Human Users Research Articles

Making New Connections: LLMs as Puzzle Generators for the New York Times' Connections Word Game

The Connections puzzle is a word association game published daily by The New York Times (NYT). In this game, players are asked to find groups of four words that are connected by a common theme. While solving a given Connections puzzle requires both semantic knowledge and abstract reasoning, generating novel puzzles additionally requires a form of metacognition: generators must be able to accurately model the downstream reasoning of potential solvers. In this paper, we investigate the ability of the GPT family of Large Language Models (LLMs) to generate challenging and creative word games for human players. We start with an analysis of the word game Connections and the unique challenges it poses as a Procedural Content Generation (PCG) domain. We then propose a method for generating Connections puzzles using LLMs by adapting a Tree of Thoughts (ToT) prompting approach. We evaluate this method by conducting a user study, asking human players to compare AI- against human-generated Connections puzzles. Our findings show that LLMs are capable puzzle creators, and can generate diverse sets of enjoyable, challenging, and creative Connections puzzles as judged by human users.

NIMG-63. LEVERAGING LLMS FOR ACCURATE DIFFERENTIATION OF RADIATION NECROSIS AND TUMOR PROGRESSION IN BRAIN MRI REPORTS: A STUDY ON AUTOMATED SCORING AND CLINICAL IMPLICATIONS

Abstract BACKGROUND Accurate differentiation between radiation necrosis and tumor progression or recurrence in patients treated with stereotactic radiosurgery for brain metastases is critical for guiding clinical management. This study leverages the advanced natural language processing capabilities of Meta Llama3, an artificial intelligence (AI) large language model (LLM), combined with prompt engineering, to rapidly categorize brain magnetic resonance imaging (MRI) radiology reports. Our objective was to develop an automated scoring system to classify concern for radiation necrosis, tumor progression, equivocal findings, or stable exams. METHODS Using a comprehensive dataset of reports annotated by expert radiologists, we ran inference on a 70-billion parameter Llama3 model (temperature 0.2, top_p 0.9) with specific prompts designed to capture the nuanced language and diagnostic criteria related to radiation necrosis or tumor progression. RESULTS The first pass was performed on a training dataset of 107 reports and did not predefine the clinical conditions. This demonstrated 43.4% accuracy in scoring when compared to a human user’s classification of each report. Agreement between the human reader and Llama3 was assessed using the Gwet agreement coefficient, AC1=0.411 (99%CI 0.396-0.426). Multiple iterations of targeted prompt engineering were then employed to narrow the definition of radiation necrosis and tumor progression, with specific examples and nuanced language used to achieve a higher degree of accuracy at 72.0%, AC1=0.719 (99%CI 0.717-0.722). DISCUSSION/FUTURE DIRECTIONS This surpasses recent demonstration of lower human-LLM agreement in radiographic score assignment, with further room for calibration on a dataset of several thousand reports. Next, we will correlate automated interpretations with actual clinical management decisions for radiation necrosis (e.g., initiation of steroids, bevacizumab, Laser Interstitial Thermal Therapy, and/or repeat imaging). This automated scoring system holds significant potential for LLMs in clinical applications. Future work will focus on integrating this model into clinical workflows and expanding its capabilities to include longitudinal monitoring of patient outcomes.

Creativity, credit, and copyright in the age of artificial art

ABSTRACT Generative artificial intelligence is transforming the way we make, and think about, art. With prompting from human users, these generative systems now produce aesthetically compelling and seemingly creative works in a variety of artistic domains. In doing so, they challenge the ways we think about artistic credit, about creativity, and about the mechanism of legal copyright, which is meant to protect and promote creativity in a capitalist art market. All of this is currently at play in the courtroom, as artists contest the ways in which their artworks can be rightfully fed into these artificial systems (“the problem of the inputs”), and other artists are challenged over whether they might be credited for the visual images these systems generate (“the problem of the outputs”). Here, we explore these problems as they arise in visual art. We argue that the contested legal landscape surrounding these artificial systems reflects the ways they challenge our received notions of artistic credit and creativity. And we suggest that clarifying or changing the application of copyright in light of their generative capacities will ultimately involve revising our conception of art.

Beyond Preferences in AI Alignment

AbstractThe dominant practice of AI alignment assumes (1) that preferences are an adequate representation of human values, (2) that human rationality can be understood in terms of maximizing the satisfaction of preferences, and (3) that AI systems should be aligned with the preferences of one or more humans to ensure that they behave safely and in accordance with our values. Whether implicitly followed or explicitly endorsed, these commitments constitute what we term a preferentist approach to AI alignment. In this paper, we characterize and challenge the preferentist approach, describing conceptual and technical alternatives that are ripe for further research. We first survey the limits of rational choice theory as a descriptive model, explaining how preferences fail to capture the thick semantic content of human values, and how utility representations neglect the possible incommensurability of those values. We then critique the normativity of expected utility theory (EUT) for humans and AI, drawing upon arguments showing how rational agents need not comply with EUT, while highlighting how EUT is silent on which preferences are normatively acceptable. Finally, we argue that these limitations motivate a reframing of the targets of AI alignment: Instead of alignment with the preferences of a human user, developer, or humanity-writ-large, AI systems should be aligned with normative standards appropriate to their social roles, such as the role of a general-purpose assistant. Furthermore, these standards should be negotiated and agreed upon by all relevant stakeholders. On this alternative conception of alignment, a multiplicity of AI systems will be able to serve diverse ends, aligned with normative standards that promote mutual benefit and limit harm despite our plural and divergent values.

Research on CAPTCHA recognition technology based on deep learning

Abstract. CAPTCHA (Completely Automated Public Turing test to tell Computers and Humans Apart), a security technique widely used on the Internet, can be utilized to distinguish human users from automated programs. The rapid development of deep learning technology has led to the emergence of graph-based recognition techniques that demonstrate excellent performance, which has prompted the investigation of CAPTCHA recognition based on deep learning as a research area of significant interest. This paper addresses the challenging problem of CAPTCHA recognition based on deep learning techniques, reviews the development and classification of CAPTCHA, examines traditional CAPTCHA recognition methods, and delves into the application of deep learning in CAPTCHA recognition. Therefore, a CAPTCHA recognition system is designed and its effectiveness is verified through experiments. This paper makes a significant contribution to the field of CAPTCHA recognition by proposing a deep learning-based approach, which not only enhances the accuracy and efficiency of CAPTCHA recognition, but also provides new ideas and methods for the further development. In the future, further research will be conducted in the field of CAPTCHA recognition to explore additional deep learning models and techniques with the aim of boosting the security and user experience of CAPTCHA.

A Roadmap of Explainable Artificial Intelligence: Explain to Whom, When, What and How?

Explainable artificial intelligence (XAI) has gained significant attention, especially in AI-powered autonomous and adaptive systems (AASs). However, a discernible disconnect exists among research efforts across different communities. The machine learning community often overlooks “explaining to whom,” while the human-computer interaction community has examined various stakeholders with diverse explanation needs without addressing which XAI methods meet these requirements. Currently, no clear guidance exists on which XAI methods suit which specific stakeholders and their distinct needs. This hinders the achievement of the goal of XAI: providing human users with understandable interpretations. To bridge this gap, this paper presents a comprehensive XAI roadmap. Based on an extensive literature review, the roadmap summarizes different stakeholders, their explanation needs at different stages of the AI system lifecycle, the questions they may pose, and existing XAI methods. Then, by utilizing stakeholders’ inquiries as a conduit, the roadmap connects their needs to prevailing XAI methods, providing a guideline to assist researchers and practitioners to determine more easily which XAI methodologies can meet the specific needs of stakeholders in AASs. Finally, the roadmap discusses the limitations of existing XAI methods and outlines directions for future research.

Interpreting Bar Charts: Effects of 3D Depth Cues on Human Gaze and User Understanding

Three-dimensional information visualizations are widely used in various fields for their aesthetic appeal. However, using them can sometimes lead to occlusion and distortion, which raises questions about when and why to use them. In this study, we aimed to investigate the effects of three-dimensional visualizations on human gaze and user understanding and analyze the perception process in detail. Our empirical research consisted of a two-part experimental study that involved both subjective and objective evaluation. We specifically focused on bar charts as they are among the most frequently used types of information visualizations. The results suggest that, for bar chart visualizations with varying gap dimensions, there is no statistically significant difference in user understanding between the two-dimensional and three-dimensional versions. Our findings indicate that, in general, three-dimensional bar chart visualizations are as comprehensible as their two-dimensional counterparts for the gap dimensions examined in this research. This study provides empirical insights demonstrating that both 3D and 2D bar charts are equally understandable, particularly when a specific gap depth is used in 3D visualizations. These findings contribute to the ongoing discussion about the effective use of three-dimensional visualizations and highlight areas for further research.

Il progetto di paesaggio come teatro di coesistenza tra specie. Parc Martin-Luther-King a Parigi

Landscape architecture plays a crucial role in defining new paradigms for urban development, capable of implementing environmental quality and favouring biodiversity. In this sense, Parc Martin-Luther-King in Paris constitutes a sophisticated device with a high ecological value as well as a surprising catalyst of social aggregation, able of regenerating the urban area of Clichy-Batignolles. The articulate plant palette, the use of water in several forms and the application of differentiated management regimes help defining a collection of heterogeneous ecosystems where various forms of controlled interaction between human and non-human users are configured. The strong spatial dynamism and ecological variety favour the coexistence of different species, allowing the park to accommodate spaces for biodiversity alongside a rich functional programme.

Performance evaluation of ChatGPT 4.0 on cardiovascular questions from the medical knowledge self-assessment program

Abstract Background Medical trainees are increasingly using online chat-based artificial intelligence (AI) platforms as supplementary resources for board exam preparation and clinical decision support. Prior studies have evaluated the performance of AI chatbots like ChatGPT on various general standardized tests such as the United States Medical Licensing Examination (USMLE), but little is known about their performance on subspecialty-focused exam questions, particularly related to clinical management and treatment. Objective This study aims to evaluate the performance of ChatGPT version 4.0 on the cardiovascular questions from the Medical Knowledge Self-Assessment Program (MKSAP) 19, a widely used resource for board exam preparation in the United States. Methods We submitted all cardiovascular questions from MKSAP 19 to ChatGPT 4.0, covering a broad range of cardiology topics in a multiple-choice format. Performance was gauged against both the official MKSAP answer key and average trainee scores obtained from the MKSAP website. Out of 129 questions, 4 were invalidated due to post-publication data, and 18 were excluded due to reliance on visual aids, leaving 107 questions for analysis. Results ChatGPT 4.0 correctly answered 93 out of 107 questions, reflecting an 87% accuracy rate, compared to a 60% accuracy rate averaged among all human users for the same questions (p<0.0001). ChatGPT accuracy rates for each question category (e.g., heart failure, electrophysiology, etc.) are provided in the figure. On the 14 questions that ChatGPT answered incorrectly, human users averaged a 47% accuracy rate. All 14 incorrectly answered questions were related to clinical management and treatment (as opposed to diagnosis or epidemiology), and the majority (10/14) involved choosing treatment medications or choosing between medications versus interventions. Conclusion ChatGPT 4.0 surpassed the typical 70% accuracy threshold required for passing the internal medicine board certification but performed less well in answering questions about clinical management and treatment. This performance suggests that ChatGPT 4.0 could be an effective adjunct tool for cardiology education, but raises some concerns about its use in clinical decision support for trainees.

Does Prosocial Automation Increase Driver’s Well-being?

As societies transition to hybrid mobility systems, interactions between automated vehicles (AVs) and human users in public spaces become more complex, highlighting the critical role of prosocial behaviors. These behaviors are essential for the seamless operation of interdependent transportation networks, helping to address integration challenges of AVs with human-operated vehicles and enhancing well-being by creating more efficient, less stressful, and inclusive environments. This study explores the impact of receiving prosocial behaviors on the cognition, riding performance, and well-being of micromobility users in simulated urban traffic scenarios. Using a mixed-method design, the research contrasts two types of social interactions—prosocial and asocial—across three temporal conditions: relaxed, neutral, and urgent. The Structural Equation Modeling underscored the intricate relationships between prosocial behaviors, satisfaction, cognition, riding performance, and well-being. Incorporating social factors into driving systems could enhance safety and improve urban mobility solutions.

Beyond humanism: telling response-able stories about significant otherness in human-chatbot relations.

AI-enabled chatbots intended to build social relations with humans are becoming increasingly common in the marketplace, with millions of registered users using these chatbots as virtual companions or therapists. These chatbots make use of what is often called the "Eliza effect"-the tendency of human users to attribute human-like knowledge and understanding to a computer program. A common interpretation of this phenomenon is to consider this form of relating in terms of delusion, error, or deception, where the user misunderstands or forgets they are talking to a computer. As an alternative, we draw on the work of feminist Science and Technology Studies scholars as providing a robust and capacious tradition of thinking and engaging with human-nonhuman relationships in non-reductive ways. We closely analyze two different stories about encounters with chatbots, taking up the feminist STS challenge to attend to the agency of significant otherness in the encounter. The first is Joseph Weizenbaum's story about rejecting the ELIZA chatbot technology he designed to mimic a therapist as a monstrosity, based on his experiences watching others engage with it. The second is a story about Julie, who experiences a mental health crisis, and her chatbot Navi, as told through her descriptions of her experiences with Navi in the recent podcast Radiotopia presents: Bot Love. We argue that a reactionary humanist narrative, as presented by Weizenbaum, is incapable of attending to the possibilities of pleasure, play, or even healing that might occur in human-chatbot relatings. Other forms of engaging with, understanding, and making sense of this new technology and its potentialities are needed both in research and mental health practice, particularly as more and more patients will begin to use these technologies alongside engaging in traditional human-led psychotherapy.

A Human-Centered View of Continual Learning: Understanding Interactions, Teaching Patterns, and Perceptions of Human Users Toward a Continual Learning Robot in Repeated Interactions

Continual learning (CL) has emerged as an important avenue of research in recent years, at the intersection of Machine Learning (ML) and Human–Robot Interaction (HRI), to allow robots to continually learn in their environments over long-term interactions with humans. Most research in CL, however, has been robot-centered to develop CL algorithms that can quickly learn new information on systematically collected static datasets. In this article, we take a human-centered approach to CL, to understand how humans interact with, teach, and perceive CL robots over the long term, and if there are variations in their teaching styles. We developed a socially guided CL system that integrates CL models for object recognition with a mobile manipulator robot and allows humans to directly teach and test the robot in real time over multiple sessions. We conducted an in-person study with 60 participants who interacted with the CL robot in 300 sessions with 5 sessions per participant. In this between-participant study, we used three different CL models deployed on a mobile manipulator robot. An extensive qualitative and quantitative analysis of the data collected in the study shows that there is significant variation among the teaching styles of individual users indicating the need for personalized adaptation to their distinct teaching styles. Our analysis shows that the constrained experimental setups that have been widely used to test most CL models are not adequate, as real users interact with and teach CL robots in a variety of ways. Finally, our analysis shows that although users have concerns about CL robots being deployed in our daily lives, they mention that with further improvements CL robots could assist older adults and people with disabilities in their homes.

LIMIT: Learning Interfaces to Maximize Information Transfer

Robots can use auditory, visual, or haptic interfaces to convey information to human users. The way these interfaces select signals is typically pre-defined by the designer: for instance, a haptic wristband might vibrate when the robot is moving and squeeze when the robot stops. But different people interpret the same signals in different ways, so that what makes sense to one person might be confusing or unintuitive to another. In this article, we introduce a unified algorithmic formalism for learning co-adaptive interfaces from scratch . Our method does not need to know the human’s task (i.e., what the human is using these signals for). Instead, our insight is that interpretable interfaces should select signals that maximize correlation between the human’s actions and the information the interface is trying to convey. Applying this insight we develop Learning Interfaces to Maximize Information Transfer (LIMIT). LIMIT optimizes a tractable, real-time proxy of information gain in continuous spaces. The first time a person works with our system the signals may appear random; but over repeated interactions, the interface learns a one-to-one mapping between displayed signals and human responses. Our resulting approach is both personalized to the current user and not tied to any specific interface modality. We compare LIMIT to state-of-the-art baselines across controlled simulations, an online survey, and an in-person user study with auditory, visual, and haptic interfaces. Overall, our results suggest that LIMIT learns interfaces that enable users to complete the task more quickly and efficiently, and users subjectively prefer LIMIT to the alternatives. See videos here: https://youtu.be/IvQ3TM1_2fA .

Self‐Deception in Human–Sex Robot Intimacy

ABSTRACTA common sentiment among anti‐sex‐robot scholars is the apprehension that sex robots will normalize and perpetuate sexual violence towards humans. In this new chapter within the feminist sex war, the authors of this article tend to agree with anti‐sex‐robot concerns and seek to further identify potential harms of sex robots. However, instead of characterizing the harm in terms of what the robots represent and symbolize, we are primarily interested in the internal state of the user and the type of relationship that will emerge between human users and sex robots, which we argue is an unprecedented sexual relation. Unlike other comparable sex products and services, sex robots occupy a liminal space between being perceived as both a sexual property and agent, oscillating based on the preferences and convenience of the user. We argue that this oscillation that enables human–sex robot intimacy requires self‐deception, which in turn entails individual moral responsibility. Thus, we articulate a novel virtue‐based approach of examining human–robot intimacy that focuses on cultivating erotic flourishing. We conclude that people have a moral responsibility to exhibit self‐awareness within the dynamics of their intimate relationship with sex robots and the (contradictory) beliefs required to maintain such intimacy.

Non-Copyrightability of Data in Scientific Publications: A Free-for-All or a Global Commons Partnership?

Scientific publications provide a wealth of peer-reviewed, high-quality data that have been maintained over time, resulting in data persistence. As data repositories with rich provenance information, publications are indispensable sources for the integration and extension of networks of interlinked Findable, Accessible, Interoperable and Reusable (FAIR*1) bio/geodiversity data. In this way, they form pivotal fact- and knowledge-based contributions to applications that address the biodiversity crisis. The mobilization of data preserved in scientific publications is hindered, however, by distinct copyright legislation contexts for publications versus the data that they contain. Moreover, legislations concerning copyright continue to lack harmonization across jurisdictions, their interpretation is difficult, and the applicable legal national scope can be uncertain. We clarify and highlight that data within scientific publications are not copyrightable and thus can be openly and freely reused once legal access has been gained to their enclosing publication*2. To ensure that publications are as accessible as possible, a joint statement supported by the Biodiversity Heritage Library (BHL), the Consortium of European Taxonomic Facilities (CETAF) and the Society for the Preservation of Natural History Collections (SPNHC) (Benichou et al. 2023) recommends that authors and publishers make their works as accessible as possible by using a CC-BY license or preferably waive copyright (CC0) to their publications. Explicitly associating a public domain mark (PDM, e.g., the PDM from Creative Commons) to their published data, provides users with certainty about reusability. Yet, by setting works and bio/geodiversity data into the public domain, they do not become a free-for-all. We stress that data need to be associated with clear provenance information in alignment with scientific best practices and the scientific community's social norms. This includes providing detailed attribution to authors of cited works and reused data. Proposed data governance labels, for example, modeled after the Local Contexts labels developed by the international Indigenous Peoples and Local Communities (IPLC) community, would enable authors to communicate social and ethical contexts and applicable rules to data users for ensuring the sustainability of a shared environmental and data commons. Categories of Local Contexts labels that are of interest and applicable in the sciences are, for example, those that communicate (1) correct citation information and ask for attribution when knowledge and/or data are reused (Traditional Knowledge label (TK) Attribution), (2) an interest in being recognized and acknowledged due to a significant relationship with and responsibility for samples and data (Biocultural label (BC) Provenance), (3) the verification of the data and their context following a community protocol (TK Verified), (4) that non-commercial use (TK Non-Commercial/BC Non-Commercial) or (5) outreach activities (TK Outreach/BC Outreach) are generally permitted, while for other uses direct contact and engagement is required, or (6) an openness to collaboration and partnerships (TK Collaboration/BC Collaboration). There are concerns about the tension between the goal of achieving open data (e.g., Anonymous 2014) to enable and promote open science (e.g., UNESCO 2021) and, at the same time, imposing restrictions on these data in the form of governance labels. Furthermore, while the reference of the publication through which data are published, as well as more specifically bibliographic references cited for specific data within the publication, provide sufficient information for attribution and provenance, much more fine-grained and nuanced contextual information (e.g., in the form of metadata) is needed for assuring responsible reuse. Such context-providing metadata unlock the full potential of the data and enable their reusability. This can be done using machine-actionable markup tags in combination with human-readable labels that inform machines and human users about the semantics of the data as well as their ethical and social dimensions that govern responsible and sustainable reuse. Future work is needed to discover, differentiate and define the quality and scope of the appropriate contexts that are necessary and sufficient for being able to fully and responsibly reuse the data in different situations.

Multimodal emotion recognition: A comprehensive review, trends, and challenges

AbstractAutomatic emotion recognition is a burgeoning field of research and has its roots in psychology and cognitive science. This article comprehensively reviews multimodal emotion recognition, covering various aspects such as emotion theories, discrete and dimensional models, emotional response systems, datasets, and current trends. This article reviewed 179 multimodal emotion recognition literature papers from 2017 to 2023 to reflect on the current trends in multimodal affective computing. This article covers various modalities used in emotion recognition based on the emotional response system under four categories: subjective experience comprising text and self‐report; peripheral physiology comprising electrodermal, cardiovascular, facial muscle, and respiration activity; central physiology comprising EEG, neuroimaging, and EOG; behavior comprising facial, vocal, whole‐body behavior, and observer ratings. This review summarizes the measures and behavior of each modality under various emotional states. This article provides an extensive list of multimodal datasets and their unique characteristics. The recent advances in multimodal emotion recognition are grouped based on the research focus areas such as emotion elicitation strategy, data collection and handling, the impact of culture and modality on multimodal emotion recognition systems, feature extraction, feature selection, alignment of signals across the modalities, and fusion strategies. The recent multimodal fusion strategies are detailed in this article, as extracting shared representations of different modalities, removing redundant features from different modalities, and learning critical features from each modality are crucial for multimodal emotion recognition. This article summarizes the strengths and weaknesses of multimodal emotion recognition based on the review outcome, along with challenges and future work in multimodal emotion recognition. This article aims to serve as a lucid introduction, covering all aspects of multimodal emotion recognition for novices.This article is categorized under: Fundamental Concepts of Data and Knowledge > Human Centricity and User Interaction Technologies > Cognitive Computing Technologies > Artificial Intelligence

WiCAM2.0: Imperceptible and Targeted Attack on Deep Learning based WiFi Sensing

With the widespread adoption of deep learning models in wireless sensing, substantial efforts have been made to develop sophisticated models that improve the accuracy and performance of sensing applications. However, the exploration of potential vulnerabilities in deep models has been limited, with existing studies primarily focusing on evaluating wireless adversarial performance in communication or sensing alone. Moreover, there is a lack of a comprehensive definition for attack imperceptibility. In this paper, we come up with a definition of the wireless attack imperceptibility for both communication and sensing. Our objective is to create an adversarial perturbation capable of degrading WiFi sensing performance while preserving WiFi communication integrity. To achieve this, we propose WiCAM2.0 to reveal the temporal and spatial attention of a DNN, capturing the crucial portions of its input. Then we design a mask to confine adversarial perturbations in the attended parts only, minimizing the impact on WiFi communication. WiCAM2.0 is a general adversarial framework that integrates adversarial methods like FGSM and PGD to generate perturbations, capable of initiating both non-targeted and targeted attacks. We carry out experiments on three popular WiFi sensing applications, including human activity recognition, gesture recognition, and user identification. Extensive experiments are conducted on both public datasets and self-collected datasets.

Explainable AI based Predictions for Workpiece Quality

It is well known that sound emissions of drilling and milling machines can be used to predict process and work piece quality. Deep Learning models have successfully been applied for that task. However, artificial neural networks are perceived as black boxes as the inference mechanisms of the network are not transparent. In this paper we present a Convolutional Neural Network trained on audio data with which process and product quality of a milling process can be controlled. The model was trained on 10-second audio snippets recorded from a milling machine that were converted into spectrograms. The trained model classifies the drilling process and predicts the expected workpiece quality. Additionally, rough estimates of the center rough and average roughness depth values are predicted. Moreover, we apply the Explainable AI methods that produce explanations by highlighting sections of the spectrogram that were relevant for the prediction. The spectrograms enriched with the explanations provide insights into the decision making process of the deep learning model in a human interpretable form. Patterns of how and why of a prediction become visible. Areas considered relevant by the model can be compared to the expectations of the human user. Through this explanation component, the model is not only able to make predictions but also to increase trust in the system which is an important aspect of acceptance of AI based quality assurance.

Team up with AI or Human? Investigating Candidates’ Self-Categorization as Fluidity and Ingroup-Serving Attribution When Judged by a Human–AI Hybrid Jury

As artificial intelligence (AI) judges are increasingly pervasive in decision-making, it is important to investigate candidates’ reactions to decisions made by human–AI hybrid juries. This study investigates candidates’ attribution of credit for success and blame for failure to the three agents in question: a human judge, an algorithmic judge, and the candidate oneself. An experiment with 3 (jury type: human-dominated, algorithm-dominated, vs. equally dominated) × 2 (decision outcome: positive vs. negative) between-subjects factorial design was conducted, with 346 valid responses. Our findings demonstrate a partial ingroup-serving attribution dependent on the outcome favorability and a significant effect of relative power status within the human–AI hybrid jury on grouping and attribution. This study reflects the fluidity of identity and self-categorization of human users when facing AI and other humans. We propose that people take a utility-oriented glance at AI in multi-agent decision-making situations.

LongT5Rank: A Novel Integrated Hybrid Approach for Text Summarisation

Text summarisation reduces text length while retaining important information, helping individuals, especially students, in managing information overload during research or assignments. However, existing text summarisation methods often lose important details, generate irrelevant or redundant sentences, or produce incoherent summaries. This study introduces a hybrid approach, LongT5Rank (coined in this study), which combines TextRank, an extractive summarisation algorithm, with LongT5, an abstractive summarisation algorithm, to automate the summarisation process. TextRank utilizes GloVe, a pre-trained word embedding model, and PageRank, a graph-based ranking algorithm, to select representative sentences. LongT5, an encoder-decoder transformer model, condenses extracted sentences into a concise and coherent summary, handling input sequences up to 16,384 tokens, for long-range sequence-to-sequence tasks. The LongT5Rank approach has shown significant achievements, including a minimum 60% compression rate, a minimum 0.6 semantic textual similarity score, and an improved F-measure compared to employing TextRank alone. Furthermore, it received positive feedback from Human Level Performance (HLP), underlining the importance of evaluating results directly from human users. This emphasizes the belief that the performance of the proposed solution should be assessed natively by humans. By combining both extractive and abstractive methods, LongT5Rank excels in generating accurate and coherent summaries.