United States

Most machine learning based decision support systems are black box models that are not interpretable for humans. However, the demand for explainable models to create comprehensible and trustworthy systems is growing, particularly in complex domains involving risky decisions. In many domains, decision making is based on visual information. We argue that nevertheless, explanations need to be verbal to communicate the relevance of specific feature values and critical relations for a classification decision. To address that claim, we introduce a fictitious visual domain from archeology where aerial views of ancient grave sites must be classified. Trustworthiness among other factors relies on the perceived or assumed correctness of a system&#39;s decisions. Models learned by induction of data, in general, cannot have perfect predictive accuracy and one can assume that unexplained erroneous system decisions might reduce trust. In a 2×2 factorial online experiment with 190 participants, we investigated the effect of verbal explanations and information about system errors. Our results show that explanations increase comprehension of the factors on which classification of grave sites is based and that explanations increase the joint performance of human and system for new decision tasks. Furthermore, explanations result in more confidence in decision making and higher trust in the system.

CogSci 2021

Explaining Machine Learned Relational Concepts in Visual Domains - Effects of Perceived Accuracy on Joint Performance and Trust

### Welcome to the 8th annual workshop on Naval Applications of Machine Learning!

NAML is the premier annual event showcasing current machine learning research relevant to Naval applications. We welcome participation from the defense community, industry, and academia. The NAML workshop is an opportunity for researchers from Navy research and development centers and our partner organizations to exchange technical ideas and to come together as a team.

The goals of NAML are:

* To spread awareness of current machine learning research relevant to Naval applications

* To connect machine learning researchers with experts in military requirements

* To build and strengthen collaborations within the defense research community

Naval Information Warfare Center Pacific is partnering with AFCEA San Diego chapter to host the workshop.

**General Chairs:**

Katie Rainey, Naval Information Warfare Center Pacific

Nicole Isoda, Naval Information Warfare Center Pacific<br><br>

**[More about NAML 2024](https://sites.google.com/go.spawar.navy.mil/naml/)**
<br><br>

Please log in to explore this event. You need to log in with the email address you registered with. Access credentials have been sent to your email. 

Please be sure to check your spam and other email folders if you do not see an email confirmation right away.

Please log in to explore this workshop.

Registration for this workshop has now closed. We appreciate your interest and hope to welcome you at a future event. If you’d like to hear about similar upcoming events, please check our website or join our mailing list.

Registration for this workshop has now closed.

NAML 2024

The 8th annual workshop on Naval Applications of Machine Learning is an opportunity for researchers from Navy research and development centers and our partner organizations to exchange technical ideas and to come together as a team. The goals of NAML are to spread awareness of current machine learning research relevant to Naval applications, to connect machine learning researchers with experts in military requirements and to build and strengthen collaborations within the defense research community.


Most machine learning based decision support systems are black box models that are not interpretable for humans. However, the demand for explainable models to create comprehensible and trustworthy systems is growing, particularly in complex domains involving risky decisions. In many domains, decision making is based on visual information. We argue that nevertheless, explanations need to be verbal to communicate the relevance of specific feature values and critical relations for a classification decision. To address that claim, we introduce a fictitious visual domain from archeology where aerial views of ancient grave sites must be classified. Trustworthiness among other factors relies on the perceived or assumed correctness of a system's decisions. Models learned by induction of data, in general, cannot have perfect predictive accuracy and one can assume that unexplained erroneous system decisions might reduce trust. In a 2×2 factorial online experiment with 190 participants, we investigated the effect of verbal explanations and information about system errors. Our results show that explanations increase comprehension of the factors on which classification of grave sites is based and that explanations increase the joint performance of human and system for new decision tasks. Furthermore, explanations result in more confidence in decision making and higher trust in the system.

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.

Lay perceptions of medical conditions and treatments determine people’s health behaviors, guide biomedical research funding, and have important consequences for both individual and societal wellbeing. Yet it has been nearly impossible to quantitatively predict lay health perceptions for hundreds of everyday diseases due to the myriad psychological forces governing health-related attitudes and beliefs. Here we present a data-driven approach that uses text explanations on healthcare websites, combined with large-scale survey data, to train a machine learning model capable of predicting lay health perception. We use our model to analyze how language influences health perceptions, interpret the psychological underpinnings of health judgment, and quantify differences between different descriptions of disease states. Our model is accurate, cost-effective, and scalable, and offers researchers and practitioners a new tool for studying health-related attitudes and beliefs.

Machine Learning Models for Predicting, Understanding, and Influencing Health Perception

Previous studies have suggested testing and feature highlighting each facilitates category learning. We investigated whether the beneficial effects of interim testing change with the presence of feature highlighting in natural category learning. Participants learned various rock categories that were divided into two sections. They studied a series of rocks with or without feature descriptions highlighted on each image and either took an interim test or not between the two sections. On the final test, participants classified studied and new rock exemplars from both sections. The testing group outperformed the no-testing group for both sections, indicating both backward and forward effects of testing. Such beneficial effects of testing occurred regardless of whether the feature highlighting was provided or not. The feature highlighting, however, showed negative effects on learning of both sections, suggesting that provision of explicit instruction may impede learning when it is not appropriately embedded in the learning material.

Effects of interim testing and feature highlighting on natural category learning

A key element of conceptual understanding in physics education is the ability to make qualitative inferences. In this study, we investigated whether people use causal notions when solving simple, qualitative physics problems. We utilized the principles of directionality and holding variables constant to assess participant's accuracy in answering questions about inferences from cause to effect (CE), effect to cause (EC), and from one cause to the other cause (CC). Participants responded with greater accuracy for CE and EC questions when alternative causal information was held constant compared to when it is made explicitly unknown. For CC questions accuracy was lower across all alternative information types. Further, participants generally treated ambiguous alternative causes as if they were held explicitly constant. These results indicate that using notions of causality can potentially help identify difficult qualitative physics questions and be used as a tool in instructional design.

Using Causality to Map Difficulties in a Qualitative Physics Problem

A set of fundamental beliefs governs our reasoning about objects and agents since infancy. Studies have shown that infants and children show enhanced exploration and learning when they observe apparent violations of these beliefs. However, little is known about whether these beliefs can be revised given counterevidence. In the present experiments, we demonstrate that 4- to 6-year-old children can revise their most fundamental beliefs in the physical domain (Experiment 1) and the psychological domain (Experiment 2) when they observe multiple pieces of belief-violating evidence.

Revising Core Beliefs in Young Children

We often organize objects around both visual and semantic boundaries in space. Across four experiments, we examined how semantically consistent partitions influenced memory for object locations. Participants learned the locations of items in a semantically partitioned display (where each partition contained objects from a single category), as well as a purely visually partitioned display (where each partition contained a random assortment of objects from different categories). While semantic partitions significantly improved location memory over the purely visually partitioned display, this advantage was significantly reduced when participants were cued to the correct partition during recall. Our results suggest that semantic category information benefits memory via strengthening the association between a given category and a spatial region delineated by a partition. Further, there was some indication that this benefit may come with the drawback of reducing memory precision for objects within a partitioned space.

The effect of semantic categorization on object location memory

Persuasion is hard. Then why do some consumers think that marketers can easily manipulate them? Three studies and an internal meta-analysis suggest that consumers’ beliefs about marketing manipulation are rooted in humans’ deeper and older psychology—motivation to understand life events. Consumers higher in motivation to make sense of their environments not only detect persuasion where it exists but also where none (make false-positive errors). Whereas consumers higher in sense-making motives believe that manipulations are more effective, objective sense-making abilities negatively predict false manipulation detection. We also tested how manipulation beliefs are related to personality traits, gender, and age. These findings help 1) understand the origins of lay (false) beliefs about the marketplace and persuasion in general, 2) consider marketing segmentation strategies to reach out to particularly skeptical consumers, and 3) understand ways to attenuate false-positive beliefs and foster accurate persuasion detection, which is critical in the era of infodemic.

Lay Theories of Manipulation: Do Consumers Believe They are Susceptible to Marketers’ Trickery?

The mind readily learns cue-outcome associations where an object predicts a specific outcome. Previous work suggested that when multiple objects associated with different outcomes were jointly presented , the mind made conjunctive predictions that represented the common property of the associated outcomes. Using attentional tracking measures, we provided more evidence for the weighted summation framework when the conjunctive predictions involved spatial locations (Experiment 1) or conceptual categories (Experiment 2). Then, we examined the reverse of conjunction, where participants were presented with a single object, which is a part of an object pair that was previously associated with an outcome (Experiment 3). Rather than making predictions based on mental operations such as subtraction, we found that participants’ predictions were purely based on previous associations. These results together demonstrated the robust tendency to make conjunctive predictions based on knowledge of cue-outcome associations.

Association knowledge guides conjunctive predictions in novel situations

In this study, we explicitly code and study the social, referential, and pragmatic features of gaze in human-human spontaneous dyadic interaction, providing novel observations that can be executed in a machine in order to improve multimodal human-agent dialogue. Gaze is an important non-verbal social signal that contains attentional cues about where to look and provides information about others' intentions and future actions. In this work, various types of gaze behaviour are annotated in detail along with speech to explore the meaning of temporal patterns in gaze cues and their co-relations. Considering that 80% of the total stimuli perceived by the brain is visual, gaze behaviour is complex and challenging; hence, implementing human-human gaze cues to an avatar/robots could improve human-agent interaction and make it more natural.

A Deep Gaze into Social and Referential Interaction

In categorization tasks optimization of performance depends on attention to relevant stimulus features. However, studies show that adults and children seem to use different strategies when learning new categories. Deng & Sloutsky (2016) found that adults attend selectively, whereas children prefer to attend diffusely, even when the stimuli possess deterministic features (i.e. with 100% probability of belonging to a certain category). Most studies of attention in category learning rely on behavioral effects to infer attentional strategy, despite the availability of eye-tracking technology. In this study, we combine the use of category training and transfer paradigms (see Miles & Minda, 2009 and Deng & Sloutsky, 2016) and eye-tracking methods to investigate attentional strategies of children and adults during category learning. Preliminary results of our adult pilot (N = 14) confirm the prediction that adults optimize performance by attending increasingly to deterministic features of the stimulus once the categorical rule is found.

Attentional strategies during category learning: an eye-tracking study

Despite recent progress in promoting gender equality, gender bias continues to post challenges to women's career advancement. Here, we use a statistically grounded framework to investigate how face-based social perception may contribute to gender biases in political and job application settings. By analyzing a large face dataset and performing a novel behavioral experiment, we find that: 1) female faces exhibit stronger anti-correlation between perceived trustworthiness and dominance, 2) this anti-correlation is due to distinct sets of facial features humans utilize to assess female and male faces for trustworthiness and dominance, 3) perceived dominance positively contributes to social decision preferences for female faces, contrary to prior suggestions that perceived dominance affects female candidates negatively, and 4) the anti-correlated perception of trustworthiness and dominance put females at a disadvantage in competitive environments. More generally, our findings reveal the important role of face-based trait perceptions underlying gender biases in social decision making.

Downloads

Next from CogSci 2021

Machine Learning Models for Predicting, Understanding, and Influencing Health Perception

Similar lecture

Interpretable Decision-Making By Case-Based Reasoning: Lessons from a Human Subjects Study

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES