Language Models (LLMs) as “copilots” for linguistic research in Frame Semantics and Construction Grammar. We examine the potential of conversational AI, like ChatGPT and OpenAssistant, to augment FrameNet, create new frames, and analyze constructions. We review ways of using locally-run models (DeepSeek-like) and local RAG to the same purpose. We review prompt engineering, qualitative and quantitative analysis, and controlled experiments. Results show LLMs can generate examples and analyses of constructions, expand FrameNet, and work in multiple languages. Careful prompting and critical evaluation are indispensable. Despite limitations, LLMs offer a novel approach to linguistic inquiry, and ongoing work focuses on fine-tuning and integrating them with other tools. We also discuss the development of pedagogical AI for Cognitive Linguistics, including the FrameChat tool, which uses open-source LLMs and FrameNet data to facilitate research and education.

Copilots for Cognitive Linguists 2025

Mathematics is a central tool for understanding the universe, yet how numbers are cognitively and neurally represented remains unclear. This is especially true when focusing on higher-level concepts, such as symbolic integers, including zero, and fractions. Twenty participants were scanned using a high-resolution 7T fMRI while performing a task in which they judged whether the current number was larger or smaller than the previous one. Behaviorally, we found that participants were slower and less accurate when two numbers were closer in distance on the number line. Neurally, our findings suggest that brain areas involved in mathematical processing – such as the intraparietal sulcus, inferior temporal gyrus, and prefrontal cortex - presented a graded neural BOLD response as a function of numerical distance on the number line. However, both behavioral and neural representations clearly distinguish between fractions and integers.

The representational space of symbolic numbers: from integers to fractions

We report a study examining developmental changes in perceptions of disagreements among speakers who use generics to describe contextually-restricted or unrestricted regularities. Sixty-five adults and 222 5-12-year-olds reacted to generic claims from speakers who attributed ostensibly contradictory attributes to a biological kind (“Xs are striped”; “Xs are spotted”). Crucially, we manipulated the scope of each speaker’s claim, described as restricted to a specific context, or unrestricted. Participants assessed faultless disagreement (whether the speakers could “both be right”). Adults were sensitive to contextual restriction: they allowed for faultless disagreement when contextual restrictions mis-aligned, and denied it when both speakers restricted to one context. Young children demonstrated striking partial competence in faultless disagreement judgments much earlier than prior developmental literature suggested. This is the first study to document faultless disagreement between differentially-restricted generics, both in adults and in children. We discuss the developmental trajectory, and implications for social functioning and learning.

Contextual restriction and faultless disagreement about generics across development

Beliefs about potential threats motivate avoidance behaviour. However, avoidance may also provide new learning experiences that can impact threat beliefs. Here, we use a novel task to explore this bidirectional relationship between aversive beliefs and avoidance. Specifically, we evaluate changes in outcome expectancies for locations on a grid-like sea-world before and after participants navigate a boat and avoid threatening locations. Using computational modelling we assess how participants generalise threat information to nearby locations and how these beliefs change as a result of avoidance. Initial piloting indicates that participants differ in the degree to which they generalise beliefs to nearby locations. Simulations further show how increased generalization influences avoidance and highlight the role of avoidance costs in shaping behavior. This work explores individual differences in belief-avoidance interplay and contributes to our understanding of how this interaction can lead to maladaptive patterns typical of anxiety.

Interaction of generalised aversive beliefs and avoidance behaviour

During language comprehension, words that are less expected tend to take more effort. This phenomenon has been described by the hypothesis that cognitive cost scales in surprisal (negative log probability; Hale, 2001; Levy, 2008), with a core justification being that surprisal quantifies the amount by which a rational comprehender's beliefs about meaning change upon encountering a word. However, this focus on next-word prediction may be too narrow. In this work we advocate measuring processing cost directly with the size of the change in beliefs about meaning, a reframing which implies a novel class of potential situations where surprisal may systematically overestimate cost. We identify typographical errors as a test case, and implement estimators of surprisal and belief-update in a noisy-channel model of comprehension as inference about intended strings. In a self-paced reading time study, we present evidence that human reading time behaves as predicted by belief-update size, rather than surprisal.

When unpredictable does not mean difficult to process

Researchers have started seriously considering the epistemic
issue of whether and when we can claim an artificial intelli-
gence (AI) has developed machine consciousness. Most cog-
nitive theories of consciousness employ a functional character-
ization of the property of consciousness. That is, they are com-
mitted to an account of consciousness as a rule-governed pro-
cess over mental states. Some cognitive scientists concerned
with AI advocate an epistemically behaviorist approach to ma-
chine consciousness; however, such approaches taken ontolog-
ically, systematically fail to satisfy reasonable intuitions about
in what consciousness ought to consist, and taken epistemi-
cally, fail to provide sufficient evidence to individuate any in-
ternal property, including consciousness, in non-human sub-
jects. Therefore, in order to assess consciousness in ways that
adequately account for reasonable intuitions as to its proper
definition, such that we can reasonably assert the presence of
machine consciousness in some AI, it is necessary to propose,
test, and revise, functional theories of consciousness.

Behavioral Evidence is Still Insufficient to Identify Consciousness

As AI models grow in power and generality, understanding how agents learn and make decisions in complex environments is critical to promoting ethical behavior. This study introduces the Odyssey, a lightweight, adaptive text-based adventure game, providing a scalable framework for exploring AI ethics and safety. The Odyssey examines the ethical implications of implementing biological drives—specifically, self-preservation—into three different agents: a Bayesian agent optimized with NEAT, a Bayesian agent optimized with stochastic variational inference, and a GPT-4o agent. The agents select actions at each scenario to survive, adapting to increasingly challenging scenarios. Post-simulation analysis evaluates the ethical scores of the agent’s decisions, uncovering the trade-offs it navigates to survive. Specifically, analysis finds that when danger increases, agents ethical behavior becomes unpredictable. Surprisingly, the GPT-4o agent outperformed the Bayesian models in both survival and ethical consistency, challenging assumptions about traditional probabilistic methods and raising questions about the source of LLMs' probabilistic reasoning.

The Odyssey of the Fittest: Can Agents Survive and Still Be Good?

Compositional capacity (i.e., chunking four objects as two sets of two) can extend 12- to 14-month-old infants’ working memory capacity from three to four1,2,3. Here we ask whether numerical composition supports 3- to 4.5-year-olds’ creating representations of large exact numbers such as five. In a non-verbal object-tracking task (See Fig1a), subset-knowers and most young CP-knowers failed to track exactly five objects. In two experimental manipulations, we provided children with spatiotemporal, linguistic, and/or color chunking cues. If tracking sets of five as a composition of two and three is within children’s compositional capacities, they should perform better than children from baseline. We found no evidence that 3- to 4.5-year-olds can represent exactly five by composing representations of two and three (Exp.1: F(4, 133) = 5.69, β = -0.03, p = .509; Exp.2: F(4, 143) = 4.94, β = -0.03, p = .587).

The Role of Compositionality in Children’s Creating Representations of Large Exact Numbers: A Case Study of the Number Five

Having reasonable expectations of how scenes will unfold is crucial in our life. One prominent hypothesis is that we perform probabilistic physics simulation in our mind. However, the question of how people infer underlying scenes from observations and how this affects downstream predictions about physics interactions is under explored. Current models usually make simplified assumptions that the 3D geometric states of objects are already given. To better understand the role of perceptual uncertainty in people’s physical predictions, we explore the idea of vision as inverse graphics and design a model that can infer a posterior distribution of object states given the raw visual inputs. This perceptual uncertainty is then propagated to a probabilistic physics simulation model to derive physical predictions. We compared the model’s predictions and generalizations on a wide range of physical scenarios from the Physion dataset and found that it captured both participants’ successes and error patterns.

From pixels to physics: an image-computable model of physical predictions

The ability to detect a partner’s trustworthiness and adjust one’s own trust decisions accordingly is crucial and adaptive for maintaining cooperative relationships. Trusting a trustworthy partner maximizes mutual benefits; withholding trust from an untrustworthy partner minimizes chances of being exploited. We sought to understand how young children learn through experience and adjust their own trust behaviors when they interact with trustworthy and untrustworthy individuals.
In this study, N = 96 6 to 11 year olds played 40 trials of repeated Trust Game with a trustworthy and an untrustworthy partner (20 trials each, order randomized). Results showed that although children across all age groups correctly identified the trustworthiness of their partners post-game, surprisingly, they did not trust the trustworthy partner more than chance level, nor did they show increasing differentiation between the two partners across the trials. These findings suggest a knowledge-behavior gap in children’s trust interaction with novel partners.

Downloads

Next from CogSci 2025

Copilots for Cognitive Linguists 2025

Similar lecture

The Forest for the Trees: Global vs. Local Advice in Human-AI Interaction

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES