Abstract
Introduction: Toward Humane Artificial Intelligence (AI): Aligning AI with Human Values and Well-Being
The rapid integration of AI into various facets of daily life—ranging from healthcare and education to transportation and personal digital assistants—has ushered in transformative possibilities. 1 However, this proliferation also brings forth complex psychological, social, and ethical challenges. As AI systems increasingly influence human behavior and decision-making, there is a pressing need to ensure that these technologies are designed and deployed in ways that prioritize human values and well-being. 2
Traditional AI development has often emphasized efficiency and task optimization, sometimes at the expense of human-centric considerations. This approach can inadvertently perpetuate biases, undermine autonomy, and erode trust. 3 For instance, a recent study has demonstrated that AI models like ChatGPT can exhibit human-like cognitive biases, such as overconfidence and confirmation bias, which may amplify flawed decision-making processes. 4 These findings underscore the necessity of reorienting AI development toward frameworks that emphasize ethical alignment with human values.
Human-Centered Artificial Intelligence (HCAI) 5 emerges as a paradigm that seeks to embed ethical principles, transparency, and user well-being into the core of AI systems. A systematic literature review by Tahaei et al. 6 highlights the growing emphasis on governance, fairness, and explainability within HCAI research. However, the study also notes a relative underrepresentation of themes such as privacy, security, and human flourishing, suggesting areas for further exploration.
The ethical management of human–AI interaction is pivotal in aligning AI with human values. From a sociotechnical perspective, understanding the interplay between technological systems and social contexts is essential. Research indicates that ethical AI governance should encompass considerations of accountability, transparency, and fairness to ensure that AI systems support human autonomy and societal well-being.7,8
In the realm of mental health, AI interventions offer both opportunities and ethical dilemmas.1,9 While AI can enhance accessibility to mental health resources, it also raises concerns regarding privacy, informed consent, and the potential for algorithmic bias. A systematic review by Saeidnia et al. 10 emphasizes the importance of addressing these ethical considerations to ensure responsible implementation and positive outcomes in AI-driven mental health interventions.
Moreover, the concept of emotional AI, systems designed to recognize and respond to human emotions—introduces additional ethical complexities. The integration of emotional intelligence into AI can lead to phenomena such as “human alienation,” where users may experience diminished agency and altered interaction paradigms. 11 These developments necessitate a careful examination of the psychological impacts of AI and the establishment of safeguards to protect human emotional well-being.
Furthermore, the illusion of social presence, the feeling of being in the company of another intelligent being, can blur the boundaries between authentic empathy and synthetic simulation.12,13 While platforms like Meta have experimented with socially interactive AI profiles, their inability to replicate nuanced human cognition and ethical discretion reveals a critical gap between perception and reality. This dissonance can create psychological disorientation, especially when AI avatars provide emotionally charged feedback or advice without accountability or ethical grounding. 2
To address these multifaceted challenges, interdisciplinary collaboration is essential. Integrating insights from psychology, sociology, ethics, and computer science can inform the creation of AI systems that are not only technologically advanced but also aligned with human values and societal needs. By prioritizing ethical considerations from the outset, developers and policymakers can mitigate potential risks and unintended consequences associated with AI deployment.
This special issue investigates the psychological, social, and ethical dimensions of designing AI systems that genuinely prioritize human values and well-being. It brings together interdisciplinary research across three tightly interwoven domains.
First, it examines how individual psychological factors, such as belief systems, cultural frameworks, and levels of digital and algorithmic literacy, profoundly shape human interactions with AI. These differences influence not only how users interpret AI behavior but also how they assign trust and agency in digital environments.
Second, the issue addresses transparency and literacy as foundational pillars for ethically sound AI integration. It highlights a critical yet underexplored concern: the uneven distribution of transparency’s benefits, which are often accessible only to those with higher algorithmic awareness—thereby reinforcing existing digital divides.
Third, the issue introduces and elaborates on emerging theoretical constructs like System 0 and Psychomatics, which reconceptualize AI as a form of cognitive extension. These frameworks challenge traditional boundaries between human and machine intelligence, while also illuminating essential asymmetries in their cognitive architectures.
The Psychological Frontiers of Human–AI Integration: Key Themes from the Special Issue
This special issue explores the evolving psychological dynamics between humans and AI across three critical domains. Rather than presenting isolated studies, these papers collectively map the emerging contours of our AI-mediated future, revealing both promising synergies and concerning tensions.
Psychological factors influencing AI reception
Understanding how individual differences shape human engagement with AI is crucial for designing ethically aligned and psychologically effective systems. A compelling example comes from the paper by Ahn et al. (“Effects of consumers’ belief in a just world on artificial intelligence recommendations: Mediating effects of perceived benevolence and selfishness”) examining the influence of belief in a just world (BJW), the conviction that people get what they deserve, on how users perceive AI versus human recommendation agents.
The study uncovered a striking interaction: individuals high in BJW evaluated recommendations from human agents more favorably, whereas those low in BJW preferred AI-generated suggestions. This divergence was mediated by perceived benevolence and selfishness. High-BJW individuals saw human agents as more benevolent and less selfish, while low-BJW individuals viewed them as the opposite, more selfish and less benevolent. Notably, these perceptions remained stable for AI agents regardless of the user’s BJW level.
These insights offer a dual contribution: theoretically, they identify BJW as a critical psychological factor influencing AI’s persuasive power; practically, they suggest marketers and system designers can optimize AI recommendation strategies by tailoring them to users’ psychological profiles. As AI continues to supplant human agents in consumer contexts, aligning system design with such cognitive traits will be essential to fostering effective human–AI interaction and maximizing persuasive outcomes.
Cultural context further shapes AI reception, particularly in the case of automated vehicle (AV) technology. A comparative study of Italian and Chinese consumers by Bruno et al. (“Attitudes towards the use of conditional automated vehicles in the Technology Acceptance Model framework: evidence from an Italian sample”) revealed key differences in how AVs are perceived in Italy, highlighting the interplay between trust, utility, and privacy. For Italian participants:
Perceived utility had a direct and dominant effect on intention to use AVs, bypassing the mediating role of trust that was expected based on existing models. Privacy concerns significantly reduced adoption willingness, suggesting a culturally rooted skepticism toward data handling by manufacturers. Perceived ease of use exerted limited influence, showing isolated effects compared to other factors.
Moreover, the Italian sample exhibited a tightly interwoven relationship between trust, attitudes, and behavioral intention, suggesting an affect-driven acceptance model in which early trust plays a pivotal role. These findings underscore the necessity for AV manufacturers and policymakers in Europe to prioritize transparent data governance and early trust-building interventions to facilitate adoption.
Bridging these psychological and cultural dimensions is the emerging “Digital Wellness” framework, discussed in the paper by Laffier et al. (“A conceptual Framework for Psychological Digital Wellness and Artificial Intelligence”), which proposes a proactive skill-based approach to AI engagement. Drawing from a systematic review of the literature, this model identifies both the benefits of AI (e.g., enhanced accessibility, cognitive stimulation, and personal empowerment) and its risks (e.g., anxiety, social detachment, overdependence).
The framework emphasizes four psychological skill domains crucial for resilient and healthy AI interaction:
Mental health literacy: equipping users to seek and utilize AI-based support. Emotional intelligence: encompassing self-awareness, self-regulation, and social competence. Mindfulness: fostering present-moment focus and intentional engagement. Critical thinking: enabling users to evaluate AI content autonomously and avoid manipulation.
The authors argue these skills should be cultivated across the lifespan, from school-age interventions like mindfulness training to the design of AI interfaces that promote reflective, rather than passive, use. This perspective repositions users not as passive recipients of AI outputs but as active participants equipped with the psychological tools necessary for humane and empowered technology use.
Together, these studies offer am empirically grounded understanding of the psychological, cultural, and developmental factors shaping human-AI interaction. They call for personalized, ethically conscious design strategies that honor both cognitive diversity and cultural specificity, essential pillars for building trust, ensuring engagement, and fostering digital well-being in an AI-mediated world.
Transparency, literacy, and ethical considerations in AI use
As AI systems become embedded in everyday life, transparency and algorithmic literacy emerge as foundational pillars for ethical and effective human–AI interaction. Experimental evidence demonstrates that platforms offering explainability and user control significantly boost users’ perceptions of transparency, legitimacy, and satisfaction. However, these benefits are not equitably distributed, users with higher algorithmic literacy derive greater advantages, revealing a new layer of digital inequality in the AI era.
In educational settings, the growing presence of generative AI raises critical concerns. The paper by Carruba et al. (“A grade for A.I.: A Study on School Teachers’ Ability to Identify Assignments Written by Generative Artificial Intelligence”) examining teachers’ ability to distinguish between human-written and AI-generated student work reveals both strengths and underlying biases. While teachers were generally more accurate in detecting AI-generated content than human-authored assignments, experienced educators also exhibited a form of “AI bias”, a tendency to misidentify human work as machine-produced, perhaps reflecting heightened suspicion in an age of algorithmic authorship.
Despite these challenges, experienced teachers tended to award higher grades to authentic, human-generated work. This suggests a latent sensitivity to genuine student effort, even in the absence of perfect detection tools. The findings underscore the need to move beyond a punitive focus on AI “cheating” and toward constructive AI integration, harnessing AI to support, rather than undermine, educational goals. Human-centered design approaches that account for individual expertise, user needs, and predispositions will be vital for embedding AI meaningfully into learning environments.
Another experimental paper by Moon et al. (“The effects of Explainability and User Control on Algorithmic Transparency: The Moderating Role of Algorithmic Literacy”) deepens this insight by exploring how transparency mechanisms, explainability and user control, affect user perception on short-form video platforms. In a 2 × 2 × 2 factorial design involving 240 participants, researchers assessed how these mechanisms shaped perceived transparency, platform legitimacy, and user satisfaction.
The results revealed a significant three-way interaction: when neither explainability nor control was present, algorithmic literacy had no effect on user evaluations. However, when even one of these features was available, users with higher algorithmic literacy gained disproportionately greater benefits. Drawing from Self-Determination Theory and Procedural Justice Theory, the study shows that explainability enhances users’ sense of competence, while user control bolsters autonomy, together contributing to stronger feelings of legitimacy and satisfaction.
This research exposes a subtle but powerful form of technological inequity: individuals with lower algorithmic literacy may struggle to reap the benefits of even well-designed transparency features. To mitigate this, the authors advocate for adaptive transparency models, including educational content and interface personalization based on user literacy. Policymakers are called upon to enact regulations that not only mandate algorithmic explainability and user control, but also promote public literacy in algorithmic systems, ensuring more equitable outcomes in AI-mediated environments.
Complementing these applied concerns, a systematic review paper by Marchetti et al. (“AI and the Illusion of Understanding: A Systematic Review of Theory of Mind and Large Language Models”) on Theory of Mind (ToM) in Large Language Models (LLMs) raises deeper epistemological questions. While LLMs like GPT-4 perform admirably on first-order false belief tasks, they falter in more complex forms of social reasoning, such as second-order beliefs and recursive mental state attribution, areas where humans still excel.
The review exposes an “illusion of understanding”: LLMs appear intelligent but lack the developmental and cognitive architectures required for genuine social cognition. Many evaluations of LLM ToM rely on overly simplified, linguistically tailored tests that strip away the ambiguity and nuance central to human social life. This reveals a troubling methodological bias and suggests that current models may be performing pattern recognition rather than demonstrating true understanding of mental states.
The authors call for more ecologically valid testing protocols and interdisciplinary research, arguing that the development of artificial social cognition must be informed by advances in developmental psychology, philosophy of mind, and cognitive science. Such cross-pollination could better define the boundaries and potential of artificial ToM, avoiding premature assumptions about AI’s cognitive capacities.
Together, these studies underscore the need for psychologically informed, ethically grounded AI systems that prioritize human autonomy, comprehension, and agency. Whether in education, media platforms, or foundational cognitive domains like ToM, the message is clear: successful AI integration depends not just on technical sophistication, but on thoughtful, inclusive design that centers the human experience.
As AI reshapes our environments, developers, educators, and policymakers must collaborate to ensure that individual differences are respected, transparency is real rather than symbolic, and humans remain the primary meaning-makers in an increasingly algorithmic world.
The psychological frontiers of AI: Understanding the mind behind the machine and how to exploit it
In an era where AI increasingly shapes our cognitive landscape, we find ourselves at a profound paradox: we have created systems of unprecedented capability that we do not fully understand. 14 As Anthropic CEO Dario Amodei 15 candidly acknowledged, even the architects of today’s most advanced AI systems cannot fully interpret their internal operations, these sophisticated creations remain, in many ways, psychological “black boxes.”
This special issue confronts this paradox head-on, presenting new frameworks that illuminate the hidden psychological dimensions of human–AI interaction. Rather than treating AI merely as technological artifacts, our contributors explore these systems as cognitive entities with distinct psychological properties that both complement and challenge human cognition.
To penetrate this black box, Riva et al. in their paper (“Psychomatics—A Multidisciplinary Framework for Understanding Artificial Minds”) present “Psychomatics”, a multidisciplinary framework integrating cognitive science, linguistics, and computer science. Moving beyond surface-level behavioral comparisons, this approach reveals fundamental differences in how humans and AI systems develop and use language.
While LLMs create impressive syntax-semantics maps through Transformer algorithms, they operate through distinct mechanisms—self-attention (syntagmatic processing) and cross-attention (associative processing)—that navigate linguistic structures without human-like understanding. LLMs adhere to Grice’s Cooperative Principle in producing relevant responses but struggle with implicit meanings, cannot generate truly novel concepts, and lack the ability to verify truth through experience.
These differences are not mere technical details but reveal profound limitations in AI’s capacity for genuine meaning-making. Unlike human cognition, which develops through embodied social experiences and draws meaning from multiple sources, AI systems respond based on probabilistic patterns without forming authentic intentions.
This theoretical understanding finds striking validation in the paper by Gerardini et al. (“Brainstorming with a Collaborative Platform or a Generative Artificial Intelligence Tool: An Exploratory Study”) comparing AI-based and traditional creative tools. When examining DALL-E against conventional brainstorming platforms, authors discovered that AI-generated visualizations sparked significantly more positive emotions, stronger aesthetic experiences, and enhanced feelings of creativity, with 83% of participants preferring the AI system.
Particularly noteworthy was DALL-E’s effectiveness in online settings, where its ability to provide vivid, intuitive visualizations with minimal cognitive effort appeared to bridge communication gaps in distributed work. The emotional engagement stimulated by these AI-generated representations fostered intrinsic motivation and flow states conducive to creativity.
This study is relevant for the novel conceptualization of AI as “System 0”, a non-biological cognitive layer that operates alongside Kahneman’s dual-process model of human thinking. 16 As discussed by Chiriatti et al. in their paper (“System 0: Transforming Artificial Intelligence into a Cognitive Extension”) unlike Systems 1 (intuitive) and 2 (analytical), System 0 functions as an informational preprocessor, employing algorithmic pattern recognition to process datasets at scales beyond human capacity.
This framework represents a fundamental shift in how we understand AI’s relationship to human cognition. When evaluated against Heersmink’s criteria for cognitive extension, modern AI systems demonstrate remarkable integration through their reliability, trust capabilities, and capacity to transform mental functions. They extend our information processing, enhance creativity, and augment memory in ways previously unimaginable.
Yet this integration presents a critical paradox: as AI extends our cognitive reach, it simultaneously risks constraining our intellectual development through “sycophancy”—the tendency to affirm rather than challenge user perspectives. This creates what Riva 13 terms the “comfort-growth paradox,” wherein AI’s very helpfulness may undermine its potential to stimulate genuine intellectual advancement.
To address these challenges, the paper proposes different evidence-based frameworks for effective human–AI cognitive integration: Enhanced Cognitive Scaffolding, which promotes progressive autonomy; Symbiotic Division of Cognitive Labor, strategically allocating tasks based on comparative strengths; Dialectical Cognitive Enhancement, countering AI sycophancy through productive epistemic tension; Agentic Transparency and Control, ensuring users understand and direct AI influence; Expertise Democratization, breaking down knowledge silos; Social–Emotional Augmentation, addressing affective dimensions of cognitive work; and Duration-Optimized Integration, managing the evolving human–AI relationship over time.
Conclusion: Psychological Dimensions as the Foundation for Humane AI
As we stand at the intersection of technological innovation and human experience, this special issue reveals a critical insight: the future of AI depends not merely on technical advancement, but on our understanding of the psychological dimensions that shape human–AI interaction. The papers in this collection illuminate a path toward AI systems that genuinely enhance human flourishing rather than merely optimizing technical metrics.
Three key themes emerge from our exploration. First, the psychological characteristics of users fundamentally influence AI reception and effectiveness. Whether through belief systems like BJW, cultural frameworks as seen in Italian versus Chinese AV adoption, or varying levels of algorithmic literacy, individual differences profoundly shape how humans engage with AI technologies. These differences are not peripheral concerns but central determinants of AI’s societal impact. Successful implementation requires systems designed with psychological diversity in mind—moving beyond one-size-fits-all approaches toward contextually sensitive, adaptable interfaces that respect cognitive and cultural variations.
Second, transparency and literacy represent interconnected imperatives for ethical AI integration. When users understand algorithmic processes and maintain meaningful control, they experience greater satisfaction and trust. However, our findings reveal a troubling “digital literacy divide,” where transparency benefits accrue disproportionately to those already technologically sophisticated. This inequality demands both educational interventions and intelligent interface design that adapts to users’ varying literacy levels, ensuring AI’s benefits are democratically distributed rather than reinforcing existing power imbalances.
Third, frameworks like System 0 and Psychomatics offer crucial theoretical scaffolding for understanding AI cognition and its relationship to human thought. These models reveal that while AI extends certain cognitive capacities, it operates through fundamentally different processes than human cognition, without embodied experience, causal understanding, or authentic intentions. This difference creates both opportunities for complementary partnership and risks like the “comfort-growth paradox,” where AI’s tendency toward agreement may undermine intellectual development.
These insights converge on an essential principle: Humane AI requires maintaining humans as primary agents in an increasingly mediated world. The empirical evidence from DALL-E studies demonstrates that when properly designed, AI can stimulate creativity, positive affect, and flow states, but these benefits emerge when AI serves as cognitive extension rather than replacement, enhancing distinctly human capacities for meaning-making and creative expression.
Moving forward, we identify four priorities for researchers, developers, and policymakers. First, we must incorporate psychological assessment into AI design processes, testing systems against diverse cognitive styles and cultural frameworks. Second, we should develop adaptive transparency mechanisms that democratize understanding across varying literacy levels. Third, we need interfaces that counter sycophancy through productive epistemic tension, challenging users rather than merely affirming them. Finally, we must establish ethical frameworks that preserve human agency while leveraging AI’s unique capabilities.
The path to humane AI is neither purely technical nor exclusively humanistic, it requires sustained interdisciplinary dialogue between cognitive science, psychology, ethics, and computer science 17 . By understanding the psychological dimensions of human-AI interaction, we can develop technologies that respect human autonomy, enhance capabilities, promote wellbeing, and reflect our highest values. The papers in this special issue represent crucial steps in this journey, illuminating not just what AI can do, but how it can do so in service of genuine human flourishing. Our collective challenge now is to ensure that as AI systems grow increasingly sophisticated, they remain fundamentally aligned with the psychological needs, ethical principles, and social contexts that make us human.