Human AI Research Articles

A Unified Model of Direct and Indirect Reciprocity in Multichannel Games

Reciprocity plays a crucial role in maintaining cooperation in human societies and AI systems. In this paper, we focus on reciprocity within multichannel games and examine how cooperation evolves in this context. We propose a unified framework that allows us to evaluate the reputations of interdependent actions across multiple channels while simultaneously exploring both direct and indirect reciprocity mechanisms. We identify partner and semi-partner strategies under both forms of reciprocity, with the former leading to full cooperation and the latter resulting in partial cooperation. Through equilibrium analysis, we characterize the conditions under which full cooperation and partial cooperation emerge. Moreover, we show that when players can link multiple interactions, they learn to coordinate their behavior across different games to maximize overall cooperation. Our findings provide new insights into the maintenance of cooperation across various reciprocity mechanisms and interaction patterns.

Fusing Predictive and Large Language Models for Actionable Recommendations In Creative Marketing

The opaqueness of modern digital advertising, exemplified by large platforms such as Meta Ads , raises concerns regarding their control over audience targeting, pricing structures, and ad relevancy assessments. Locked in place by network effects, these natural monopolies attract countless advertisers who rely on subjective intuition, with billions of dollars lost on ineffective social media advertisements. The platforms’ algorithms rely on huge amounts of data unavailable to advertisers, and the algorithms themselves are opaque too, so advertisers often cannot make informed decisions. To promote transparency and help individual advertisers, we first propose novel ways to optimize advertising strategies, predicting click-through rates of novel advertising content based on the content itself. However, advertisers face both opaqueness and a vast abundance of data: a large platform has so many competitor ads that it is hard to derive meaningful insights. Drawing inspiration from the success of large language models (LLM), we propose a system that merges multimodal LLMs and pretrained AI models with an emphasis on digital marketing and advertising data analysis. Leveraging the capabilities of LLMs and incorporating explainability features, including modern text-image models, we aim to improve efficiency and produce synergy between human marketers and AI systems.

The Evolution of Human-AI Collaboration in Cloud Technology: Transforming Industrial Operations

The integration of Artificial Intelligence with cloud technology has revolutionized human-machine collaboration across industrial sectors, transforming traditional workflows and decision-making processes. This technological convergence has particularly impacted healthcare delivery, customer relationship management, and inventory control systems. Organizations implementing AI-cloud solutions have experienced enhanced operational efficiency, improved data processing capabilities, and more sophisticated approaches to resource management. The synergy between human expertise and AI capabilities has created a balanced framework where automated systems complement human judgment, leading to more nuanced decision-making and strategic planning. Implementation strategies focusing on comprehensive change management, workforce development, and security considerations have proven crucial for successful integration. The evolution of these technologies continues to shape the future of industrial operations, promising further advancements in personalization, analytics, and operational excellence.

Reimagining Flipped Learning via Bloom’s Taxonomy and Student–Teacher–GenAI Interactions

This paper explores how generative artificial intelligence (GenAI) technologies, such as ChatGPT 4o and other AI-based conversational models, can be applied to flipped learning pedagogy to achieve enhanced learning outcomes for students. By applying Bloom’s taxonomy to intentionally align educational objectives to the key phases of flipped learning, our study proposes a model for assigning learning activities to pre-class, in-class, and post-class contexts that can be enhanced by the integration of GenAI. In the pre-class phase, GenAI tools can facilitate personalised content delivery, enabling students to grasp fundamental concepts at their own pace. During class, the interactions between students, teacher, and GenAI encourage collaborative learning and real-time feedback. Post-class activities utilise GenAI to reinforce knowledge, provide instant feedback, and support continuous learning through summarisation and content generation. Furthermore, our model articulates the synergies between the three key actors: interactions between students and teachers, learning support provided by GenAI to students, and use of GenAI by teachers to enhance their teaching strategies. These human–AI interactions fundamentally reshape the flipped learning experience, making it more adaptive, engaging, and supportive of the development of 21st-century skills such as critical thinking, collaboration, communication, and creativity.

Human Factors Requirements for Human-AI Teaming in Aviation

The advent of Artificial Intelligence in the cockpit and the air traffic control centre in the coming decade could mark a step-change improvement in aviation safety, or else could usher in a flush of ‘AI-induced’ accidents. Given that contemporary AI has well-known weaknesses, from data biases and edge or corner effects, to outright ‘hallucinations’, in the mid-term AI will almost certainly be partnered with human expertise, its outputs monitored and tempered by human judgement. This is already enshrined in the EU Act on AI, with adherence to principles of human agency and oversight required in safety-critical domains such as aviation. However, such sound policies and principles are unlikely to be enough. Human interactions with current automation in the cockpit or air traffic control tower require extensive requirements, methods, and validations to ensure a robust (accident-free) partnership. Since AI will inevitably push the boundaries of traditional human-automation interaction, there is a need to revisit Human Factors to meet the challenges of future human-AI interaction design. This paper briefly reviews the types of AI and ‘Intelligent Agents’ along with their associated levels of AI autonomy being considered for future aviation applications. It then reviews the evolution of Human Factors to identify the critical areas where Human Factors can aid future human-AI teaming performance and safety, to generate a detailed requirements set organised for Human AI Teaming design. The resultant requirements set comprises eight Human Factors areas, from Human-Centred Design to Organisational Readiness, and 165 detailed requirements, and has been applied to three AI-based Intelligent Agent prototypes (two cockpit, one air traffic control tower). These early applications suggest that the new requirements set is scalable to different design maturity levels and different levels of AI autonomy, and acceptable as an approach to Human-AI Teaming design teams.

Achieving effectiveness of public service in AI-enabled service from public value theory: does human–AI collaboration matters?

PurposeArtificial intelligence (AI) integration into public services addresses long-standing challenges of trust, transparency and citizen satisfaction, yet gaps remain in understanding its role in fostering trust and meeting citizen expectations in complex contexts. This study aims to examine how service accuracy, transparency and citizen trust influence perceived service value and satisfaction, and investigate the role of human−AI collaboration as a moderating mechanism.Design/methodology/approachGrounded in public value theory (PVT), this research uses partial least squares structural equation modeling and necessary condition analysis (NCA) to analyze data collected from Indonesian citizens.FindingsThe results indicate that service accuracy, transparency and citizen trust positively influence perceived service value, with service accuracy as the strongest driver. Perceived service value significantly predicts citizen satisfaction. Human−AI collaboration strengthens the impact of transparency and trust on perceived service value but shows no significant effect on the relationship between service accuracy and perceived service value. The NCA identifies transparency and citizen trust as critical prerequisites for achieving citizen satisfaction, emphasizing their indispensable role in ensuring effective and trustworthy public service delivery.Originality/valueThis study advances PVT by examining the effectiveness of AI-enabled public services, showing how service accuracy, transparency and trust drive perceived value and citizen satisfaction. It highlights transparency and trust as essential for achieving effectiveness and positions human−AI collaboration as a key enabler for optimizing outcomes. These findings offer practical guidance for designing effective, citizen-centered public services.

LLM Fine-Tuning: Concepts, Opportunities, and Challenges

As a foundation of large language models, fine-tuning drives rapid progress, broad applicability, and profound impacts on human–AI collaboration, surpassing earlier technological advancements. This paper provides a comprehensive overview of large language model (LLM) fine-tuning by integrating hermeneutic theories of human comprehension, with a focus on the essential cognitive conditions that underpin this process. Drawing on Gadamer’s concepts of Vorverständnis, Distanciation, and the Hermeneutic Circle, the paper explores how LLM fine-tuning evolves from initial learning to deeper comprehension, ultimately advancing toward self-awareness. It examines the core principles, development, and applications of fine-tuning techniques, emphasizing its growing significance across diverse field and industries. The paper introduces a new term, “Tutorial Fine-Tuning (TFT)”, which annotates a process of intensive tuition given by a “tutor” to a small number of “students”, to define the latest round of LLM fine-tuning advancements. By addressing key challenges associated with fine-tuning, including ensuring adaptability, precision, credibility and reliability, this paper explores potential future directions for the co-evolution of humans and AI. By bridging theoretical perspectives with practical implications, this work provides valuable insights into the ongoing development of LLMs, emphasizing their potential to achieve higher levels of cognitive and operational intelligence.

Accelerated experimental design using a human–AI teaming framework

Accelerated experimental design using a human–AI teaming framework

Understanding the processes of trust and distrust contagion in Human–AI Teams: A qualitative approach

Understanding the processes of trust and distrust contagion in Human–AI Teams: A qualitative approach

Impact of Different Employee–AI Interaction: Instrumental vs. Emotional Support and Gender Differences

Previous studies have shown that employee–AI interactions increase job insecurity and negative behaviors, mainly due to AI’s focus on task assistance while neglecting its role in emotional support. We categorize employee–AI interactions into two types: using AI for emotional support and using AI for instrumental support. We further differentiate their impact on job insecurity and knowledge hiding. Results from a multi-wave survey involving 495 participants showed that using AI for instrumental support increases job insecurity, subsequently leading to knowledge hiding, while using AI for emotional support reduces these detrimental effects. Furthermore, the mitigating effect of emotional support on job insecurity and knowledge hiding is more pronounced for women than men. No gender differences were found in the impact of instrumental support. Our study contributes to the human–AI interaction literature by exploring how the nature of AI interactions influences workplace behavior.

A vision of human–AI collaboration for enhanced biological collection curation and research

Abstract Natural history collections play a crucial role in our understanding of biodiversity, informing research, management, and policy in areas such as biosecurity, conservation, climate change, and food security. However, the growing volume of specimens and associated data presents significant challenges for curation and management. By leveraging human–AI collaborations, we aim to transform the way biological collections are curated and managed, realizing their full potential in addressing global challenges. In this article, we discuss our vision for improving biological collections curation and management using human–AI collaboration. We explore the rationale behind this approach, the challenges faced in data management, general curation problems, and the potential benefits that could be derived from incorporating AI-based assistants in collection teams. Finally, we examine future possibilities for collaborations between human and digital curators and collection-based research.

Bridging the human–AI knowledge gap through concept discovery and transfer in AlphaZero

AI systems have attained superhuman performance across various domains. If the hidden knowledge encoded in these highly capable systems can be leveraged, human knowledge and performance can be advanced. Yet, this internal knowledge is difficult to extract. Due to the vast space of possible internal representations, searching for meaningful new conceptual knowledge can be like finding a needle in a haystack. Here, we introduce a method that extracts new chess concepts from AlphaZero, an AI system that mastered chess via self-play without human supervision. Our method excavates vectors that represent concepts from AlphaZero's internal representations using convex optimization, and filters the concepts based on teachability (whether the concept is transferable to another AI agent) and novelty (whether the concept contains information not present in human chess games). These steps ensure that the discovered concepts are useful and meaningful. For the resulting set of concepts, prototypes (chess puzzle-solution pairs) are presented to experts for final validation. In a preliminary human study, four top chess grandmasters (all former or current world chess champions) were evaluated on their ability to solve concept prototype positions. All grandmasters showed improvement after the learning phase, suggesting that the concepts are at the frontier of human understanding. Despite the small scale, our result is a proof of concept demonstrating the possibility of leveraging knowledge from a highly capable AI system to advance the frontier of human knowledge; a development that could bear profound implications and shape how we interact with AI systems across many applications.

Различие между паттернами естественного и искусственного интеллекта: от микро- к макроуровню

Introduction. The convergence between brain science and artificial intelligence technologies makes the relationship between human thinking and AI systems relevant. Theoretical analysis. The formation of artificial and natural intelligences takes place in opposite vectors. In the first case, the formation of thought operations occurs according to the logic of Integration. Discrete units of the lower level, “words” of computer language, are combined into larger units of the higher level – blocks. In the second case, the development of thought operations, on the contrary, takes place according to the logic of Differentiation. The original syncretic unity of thought-forms is divided into discrete units – words with definite meanings. There is a similar logic: the differentiation of syncretic unity into discrete units (rather than the integration of discrete units into blocks) is tracked both in the development of thinking in children and in the evolution of languages in the history of mankind. Conclusion. In line with evolutionary epistemology, the conclusion is made, that the formation of the initial patterns of natural intelligence takes place not inside the “apparatus” of the brain, but in the process of interaction between the body and the external environment. This is consistent with the concept of intentionality of Phenomenological philosophy.

The impact of human interviewers versus AI interviewers on applicant self‐efficacy and offer acceptance

The impact of human interviewers versus AI interviewers on applicant self‐efficacy and offer acceptance

Enablers and inhibitors of AI assimilation in hiring: mitigating the effects of inhibitors through human–AI collaboration

PurposeMost prior studies have primarily investigated AI adoption, with less attention given to AI assimilation in human resource management (HRM). Additionally, prior studies often lack empirical verification of the extent to which human–AI collaboration might alleviate challenges and promote AI assimilation in the HRM context. Thus, this study aims to explore AI assimilation in recruitment with a balanced view that identifies both enabling and inhibiting factors while examining the role of human–AI collaboration in mitigating the effects of inhibiting factors.Design/methodology/approachWe used a mixed-method approach. Using an open-ended survey questionnaire approach and collecting data from 26 HR professionals, we identified five factors, namely, AI competency, recruitment agility, AI opacity, AI empathy and human–AI collaboration, potentially impacting AI assimilation. Thereafter, drawing from the enabler–inhibitor perspective, we theorize that AI competency and recruitment agility are the enablers, whereas AI opacity and AI empathy are the inhibitors of an organization’s efforts to assimilate AI in recruitment practices. We tested our proposed model by collecting data from 309 HR professionals.FindingsThe findings showed that both enablers, AI competency and recruitment agility, significantly influence AI assimilation; however, both inhibitors, AI opacity and AI empathy, are non-significant for AI assimilation. While looking into the reasons for these non-significant effects, we observed that the interaction term between AI empathy and human–AI-collaboration as well as between AI opacity and human–AI-collaboration both had significant effects on AI assimilation. These interaction effects suggest that human–AI collaboration mitigates the constraining impact of both inhibitors.Originality/valueDrawing from the enabler–inhibitor perspective and by empirically testing our proposed model, this paper significantly contributes to the IS literature. Our study not only identifies factors that promote and inhibit AI assimilation in the context of HRM practices but also reveals how human–AI collaboration may mitigate the effects of inhibitors. Our findings suggest that organizations should have a collaborative recruitment environment where AI handles repetitive tasks, and humans focus on roles requiring emotional intelligence. This approach enhances the integration of AI-powered tools, addresses AI assimilation inhibitors and optimizes recruitment effectiveness.

Human–AI Enhancement of Cyber Threat Intelligence

This study proposes a human-AI collaboration to model the landscape of cyber threat intelligence (CTI) and use it to detect suspicious communication indicating impending cybersecurity incidents. We show how the collaboration between cybersecurity experts and AI-based text-classification methods develops an understanding of professional hackers and helps detect cybersecurity threats more accurately. The human-AI collaboration rests on a Reciprocal Human–Machine Learning (RHML) model, in which a human expert and a machine interact repeatedly over time and simultaneously continually learn to detect professional hackers. Two cybersecurity experts employed qualitative data analysis and worked with RHML software assistance to classify 6651 messages from an online hackers’ forum. We discovered an improvement, over time, of both the detection accuracy and the experts’ understanding of the threat landscape as represented by their concept maps. In particular, the concept map refers to the hacker’s capabilities, intent, and behaviour to define the threat landscape needed for professional detection, in contrast to amateur hackers. We believe this approach may ultimately lead to a more robust and proactive cybersecurity posture and translate into operational advantages in the field of CTI.

Designing High-Impact Experiments for Human–Autonomy / AI Teaming

The potential to create autonomous teammates that work alongside humans has increased with continued advancements in AI and autonomous technology. Research in human–AI teams and human–autonomy teams (HATs) has seen an influx of new and diverse researchers from human factors, computing, and teamwork, yielding one of the most interdisciplinary domains in modern research. However, the HAT domain’s interdisciplinary nature can make the design of research, especially experiments, more complex, and new researchers may not fully grasp the numerous decisions required to perform high-impact HAT research. To aid researchers in designing high-impact experiments, this article itemizes four initial decision points needed to form a HAT experiment: deciding on a research question, deciding on a team composition, deciding on a research environment, and deciding on data collection. For each decision point, this article discusses these decisions in practice, providing related works to guide researchers toward different options available to them. These decision points are then synthesized through actionable recommendations to guide future researchers. The contribution of this article will increase the impact and knowledge of HAT experiments.

Human insight vs. AI in extreme financial crises: the case for human decision-making

Human insight vs. AI in extreme financial crises: the case for human decision-making

Collaborative human-AI risk annotation: co-annotating online incivility with CHAIRA

Introduction. Collaborative human-AI annotation is a promising approach for various tasks with large-scale and complex data. Tools and methods to support effective human-AI collaboration for data annotation are an important direction for research. In this paper, we present CHAIRA: a Collaborative Human-AI Risk Annotation tool that enables human and AI agents to collaboratively annotate online incivility. Method. We leveraged Large Language Models (LLMs) to facilitate the interaction between human and AI annotators and examine four different prompting strategies. The developed CHAIRA system combines multiple prompting approaches with human-AI collaboration for online incivility data annotation. Analysis. We evaluated CHAIRA on 457 user comments with ground truth labels based on the inter-rater agreement between human and AI coders. Results. We found that the most collaborative prompt supported a high level of agreement between a human agent and AI, comparable to that of two human coders. While the AI missed some implicit incivility that human coders easily identified, it also spotted politically nuanced incivility that human coders overlooked. Conclusions. Our study reveals the benefits and challenges of using AI agents for incivility annotation and provides design implications and best practices for human-AI collaboration in subjective data annotation.

AI and Human-AI Collaboration in Enterprise Integration and Document Automation

The integration of artificial intelligence with human expertise is transforming enterprise document processing and system integration workflows. This synergistic relationship creates a powerful framework where AI handles routine, high-volume tasks while humans manage exceptions, provide oversight, and handle complex decision-making. Organizations implementing collaborative approaches experience significant improvements in processing speed, accuracy, compliance, and operational elasticity compared to either purely manual or fully automated alternatives. These hybrid systems continuously improve through structured feedback mechanisms, with human experts training AI to handle increasingly complex scenarios while maintaining appropriate ethical guardrails. Despite implementation challenges including bias concerns, employee resistance, data privacy considerations, and integration complexity, successful deployments fundamentally reimagine organizational workflows rather than simply automating existing processes. Future developments in hyperautomation, AI-augmented APIs, conversational interfaces, and digital twins promise to further enhance this partnership model, enabling unprecedented levels of operational excellence while preserving critical human judgment for scenarios requiring contextual understanding, ethical reasoning, and novel situation handling.