Search

Usability—the extent to which a system can be operated to achieve specified goals with effectiveness, efficiency, and satisfaction—is a hallmark of many successful systems. Identifying usability concerns with existing measures, however, can be problematic. We use biased competition theory (BCT) to explain how interfaces with low usability create attentional interference, which can be measured through mouse movement efficiency (MME). We tested our hypotheses in four studies. Study 1 manipulates attentional interference and examines its influence on MME using eye tracking to validate BCT as an appropriate theoretical lens for our research. Study 2 manipulates usability and shows similar markers of attentional interference in MME on a website with two different populations. Study 3 is a field test that compares the reliability of MME in ranking the usability of different components of a commercial web application to rankings produced by a common perceptual measure of usability: perceived ease of use. Collaborating with a Fortune 100 software vendor, Study 4 further demonstrates the efficacy of MME in ranking the usability of various system components compared to a standard industry measure of usability across three different products. The results consistently show that MME can help identify usability differences among parts of systems. The results have implications for facilitating more efficient and temporally precise usability research and mass-deployable usability testing. [ABSTRACT FROM AUTHOR]

Customer care is important for its role in relationship building. This role has traditionally been performed by human customer agents; however, the emergence of interactive generative AI (GenAI) shows potential for using AI for customer care in emotionally charged interactions. Bridging practice and the academic literatures in marketing and computer science, this article develops an AI-enabled customer care journey, from accurate emotion recognition to empathetic response, emotional management support, and, finally, the establishment of an emotional connection. Marketing requirements for each of the stages are derived from in-depth interviews with top managers and a survey of chief marketing officers. By juxtaposing these requirements against the current feeling capabilities of GenAI, the authors highlight the technological challenges engineers must tackle. The article concludes with a set of marketing tenets for implementing and researching the caring machine. These include verifying emotion recognition accuracy using marketing emotion theories through multiple emotion signals and methods, utilizing prompt engineering to enhance GenAI's emotion understanding, employing "response engineering" to personalize emotion management recommendations, and strategically deploying GenAI for emotional connection to simultaneously enhance customer emotional well-being and customer lifetime value. [ABSTRACT FROM AUTHOR]

Previous research has shown that consumers respond differently to decisions made by humans versus algorithms. Many tasks, however, are not performed by humans anymore but entirely by algorithms. In fact, consumers increasingly encounter algorithm-controlled products, such as robotic vacuum cleaners or smart refrigerators, which are steered by different types of algorithms. Building on insights from computer science and consumer research on algorithm perception, this research investigates how consumers respond to different types of algorithms within these products. This research compares high-adaptivity algorithms, which can learn and adapt, versus low-adaptivity algorithms, which are entirely pre-programmed, and explore their impact on consumers' product preferences. Six empirical studies show that, in general, consumers prefer products with high-adaptivity algorithms. However, this preference depends on the desired level of product outcome range—the number of solutions a product is expected to provide within a task or across tasks. The findings also demonstrate that perceived algorithm creativity and predictability drive the observed effects. This research highlights the distinctive role of algorithm types in the perception of consumer goods and reveals the consequences of unveiling the mind of the machine to consumers. [ABSTRACT FROM AUTHOR]

This research explores prospective determinants of trust in the recommendations of artificial agents regarding decisions to kill, using a novel visual challenge paradigm simulating threat-identification (enemy combatants vs. civilians) under uncertainty. In Experiment 1, we compared trust in the advice of a physically embodied versus screen-mediated anthropomorphic robot, observing no effects of embodiment; in Experiment 2, we manipulated the relative anthropomorphism of virtual robots, observing modestly greater trust in the most anthropomorphic agent relative to the least. Across studies, when any version of the agent randomly disagreed, participants reversed their threat-identifications and decisions to kill in the majority of cases, substantially degrading their initial performance. Participants subjective confidence in their decisions tracked whether the agent (dis)agreed, while both decision-reversals and confidence were moderated by appraisals of the agents intelligence. The overall findings indicate a strong propensity to overtrust unreliable AI in life-or-death decisions made under uncertainty.

Artificial intelligence (AI) has the potential to improve human decision-making by providing decision recommendations and problem-relevant information to assist human decision-makers. However, the full realization of the potential of human-AI collaboration continues to face several challenges. First, the conditions that support complementarity (i.e., situations in which the performance of a human with AI assistance exceeds the performance of an unassisted human or the AI in isolation) must be understood. This task requires humans to be able to recognize situations in which the AI should be leveraged and to develop new AI systems that can learn to complement the human decision-maker. Second, human mental models of the AI, which contain both expectations of the AI and reliance strategies, must be accurately assessed. Third, the effects of different design choices for human-AI interaction must be understood, including both the timing of AI assistance and the amount of model information that should be presented to the human decision-maker to avoid cognitive overload and ineffective reliance strategies. In response to each of these three challenges, we present an interdisciplinary perspective based on recent empirical and theoretical findings and discuss new research directions.

This study compares how English-speaking adults and children from the United States adapt their speech when talking to a real person and a smart speaker (Amazon Alexa) in a psycholinguistic experiment. Overall, participants produced more effortful speech when talking to a device (longer duration and higher pitch). These differences also varied by age: children produced even higher pitch in device-directed speech, suggesting a stronger expectation to be misunderstood by the system. In support of this, we see that after a staged recognition error by the device, children increased pitch even more. Furthermore, both adults and children displayed the same degree of variation in their responses for whether Alexa seems like a real person or not, further indicating that childrens conceptualization of the systems competence shaped their register adjustments, rather than an increased anthropomorphism response. This work speaks to models on the mechanisms underlying speech production, and human-computer interaction frameworks, providing support for routinized theories of spoken interaction with technology.

In recent years, the field of computer graphics has achieved its longstanding dream of photorealism: modern graphics algorithms produce images that are indistinguishable from reality. Much like art at the advent of photography, then, computer graphics is now turning its gaze to the beholder: researchers are increasingly looking to cognitive science to engineer new modes of visual expression. Recent work has sought to apply insights from cognitive science to a variety of traditional graphics topics: from taking a perceptual approach to perspective, to studying the theory of mind behind animation, to applying theories of abstraction learning to build tools for geometry processing. At the same time, a wave of recent work in cognitive science has addressed fundamental questions about visual expression: for example, how humans understand and create sketches, shapes, and symbols. The field has also benefited greatly from tools and methods from computer graphics: differentiable rendering, physics simulation, and game engines have become indispensable in modeling human perception and intuitive physics. Recognizing this growing interdisciplinary exchange of ideas, we are proposing a workshop to begin building formal bridges between the cognitive science and computer graphics communities.

Emotions, including stress and anxiety, strongly influence cognition and learning experiences. This study investigates the impacts of stress on cognitive load during learning, considering baseline anxiety levels and fluctuating stress. With a focus on technology-based learning, a web-based HTML introduction module was used. Using a social stress test, 15 participants underwent a stressful situation during learning, while the control group of 15 were in a neutral condition. Results indicate significantly elevated stress levels in the experimental group throughout the experiment, with a corresponding decrease in learning performance. For high perceived difficulty, the stressed condition demonstrated a significant increase in response time compared to the control condition. In contrast, when experiencing low perceived difficulty, a significant difference in response time across conditions was not found. Findings emphasise the importance of managing stress in educational contexts to optimise learning outcomes in the evolving landscape of technology-based learning.

Music-making relies on precise temporal control and mutual coordination among performers, particularly to maintain tempo. We evaluate the impact of human-machine interaction and rhythmic subdivisions on tempo change in musical trios. A synchronization-continuation task was performed by trios of human participants interacting with confederates or with algorithmic (i.e. machine) models. Sounded tone onsets were produced by a linear error-correction model, delay-coupled model, and Kuramoto model that replaced a human participant. Inter-onset intervals were examined from participants who performed rhythms in both in-phase and anti-phase conditions while a third group member was either a human or algorithmic model. Trios drifted toward faster tempi more when they contained a human than an algorithmic model. Tempo drift also increased for the aligned rhythms (in-phase) compared to rhythms with rhythmic subdivisions (anti-phase). Finally, the tested algorithmic models replicated the confederate's tempo drift without the use of any period correction mechanisms. This research advances our understanding of unintentional tempo drift, offering insights into ensemble dynamics and models of temporal coordination in groups. Implications for musical coordination and avenues for future research are discussed.

Content-curating algorithms provide a crucial service for social media users by surfacing relevant content, but they can also bring about harms when their objectives are misaligned with user values and welfare. Yet, potential behavioral consequences of this alignment problem remain understudied in controlled experiments. In a preregistered, two-wave, collaborative filtering experiment, we demonstrate that small changes to the metrics used for sampling and ranking posts affect the beliefs people form. Our results show observable differences in two types of outcomes within statisticized groups: belief accuracy and consensus. We find partial support for hypotheses that the recently proposed approaches of "bridging-based ranking" and "intelligence-based ranking" promote consensus and belief accuracy, respectively. We also find that while personalized, engagement-based ranking promotes posts that participants perceive favorably, it simultaneously leads those participants to form more polarized and less accurate beliefs than any of the other algorithms considered.

Predicting Click-Through Rate (CTR) is crucial in product and content recommendation, as it involves estimating the likelihood of a user engaging with a specific advertisement or content link. This task encompasses understanding the complex cognitive processes behind human interactions with recommended content. Learning varied feature embeddings that reflect different cognitive responses in various circumstances is significantly important. However, traditional methods typically learn fixed feature representations, leading to suboptimal performance. Some recent approaches attempt to address this issue by learning bit-wise weights or augmented embeddings for feature representations, but suffer from uninformative or redundant features in the context. To tackle this problem, inspired by the Global Workspace Theory in conscious processing, which posits that only a specific subset of the product features are pertinent while the rest can be noisy and even detrimental to human-click behaviors, we propose a CTR model that enables Dynamic Embedding Learning with Truncated Conscious Attention for CTR prediction, termed DELTA. DELTA contains two key components:(I) conscious truncation module (CTM), which utilizes curriculum learning to apply adaptive truncation on attention weights to select the most critical feature in the context;(II) explicit embedding optimization (EEO), which applies an auxiliary task during training that directly and independently propagates the gradient from the loss layer to the embedding layer, thereby optimizing the embedding explicitly via linear feature crossing. Extensive experiments on five challenging CTR datasets demonstrate that DELTA achieves new state-of-the-art performance among current CTR methods.

The study of mind perception, particularly how one perceives the mental states of `others,' has attracted considerable interest in cognitive science. The present study contributes to the investigation of mind perception in a human-robot interaction context, by testing a humanoid robot and a human and their communicative and noncommunicative actions. We examine mind perception across its two primary dimensions: Agency and Experience and in their High and Low ends. The novelty of our study lies in its real-time and implicit nature---both identified as crucial elements in current debates within the field. Our results indicate that testing physically present and active agents, as well as exposing participants to various types of live actions, influences mental capacity attributions across different capacities. Additionally, the integration of behavioral measurements alongside verbal data holds promise for a detailed interpretation of the mind perception process.

Saliency-based Explainable AI (XAI) methods have been commonly used for explaining computer vision models, but whether they could indeed enhance user understanding at different levels remains unclear. We showed that for object detection AI, presenting users with AI 's output for a given input was sufficient for improving feature-level and some instance-level user understanding, particularly for false alarms, and providing saliency-based explanations did not have additional benefit. This was in contrast to previous research on image classification models where such explanations enhanced understanding. Analyses with human attention maps suggested that humans already attended to features important for AI's output in object detection and thus could infer AI's decision-making processes without saliency-based explanations. However, it did not enhance users' ability to distinguish AI's misses and hits, or system-level understanding. Therefore, the effectiveness of saliency-based explanations is task-dependent, and alternative XAI methods are required for object detection models to better enhance understanding.

Self-regulated learning influences students' learning behaviors and is a significant academic performance factor. Resource management strategy based on self-regulated learning theory is an important indicator for students to demonstrate Self-regulated learning. However, current self-regulated learning and resource management strategy assessments still rely on subjective evaluations and self-assessments, which are time-consuming and laborious. Therefore, we propose a novel method combined with mobile sensing by collecting detailed learning strategy subscales and objective mobile sensing data from 211 college students to explore a new approach to assessing self-regulated learning and resource management strategy. We are the first to propose a mobile sensing approach for assessing self-regulated learning and learning strategies. The method studies the associations between the learning strategy subscales and these daily behavior patterns and presents features for behavior patterns from mobile sensing data. Our study helps to reveal new forms of assessing self-regulated learning and opens the way for personalized interventions.

This study aimed to test the benefit of note-taking on pauses and learning in online educational videos. Participants (N = 72) were randomly assigned to one of two note-taking conditions (allowed or not) with the possibility to take pauses during a 10-minute online instructional video on the autonomic nervous system. The results did not reveal a significant correlation between note-taking and the number of pauses. Moreover, we observed no significant effect of note-taking on learning performance. However, prior knowledge and age affected significantly the relationship between pauses, note-taking and learning performance. We discuss the importance of prior knowledge and age for future research.

In teleoperation, the spatiotemporal calibration of the system can significantly impact both performance and user experience, which may not necessarily be causally linked. This study asks if Sense of Embodiment (SoE) varies with spatiotemporal calibration of a teleoperated system, which in turn affects task performance. Most SoE studies are passive and they do not represent a great paradigm to study the impact of calibration on SoE in active teleoperation. Therefore, we designed an active RHI in mixed reality where we manipulated both the spatial calibration (shifts) and visuo-proprioceptive synchronicity (temporal delay). We investigated if this manipulation affected performance, proprioceptive mapping, SoE and the perception of the setup as a mediator. The results suggest a potential direct influence of SoE on task performance, particularly through enhanced calibration due to synchronicity, indicating potential benefits for sustained usage. Additionally, SoE is explored comprehensively, employing multiple tests assessing implicit and explicit dimensions of calibration.

The proliferation of artificial intelligence (AI) agents has introduced a new dynamic into the human social environment. This study investigates prosocial behavior in a hybrid human-AI setting, particularly within a gaming environment. Many existing studies on prosocial behavior are conducted in economic game settings in which the agents' intentions, and whether or not prosocial actions offer benefits, are explicit. This project explores prosocial interactions in spatial environments where the need for help by another agent might not be immediately obvious, and where cognitive processes such as attention, and decision-making processes about the cost of helping are thus likely to play a role. In a baseline study (N = 177), we investigated the likelihood of human agents reciprocating prosocial behavior initiated by an AI player. Results indicated that the low saliency of the AI player's actions was a primary reason for non-reciprocation. A follow-up study (N = 164) tested whether increasing the salience of the AI's actions would enhance human prosocial responses. We found support for our hypothesis from analysis of the time-series data and participants' self-reported post-game questionnaires. This research contributes to the growing field of human-AI cooperation, outlining a vision for a future where technology actively contributes to our collective well-being, and opening up new possibilities for positive transformation in a world increasingly populated by intelligent machines.

In this paper, we developed three ACT-R cognitive models to simulate the learning process of abacus gestures. Abacus gestures are mid-air gestures, each representing a number between 0 and 99. Our models learn to predict the response time of making an abacus gesture. We found the accuracy of a model's predictions depends on the structure of its declarative memory. A model with 100 chunks cannot simulate human response, whereas models using fewer chunks can, as segmenting chunks increase both the frequency and recency of information retrieval. Furthermore, our findings suggest that the mind is more likely to represent abacus gestures by dividing attention between two hands rather than memorizing and outputting all gestures directly. These insights have important implications for future research in cognitive science and human-computer interaction, particularly in developing vision and motor modules for mental states in existing cognitive architectures and designing intuitive and efficient mid-air gesture interfaces.

Despite computational algorithms outperforming humans in certain tasks, algorithmic advice is less used than human advice (algorithm aversion). Thus, algorithmic advice should be designed to avoid algorithm aversion. However, few studies have discussed the use of advice with an interval (e.g., 60.0 ± 2.0 %), a common format in algorithmic advice. This study confirmed in two behavioral experiments (N = 200) that differences in advice sources lead to differences in advice use, mainly by influencing the step at which the judge decides whether to ignore the advice. Therefore, we proposed to individualize the presentation of advice so that the advice would be such that decreases the rate advice being ignored. Our individualization of the advice presentation focused on the distance between the advice and the initial judgment, a significant factor in advice utilization. Another behavioral experiment (N = 100) confirmed that our proposed advice design overcomes differences among advisors.

Pragmatic implicature derivation presupposes that the cooperative principle is observed and critically depends on interlocutors expecting each other to behave cooperatively. It is much less clear, however, whether people extend this assumption to communication with artificial agents. People might therefore not draw the same pragmatic inferences when interacting with an artificial agent as they would with other conversationally competent humans, even if the agent is in principle believed to be similarly competent. In our study, we ask participants to interpret messages in a pragmatic reference game which they are told were generated by ChatGPT. Additionally, participants report whether they believe ChatGPT to be capable of the reasoning needed to select the optimal message. We observe a noteworthy discrepancy: in the reference game, participants interpret ChatGPT's messages less pragmatically than those of another adult human, but in the post-test questionnaire, they overwhelmingly rate ChatGPT's pragmatic ability very highly.

Transcription with Incremental Presentation of the Stimulus (TIPS) is a novel approach relying on micro-behaviours proposed by Colarusso and colleagues (2023) to study users' cognition with data visualizations. The study in this paper has two primary objectives: (a) investigate whether TIPS can measure an individual's competence with data visualizations; and (b) explore the potential enhancement of TIPS measures by normalizing them with the individual's performance on tests of visuo-spatial abilities and memory capacity. We test 30 participants with different expertise and cognitive skills. Results reveal that TIPS provides some promise for individual competence assessment, but only when normalized with the individual's performance on a test of rigid transformation of mental images. Other tests measuring visuospatial abilities or memory capacity did not produce effective normalizations.

Intelligent systems that record, analyze, and respond to events have become major parts of our lives. They are available as Decision Support (DS) for many tasks and can enhance the information on which decision-makers can base their decisions. Decision makers need to evaluate the available information, and they also have to decide whether to seek information from additional information sources. The information is often costly, and its costs and benefits must be weighted. Also, integrating information from multiple sources can complicate the decision task. Here, we study the combined decision process that chooses information sources and integrates them, if chosen, in a classification decision.In an online experiment with 75 engineering students, we manipulated the redundancy level of information received from DS with already existing information. Participants' task in two between-subjects conditions was to classify binary events with the option to access up to two DS systems. In one of the conditions, the two DSs provided non-redundant information, and in the second condition, one of them provided fully redundant information, and the other provided non-redundant information. We found that the decision to access information was not affected by whether some information was redundant (strongly correlated with already available information). Participants used the information to improve classification performance, and the improvement was significantly higher when they used non-redundant information. However, the benefits gained were smaller than predicted from a normative model. Moreover, the use of information from multiple non-correlated sources can increase mental workload, as was evident in our results, possibly because of conflicting information from different sources.

The photo-taking-impairment effect refers to the detrimental impact photo taking has on one's memory for the photographed object. We explored the role of two mechanisms that have been said to underlie the effect, namely offloading and attentional disengagement. In this online study 107 undergraduate students were shown 3x5 paintings and were instructed to simply observe them, take a picture that would be available for later usage, or take a picture and delete it. Afterwards, they were presented with a visual details multiple-choice test. It was expected that if attentional disengagement was the mechanism underlying the photo-taking-impairment effect, then the effect would not be present in the current study. This expectation was due to the non-distracting nature of the photo taking task that was used in the study. The results were in line with the expectations and did not indicate the presence of the photo-taking-impairment effect. Consequentially, it supported the hypothesis that attentional disengagement, rather than offloading, is the mechanism underlying this effect.

This paper presents an approach to model the internal cognitive state of decision-makers when interacting with AI to understand exchanges between agents and improve future interactions. We focus on understanding how AI suggestions are perceived by a human agent using an approach based on the technology acceptance model. The variation in the user's state is investigated when perceiving the interaction with AI by considering it as a hidden (latent) state. Using human evaluation data collected from two cases of clinical decision-making and software development scenarios, we analyse and explore the user's perceptional state during interaction. The experiment conducted employs the Bayesian belief network to represent the human perceptional model and provide a prediction of the usefulness of AI model's suggestions in the considered case. Upon introduction of cognitive states in the model, we observed an increase in predictive performance by 76–77%. Our investigation can be concluded as an attempt to identify implicit static and dynamic cognitive characteristics of users to provide personalized assistance in human-AI interaction (HAI) and collaboration in complex domains of decision-making

The steady increase in time spent on smartphone applications and particularly on social networks, raises questions about the environmental and societal sustainability of such a phenomenon. Utility and enjoyment have a key role in such practices, but other factors such as passing time may also contribute. From May to November 2023, 5,028 people took part in a web survey aiming at producing durations prospectively using mobile applications like Facebook, Instagram, TikTok, Reading. The protocol introduces variables known to have an effect on time perception. On average, the produced durations were underestimated. This result is in line with the notion that tracking information and tracking time compete for the brain's limited attentional resources and, hence, that attention plays a critical role in time estimation. Significant differences emerged between the applications tested. TikTok and Reading tasks appear the most underestimate but with opposite dynamics as the level of satisfaction and familiarity are lower for the first compared to the former. Among the variables studied to explain the difficulties in evaluating time spent, the importance of familiarity with the activity is undoubtedly something worth exploring in the context of the race between new algorithms and cognitive adaptability.

Previous research on perception of correlation of scatterplots used scatterplots as stimuli and asked participants to estimate or compare correlations of those scatterplots. This literature has shown a tendency for people to underestimate correlation in some correlation ranges. We flipped the task: instead of estimating correlation from visual stimuli, participants drew a scatterplot based on a given correlation: 0, 0.25, 0.5, 0.75 and 1 using 20 dots. Participants drew greater correlations for r = 0.25 and r = 0.5 (0.59 and 0.71 respectively), which is analogous to underestimating correlation in previous viewing tasks. Drawn correlations for r = 0, 0.75 and 1 were more accurate. The number of statistics courses taken did not improve correlation drawing accuracy in a strong or meaningful way. We discuss possible interpretations of these results and future directions.

In optimal-stopping problems, people encounter options sequentially with the goal of finding the best one; once it is rejected, it is no longer available. Previous research indicates that people often do not make optimal choices in these tasks. We examined whether additional information about the task's environment enhances choices, aligning people's behaviour closer to the optimal policy. Our study implemented two additional-information conditions: (1) a transparent presentation of the underlying distribution and (2) a provision of the optimal policy. Our results indicated that while choice patterns varied weakly with additional information when providing the optimal policy, it did not significantly enhance participants' performance. This finding suggests that the challenge in following the optimal strategy is not only due to its computational complexity; even with access to the optimal policy, participants often chose suboptimal options. These results align with other studies showing people's reluctance to rely on algorithmic or AI-generated advice.

The recent surge in the use of AI-powered chatbots such as ChatGPT has led to new challenges in academia. These chatbots can enable student plagiarism and the submission of misleading content, undermining educational objectives. With plagiarism detectors unreliable in the face of this issue, educational institutions have been struggling to update their policies apace. This study assesses the effectiveness of sending warning messages - a common strategy used to discourage unethical use of ChatGPT - and investigates the use of the illusion of explanatory depth (IOED) paradigm as an alternative intervention. An international sample of students was asked to rate their understanding of, likelihood to use, and moral stance toward ChatGPT-generated text in assignments both before and after either reading a cautionary university message or explaining how ChatGPT works. Results showed that the explanation task did lead to the expected reduction in ratings of understanding, but despite this, neither moral acceptability nor likelihood to use decreased along with it. Similarly, reading the cautionary message neither resulted in a change in likelihood to use nor in moral acceptability, although it unexpectedly increased ratings of understanding. The results suggest that tackling students' understanding of ChatGPT is insufficient when it comes to deterring its unethical use, and that future interventions might want to have students reflect on moral issues surrounding the use of AI-powered chatbots.

This study examines 205 4- to 12-year-old children's beliefs about the abilities of three technological informants (the internet as a whole, Google search, and Amazon's Alexa smart speaker) to answer questions about celebrity and non-celebrity people and near future and far future events. The results indicate that, with increasing age, children increasingly indicate that these sources can accurately answer questions about near future events and celebrities but not about non-celebrities or far future events. Although children increasingly indicate that these sources cannot tell them about everyday people, the oldest children in the sample believe that the internet is more likely to be able to tell you about non-celebrities than Alexa or a Google search. Children's understanding of the capacities of technology change with age and information type, perhaps reflecting changes in children's experiences online. Implications for children's learning and understanding of privacy are discussed.

Empirical work on people's perceptions of AI advisors has found evidence for both “algorithm aversion” and “algorithm appreciation.” We investigated whether these differing reactions stem from two different paths of processing: assessing the content of the advice and evaluating the source (AI vs. human advisor). In two survey studies, people were as strongly persuaded by the advice of an AI as that of a human advisor; nonetheless, people's approval of and trust in the AI advisor was consistently lower. This pattern of dissociation suggests that algorithm aversion and algorithm appreciation can occur at the same time, but along different response paths.

LLMs can interact as though they understand language, yet they remain algorithms and can be used as such. This study explores a novel guided interaction design for modeling users' information foraging behavior when navigating GPT-generated content and the role of Computational Thinking skills in shaping such behavior. Conducted with nine educational researchers in a doctoral-level AIEd course, our research used editable prompt templates and keywords to structure the prompt crafting process. We modeled and analyzed participants' behaviors in terms of \textit{exploration} (to generate and explore various information landscapes) and \textit{exploitation} (to delve deeper in a specific landscape). Our data, including responses from the Computational Thinking Scale, suggests that Algorithmic Thinking and Creativity might encourage exploitation behavior, leaning more on AI-generated information rather than pre-defined design elements.

Bar graphs are commonly used graphs, but what do students infer about the data that created the bar graph? Previously, a drawing task revealed that a minority of students conflate mean bar graphs with count bar graphs and draw all data points within the bar of a mean bar graph (bar tip limit error, BTLE). The present study extends this literature by manipulating the instructional text for the drawing task, interviewing the participants on their drawings, and recording their drawings and drawing session for further analysis. While we did not see any differences in the BLTE rates across instructional conditions, we did see significant differences in their drawing explanations and drawn data distributions based on condition, and in their drawing explanations based on whether they expressed confusion and whether they committed the BTLE. We discuss possible explanations and their implications.

How to best and robustly detect human systemic cognitive states like workload, sense of urgency, mind wandering, interference, and others is still an open question as the answer essentially depends both on the employed physiological measurements as well as the trained computational classification models. In this paper, we analyze data from a human driving experiment to explore the validity of eye gaze in assessing different systemic cognitive states and relations among them. Our statistical analyses and classification results indicate that eye gaze, in particular the percentage change in pupil size (PCPS), is a reliable physiological biomarker in assessing multiple systemic human cognitive states including workload, sense of urgency (SoU), and mind wandering (MW) while it does not seem suitable to detect task interference (which can be assessed based on participant's response times.

Termed the 'Made-by-Human Hypothesis,' I challenge the commonly accepted notion that Artificial Intelligence (AI) is exclusively crafted by humans, emphasizing its impediment to progress. I argue that influences beyond human agency significantly shape AI's trajectory. Introducing the 'Hybrid Hypothesis,' I suggest that the creation of AI is multi-sourced; methods such as evolutionary algorithms influencing AI originate from diverse sources and yield varied impacts. I argue that the development of AI models will increasingly adopt a 'Human+' hybrid composition, where human expertise merges with AI's intrinsic mechanisms, which themselves are influenced by non-human sources. The Hybrid Hypothesis posits that the origin of AI extends beyond human influence, prompting a thorough exploration of unresolved issues in the field of artificial intelligence.

This study investigates the differential impacts of virtual and human influencers on consumer purchase intentions, focusing particularly on the roles of perceived psychological distance and credibility. Utilizing image recognition algorithms, two influencers with facial similarities were stringently selected, and surveys from 427 consumers on their perceptions of the products endorsed by these influencers were analyzed. Results show human influencers outperform virtual ones, yet the latter still positively affect purchase intentions, revealing their potential as effective marketing tools. The study further reveals that perceived psychological distance can independently mediate the relationship between influencer type and purchase intention, and also acts in tandem with perceived credibility in this mediation. This research not only offers empirical insights into the comparative effectiveness of virtual versus human influencers in digital marketing but also advances understanding of the psychological mechanisms underpinning consumer behavior in the digital era.

Interlingual Respeaking (IR) is a new technique that enables real-time subtitling in a different language. This cognitively demanding technique involves collaboration between a language professional and automatic speech recognition software (ASR), creating a human-AI interaction (HAII) environment. Integrating technological tools with an individual's internal cognitive resources establishes an extended cognitive system. However, different types of errors are observed in terms of output accuracy. Our ESRC-funded research found that working memory (WM) (backward span) has a negative relationship with omissions, where content is dropped out (e.g., to save time). Nevertheless, additions, where the human adds content (e.g., to clarify meaning) and correctness, where form-related issues arise (such as grammar mistakes), had an inverse relationship with the N-back Task (the simultaneous maintenance, updating, and processing of WM). These findings suggest that the IR errors involve diverse types of WM resources.

In our study, we address conflicts between individuals and groups, such as cyberbully on social media, as a challenge related to the distinction between the self and others. To address this issue using technology, we propose the concept of introducing facial synchronization in the virtual realm as a means to manipulate the boundary between oneself and others.We designed an experiment using Cyberball that simulates an ostracism environment, effectively partitioning the boundary between the self and others. This task was conducted in Virtual Reality (VR), with the agent's facial expressions synchronized with those of the participant.Our findings indicated a reduction in feelings of alienation within the ostracism environment. This discovery has potential implications for communication media, particularly in enhancing interfaces for individuals who may experience exclusionary behavior on social media.

Eye movement detection plays a crucial role in various fields, including eye tracking applications and understanding human perception and cognitive states. Existing detection methods typically rely on gaze positions predicted by gaze estimation algorithms, which may introduce cumulative errors. While certain video-based methods, directly classifying behaviours from videos, have been introduced to address this issue, they often have limitations as they primarily focus on detecting blinks. In this paper, we propose a video-based two-stream framework designed to detect four eye movement behaviours—fixations, saccades, smooth pursuits, and blinks—from infrared near-eye videos. To explicitly capture motion information, we introduce optical flow as the input for one stream. Additionally, we propose a spatio-temporal feature fusion module to combine information from the two streams. The framework is evaluated on a large-scale eye movement dataset and performs excellent results.