Wearable cameras provide an efficient and convenient way to record our lives, supporting real-time documentation and analysis across various domains. Recent research has explored diverse methods for temporal segmentation, which aim to transform unstructured video data into structured events. This transformation facilitates deeper video understanding, optimizes computational resources, and improves the accessibility and interpretability of video content for both machines and humans. However, unlike conventional videos, wearable camera recordings present unique challenges. These include highly unstable camera perspectives, diverse activities across various environments, and flexible duration. As a result, traditional temporal segmentation methods often fail to return effective results. This paper introduces SmartSeg, an unsupervised, non-parametric approach for segmenting wearable camera videos without labeled data. By capturing the fundamental meanings of the video, SmartSeg aggregates the video through the Temporal Self-Similarity Metric encoder and groups sequences of frames into coherent events through clustering techniques. We evaluated SmartSeg on three diverse datasets. We achieved a 50% increase in Mean-over-Frames(MoF) compared to the state-of-the-art on one egocentric dataset. We conducted a real-world case study on nursing simulations, demonstrating SmartSeg’s ability to effectively segment complex, noisy interactions with diverse activity transitions. The results highlight SmartSeg’s robustness in handling long, unstructured, and visually challenging wearable camera videos, establishing it as a promising tool for real-world video temporal segmentation tasks.

Reading code is a fundamental activity in both software engineering and computer science education. Understanding the cognitive processes involved in reading code is crucial for identifying effective cognitive strategies, which can inform teaching methods and tooling support for developers. However, collecting large human subject eye tracking datasets, especially for programming tasks, is often costly and time-consuming, limiting its scalability and applicability. To address this issue, we present CodeACT-R, the first cognitive simulation framework tailored for code reading, based on the well-established Adaptive Control of Thought—Rational (ACT-R) architecture from cognitive science. CodeACT-R simulates how humans read code and requires only a small, manageable amount of human data to initiate the simulator design, offering a cost-effective and scalable alternative to traditional data collection methods like eye tracking.Specifically, we first collected real human visual attention data from 48 programmers reading code using eye tracking. These data were then used to develop CodeACT-R, enabling the simulation of human-like code reading behaviors. Our evaluation demonstrates that CodeACT-R is capable of simulating visual attention patterns (i.e., scanpaths) that closely resemble real-world human attention patterns, also accounting for up to 87% of observed pattern variations.

Eye-tracking technology has expanded our ability to quantitatively measure human perception. This rich data source has been widely used to characterize human behavior and cognition. However, eye-tracking analysis has been limited in its applicability, as contextualizing gaze to environmental artifacts is non-trivial. Moreover, the temporal evolution of gaze behavior through open-ended environments where learners are alternating between tasks often remains unclear. In this paper, we propose temporal fully-weighted scanpath graphs as a novel representation of gaze behavior and combine it with a clustering scheme to obtain high-level gaze summaries that can be mapped to cognitive tasks via network metrics and cluster mean graphs. In a case study with nurse simulation-based team training, our approach was able to explain changes in gaze behavior with respect to key events during the simulation. By identifying cognitive tasks via gaze behavior, learners’ strategies can be evaluated to create online performance metrics and personalized feedback.

Just as our interactions with other people are shaped by our concepts about their beliefs, desires, and goals (i.e., "theory of mind"), our interactions with intelligent technologies such as robots are shaped by our concepts about their internal operations. Multiple studies have demonstrated that people attribute anthropomorphic features to technological agents in certain contexts, but researchers remain divided on how these attributions arise: What default assumptions do people make about the internal operations of intelligent technology, and what events or additional information cause us to alter those default assumptions? This article explores these open questions and some of their implications for law and policy. First, we review psychological research exploring people's attributions of agency, with particular focus on attributions to technological entities. Next, we define and describe one popular account of this research---a "promiscuous agency" account that assumes a reflexive tendency to broadly attribute humanlike properties to technological agents. We then summarize mounting evidence that people are often more cautious in attributing human properties than the promiscuous agency account suggests. We seek to integrate the mounting evidence for a "selective agency" account with the promiscuous agency account through the transition model of agency. Finally, we explore how selective agency, promiscuous agency, and the transition model relate to a sample of robotics law and policy issues. We address, in turn, issues related to Fourth Amendment protection, copyright law, statutory and regulatory interpretation, and negligence litigation, identifying specific implications of the transition model of agency for each issue.

This research investigates adults' understanding of differences in the basic nature of intelligence exhibited by humans and by machines such as computers and robots. We tested these intuitions by asking participants to make predictions about the behaviors of different entities in situations where actions could be based on either goal-directed intentional thought or more mechanical nonintentional thought. Across several studies, adults made more intentional predictions about the behavior of humans than about the behavior of robots or computers. Although initial experiments demonstrated that participants made very similar predictions for computers and anthropomorphic robots, when asked to track robots' attention to objects, participants began to predict more intentional behaviors for the robot. A multiple regression demonstrated that differential behavioral predictions about mechanical and human entities were associated with ratings of goal understanding but not overall intelligence of current computers/rob...

When people interact with intelligent agents, they likely rely upon a wide range of existing knowledge about machines, minds, and intelligence. This knowledge not only guides these interactions, but it can be challenged and potentially changed by interaction experiences. We hypothesized that a key factor mediating conceptual change in response to human-machine interactions is cognitive conflict, or dissonance. In this experiment, we evaluated whether interactions with a robot partner during a realistic medical triage scenario caused increased levels of cognitive dissonance relative to a control condition in which the same task was performed with a human partner. In addition, we evaluated whether heightened levels of dissonance affected concepts about agents. We observed increased cognitive dissonance after the human-robot interaction and found that this dissonance was correlated with a significantly less intentional (e.g., less human-like) view of the intelligence inherent to computers.

Although a number of recent studies have explored people's concepts about robots, almost no research has tested the degree to which these concepts affect people's capacity to understand and remember a robot's actions. In this study, we tested whether a narrative describing a robot performing basic intentional acts would be easier to remember than a narrative that described similar non-intentional actions. Participants read one of two stories about a robot in which it was either described as having intentional or non-intentional mental representations. Participants who read about the intentional robot were more likely to recall information about the robotic agent, but there was no difference between the two groups in accuracy for questions unrelated to the agent. Additionally, participants who read about the intentional robot were marginally more likely to falsely recall a non-present object that was similar to the objects that the robot did interact with. We conclude that beliefs about a robot affect encoding and recall of its actions, possibly due to a focus on the type of information the agent is believed to "mentally" represent.

Computerized learning environments often include implementations of simulated social agents, incorporating the reasonable design assumption that learning interactions that are more social are more effective. However, recent evidence suggests that computerized social agents can in some circumstances fail to promote or even hinder learning. The current paper outlines evidence both supporting and arguing against the utility of computerized social agents for learning. We propose a framework reconciling this evidence by delineating impacts from a social agent that impinge upon different points of learning, from shallow to deep phases. Shallow social impacts when ineffectual carry little to no potential to actively impede learning, but any potential positive impacts on shallow learning phases are relatively limited in the absence of positive deep impacts. Deep social impacts, on the other hand, carry the potential to strongly drive deep learning, but when ineffectual carry the risk to impede it. The paper concludes with a proposal for future research based on this framework.

In previous research, we demonstrated that people distinguish between human and nonhuman intelligence by assuming that humans are more likely to engage in intentional goal-directed behaviors than computers or robots. In the present study, we tested whether participants who respond relatively quickly when making predictions about an entity are more or less likely to distinguish between human and nonhuman agents on the dimension of intentionality. Participants responded to a series of five scenarios in which they chose between intentional and nonintentional actions for a human, a computer, and a robot. Results indicated that participants who chose quickly were more likely to distinguish human and nonhuman agents than participants who deliberated more over their responses. We suggest that the short-RT participants were employing a first-line default to distinguish between human intentionality and more mechanical nonhuman behavior, and that the slower, more deliberative participants engaged in deeper second-line reasoning that led them to change their predictions for the behavior of a human agent.

We have previously demonstrated that people apply fundamentally different concepts to mechanical agents and human agents, assuming that mechanical agents engage in more location-based, and feature-based behaviors whereas humans engage in more goal-based, and category-based behavior. We also found that attributions about anthropomorphic agents such as robots are very similar to those about computers, unless subjects are asked to attend closely to specific intentional-appearing behaviors. In the present studies, we ask whether subjects initially do not attribute intentionality to robots because they believe that temporary limits in current technology preclude real intelligent behavior. In addition, we ask whether a basic categorization as an artifact affords lessened attributions of intentionality. We find that subjects assume that robots created with future technology may become more intentional, but will not be fully equivalent to humans, and that even a fully human-controlled robot will not be as intentional as a human. These results suggest that subjects strongly distinguish intelligent agents based on intentionality, and that the basic living/mechanical distinction is powerful enough, even in adults, to make it difficult for adults to assent to the possibility that mechanical things can be fully intentional.

Because computers often rely on visual displays as a way to convey information to a user, recent research suggesting that people have detailed awareness of only a small subset of the visual environment has important implications for human-computer interface design. Equally important to basic limits of awareness is the fact that people often over-predict what they will see and become aware of. Together, basic failures of awareness and people's failure to intuitively understand them may account for situations where computer users fail to obtain critical information from a display even when the designer intended to make the information highly visible and easy to apprehend. To minimize the deleterious effects of failures of awareness, it is important for users and especially designers to be mindful of the circumscribed nature of visual awareness. In this article, we review basic and applied research documenting failures of visual awareness and the related metacognitive failure and then discuss misplaced beliefs that could accentuate both in the context of the human-computer interface.