Integrating Psychological Theories into AI Design: A Narrative Review of Human-Centred Artificial Intelligence (HCAI) ()
1. Introduction
Artificial Intelligence (AI) has become deeply embedded within contemporary society, reshaping sectors such as healthcare, finance, education, and daily personal activities (Williams et al., 2022). Its rapid evolution and widespread adoption have significantly transformed how people interact with technology, influencing everything from automated customer service to advanced medical diagnostics and personal digital assistants (Sinha, 2024). However, despite these technological advancements, many AI systems continue to suffer from shortcomings related to user acceptance, usability, cognitive alignment, and trust (Cabrera-Sánchez et al., 2020). These deficiencies often result from a limited consideration of psychological and behavioural theories in AI system design, suggesting a critical gap between AI’s technological capabilities and its psychological integration into human lives (Kirk et al., 2025).
In response to this gap, the concept of Human-Centred Artificial Intelligence (HCAI) has gained considerable traction. HCAI seeks to place human psychological needs, cognitive processes, emotional responses, and ethical considerations at the core of AI development (Gomaa & Mahdy, 2024). It recognises that for AI technologies to be genuinely effective, widely adopted, and ethically responsible, they must align closely with human psychological attributes. This systematic review addresses this need by exploring how recent research integrates psychological theories into AI design, thereby promoting trustworthiness, usability, emotional resonance, and cognitive efficiency (Nakao et al., 2022).
Psychological integration into AI involves multiple aspects, with several key areas identified as particularly influential. Firstly, the Theory of Mind (ToM) is crucial for enabling AI systems to understand and predict human behaviours by interpreting mental states such as beliefs, intentions, and emotions (Langley et al., 2022). By embedding ToM principles, AI systems can more accurately anticipate user needs, improving interaction fluency and reducing misunderstandings. Secondly, the emotional dimensions of AI, encompassing empathy and emotional intelligence, have emerged as central to fostering deeper trust and rapport between users and AI systems (Zhang & Wang, 2024). Emotional intelligence allows AI to recognise, interpret, and respond appropriately to user emotions, significantly enhancing the quality and perceived value of interactions, especially in emotionally sensitive contexts such as mental health and interpersonal communications (Vicci, 2024).
Furthermore, cognitive load theory (CLT) plays a vital role in AI interaction design by addressing the human cognitive capacity to process information effectively (Baxter et al., 2025). AI systems designed with cognitive load principles minimise unnecessary mental effort, enhancing decision-making efficiency and user satisfaction (Ayres & Paas, 2012). This is particularly relevant in professional and high-stakes environments, where efficient cognitive proccessing directly affects performance outcomes. User-Centred Design further reinforces the importance of aligning AI development with human factors, advocating for iterative design processes that actively involve end-users in shaping system functionalities and interfaces (Schmager et al., 2025).
2. Methods
This review followed a narrative synthesis approach, focusing on conceptual integration rather than exhaustive systematic coverage. The aim was to explore how psychological theories have been applied within the design of Human-Centred Artificial Intelligence (HCAI), highlight recurring themes, and identify conceptual gaps. Unlike systematic reviews, which adhere to rigid protocols for literature inclusion, a narrative approach allowed for greater flexibility in selecting and interpreting studies across diverse disciplines, including psychology, computer science, and human-computer interaction (Braun & Clarke, 2006).
The review process involved three broad stages:
Exploratory Scanning—An initial mapping of recent research (2020-2025) to identify recurring psychological constructs in AI design.
Focused Selection—Prioritizing studies that explicitly discussed psychological theories (e.g., Theory of Mind, cognitive load, emotional intelligence) in relation to AI usability, trust, or ethics.
Thematic Synthesis—Organizing selected literature into key thematic domains, synthesizing conceptual linkages, and critically evaluating implications for future HCAI development. The approach emphasized depth and interpretive synthesis over comprehensive retrieval, enabling a more integrative account of current theoretical and applied perspectives.
3. Search Strategy
The literature search was narrative and iterative rather than systematic. Relevant studies were identified through searches of multidisciplinary databases (e.g., PsycINFO, PubMed, Web of Science, and Google Scholar) and by examining reference lists of influential papers. Keywords such as “psychological theories”, “artificial intelligence design”, “Theory of Mind”, “trust”, “emotional intelligence”, “cognitive load”, “user-centred design”, “mental health AI”, and “ethical AI” were used in varying combinations.
Given the interpretive nature of narrative reviews, inclusion was guided by relevance and conceptual contribution, rather than rigid eligibility criteria. Preference was given to peer-reviewed journal articles and conference proceedings published between 2020 and 2025, though seminal earlier works were also referenced when theoretically necessary (e.g., cognitive load theory, thematic analysis). This flexible search strategy enabled the incorporation of diverse disciplinary perspectives, ensuring that the synthesis reflects both psychological foundations and technological applications of HCAI.
4. Results
The narrative review identified relevant articles that met the inclusion criteria. Through thematic analysis, six prominent themes emerged: (1) Theory of Mind (ToM) and Mental Models, (2) Trust, Empathy, and Emotional Intelligence, (3) Cognitive Load and Adaptation, (4) Human-Centred and User-Centred AI Design, (5) AI in Mental Health Applications, and (6) Ethical Considerations.
4.1. Theory of Mind (ToM) and Mental Models
Theory of Mind (ToM) emerged prominently as a foundational psychological theory informing AI design. Studies consistently demonstrated the significance of ToM in enabling AI systems to interpret and predict human behaviours accurately (Williams et al., 2022). This capability significantly enhances AI’s effectiveness in collaborative contexts, enabling smoother interactions and more intuitive user experiences. Bara, Wang, and Chai’s (2021) MindCraft study illustrated that AI systems incorporating ToM could dynamically infer and adapt to user intentions, dramatically reducing communication misunderstandings and enhancing collaborative efficiency (Li et al., 2025). Similarly, research by Liao et al. (2020) showed adaptive dialogue systems using ToM significantly improved user trust and satisfaction by providing contextually appropriate responses. ToM integration also helps manage user expectations, aligning AI outputs closely with user mental models and thus fostering user confidence and trust (Li et al., 2025). Collectively, these findings underscore the crucial role of ToM in bridging the gap between AI functionalities and user expectations, highlighting its value in creating more psychologically attuned and user-friendly AI systems (Kosinski, 2024).
4.2. Trust, Empathy, and Emotional Intelligence
Trust was consistently identified as central to user acceptance of AI technologies, strongly linked to empathy and emotional intelligence capabilities in AI systems (OpenAI, 2024). Studies found that AI demonstrating emotional awareness and empathetic responsiveness could significantly enhance user trust, particularly in emotionally sensitive interactions such as healthcare, counseling, and personal support applications (Vicci, 2024). Liu and Sundar (2018) provided empirical evidence that emotionally intelligent AI, such as empathetic chatbots, notably increased perceived reliability and user trust. However, researchers also cautioned against emotional misalignment or overly anthropomorphic behaviours, as these could create user discomfort or mistrust. Becker et al. (2023) illustrated the potential pitfalls of inappropriate emotional expressions, highlighting the necessity for carefully balanced emotional intelligence to maintain trust and interaction quality. Thus, emotional intelligence in AI must be precisely calibrated to reflect authentic emotional cues without overstepping psychological boundaries. Ensuring emotional congruence and contextually appropriate interactions emerged as essential for maintaining user trust and engagement, highlighting the nuanced relationship between emotional intelligence and user perceptions in AI interactions (Sousa et al., 2023).
4.3. Cognitive Load and Adaptation
The theme of cognitive load adaptation consistently emerged as critical to enhancing AI system usability. Cognitive Load Theory (CLT) emphasizes minimizing unnecessary cognitive demands to optimize user performance and satisfaction. Studies demonstrated that AI systems designed with cognitive load principles significantly reduced users’ mental fatigue, enhanced decision-making efficiency, and improved overall task performance (Sun et al., 2025). Liefooghe and Van Maanen (2023) highlighted that personalized cognitive load adaptations, such as simplifying task instructions or providing cognitively aligned outputs, led to marked improvements in user efficiency and cognitive comfort. Riefle & Weber (2022) similarly found that AI systems adjusting output complexity according to users’ cognitive styles enhanced interaction satisfaction and performance outcomes. Additionally, systems implementing progressive disclosure techniques, presenting information incrementally based on user interactions, effectively reduced cognitive overload, promoting sustained user engagement and task effectiveness (Moreno et al., 2023). The reviewed literature strongly supports cognitive load management as a fundamental consideration for designing AI interactions that respect human cognitive limitations and enhance user experiences (Latif et al., 2024).
4.4. Human-Centred and User-Centred AI Design
Human-centred and user-centred design principles were consistently advocated as essential for effective AI integration. These approaches emphasize involving users directly in the design process, ensuring AI technologies closely align with actual user needs, preferences, and contextual nuances (Nakao et al., 2022). Studies showed that user involvement through participatory design methodologies significantly enhanced AI usability, acceptance, and practical relevance. Torkamaan et al. (2024) demonstrated that iterative, user-driven refinements effectively improved AI functionalities and user satisfaction, underscoring the value of continuous user feedback in AI development. AI personalization and adaptive interfaces were identified as critical elements of human-centred design, enhancing user engagement by tailoring system responses to individual preferences and behavioural patterns (Bier et al., 2024). Furthermore, accessibility considerations were highlighted, with inclusive design principles ensuring AI systems cater effectively to diverse user populations, including those with cognitive or physical disabilities. The findings collectively reinforce the necessity of user-centric methodologies, advocating for AI development that actively incorporates user perspectives and iterative refinements to create genuinely usable and inclusive systems (Ayeni et al., 2024).
4.5. AI in Mental Health Applications
AI’s application in mental health contexts emerged as a promising yet ethically sensitive domain. Studies demonstrated significant potential for AI-driven psychological support systems, particularly emotionally intelligent chatbots delivering Cognitive Behavioral Therapy (CBT), mindfulness interventions, and stress reduction strategies (Casu et al., 2024). Thieme et al. (2023) provided evidence of AI’s effectiveness in enhancing user accessibility and reducing stigma associated with mental health support. Users reported increased comfort and therapeutic adherence when interacting with emotionally responsive AI, reflecting enhanced trust and perceived support. Springer and Whittaker (2020) similarly noted the benefits of transparent, emotionally calibrated AI interactions in reducing user anxiety and promoting sustained engagement. However, ethical concerns surrounding emotional authenticity, transparency, and the risk of emotional manipulation were prominent (Xu et al., 2023). Researchers emphasized the necessity of clearly defined emotional boundaries, ensuring users are accurately informed about AI capabilities and limitations. The literature highlights that while AI holds considerable promise for mental health support, maintaining rigorous ethical standards and transparent communication is essential for responsible implementation (Saeidnia et al., 2024).
4.6. Ethical Considerations
Ethical considerations emerged as critical to the responsible integration of psychological theories into AI design, focusing on fairness, accountability, transparency, and user autonomy. Studies consistently highlighted the risks associated with algorithmic biases, privacy violations, and emotional manipulation, emphasizing the necessity of comprehensive ethical frameworks and transparency mechanisms (Saeidnia et al., 2024). de Filippis and Al Foysal (2024) advocated for multi-level bias mitigation strategies, including dataset diversification, fairness constraints, and transparent reporting of AI decision processes. Regulatory guidelines, notably the European Union’s AI Act, were highlighted as vital for standardizing ethical practices and ensuring responsible AI deployment (Dhopte & Bagde, 2023). Ethical vigilance was emphasized as essential not only to prevent discriminatory outcomes but also to foster sustained user trust and acceptance of AI technologies. Additionally, researchers underscored the importance of ethical transparency, advocating user-friendly mechanisms that clearly communicate AI capabilities, decision-making processes, and limitations (Dhopte & Bagde, 2023). Overall, the literature reinforces that robust ethical frameworks and transparency are indispensable for maintaining user trust, ethical integrity, and responsible AI development and deployment (Li et al., 2021).
5. Discussion
The findings from this narrative review highlight the critical role psychological theories play in enhancing AI system design, aligning closely with the principles of Human-Centred Artificial Intelligence (HCAI) (Saeidnia et al., 2024). By systematically analyzing recent literature, this review demonstrated that incorporating psychological constructs significantly improves AI usability, trustworthiness, and ethical integrity. Each theme identified provides specific psychological insights that can guide future AI developments and inform practical design considerations (Di Plinio, 2025).
Firstly, Theory of Mind (ToM) was shown to be highly effective in enabling AI systems to interpret user intentions and predict behaviours accurately, thereby improving user interaction quality and satisfaction. ToM integration into AI systems facilitates more natural and intuitive interactions, reflecting the necessity for AI to emulate human social cognition accurately (Kosinski, 2024).
The theme of Trust, Empathy, and Emotional Intelligence further illustrated the nuanced relationship between emotional responsiveness in AI and user trust. While emotional intelligence in AI interactions significantly enhanced perceived reliability and user engagement, it also raised ethical concerns about authenticity and emotional manipulation (Velagaleti, 2024). These findings suggest that emotional intelligence must be carefully balanced, calibrated accurately, and transparently communicated to maintain genuine user trust (Chavan et al., 2025).
Cognitive Load and Adaptation emerged as essential components for ensuring cognitive alignment between AI functionalities and human cognitive capacities. Systems designed according to cognitive load theory effectively enhanced user efficiency and reduced mental fatigue, emphasizing the practical importance of psychological integration into AI interaction design. Personalized cognitive load adaptations were particularly impactful, suggesting AI systems should dynamically adjust their interactions to user-specific cognitive profiles (Gkintoni et al., 2025).
The importance of Human-Centred and User-Centred AI Design was consistently reinforced across the reviewed studies, highlighting the necessity for active user involvement throughout the AI design process. Iterative, participatory methodologies significantly enhanced user satisfaction, usability, and acceptance (Xu, 2021). This emphasizes that AI development should incorporate continuous user feedback, adapting functionalities to align closely with actual user needs, preferences, and diverse contextual realities (Fucs et al., 2020).
AI’s application within mental health contexts was notably promising yet ethically complex. The review demonstrated AI’s substantial potential for delivering accessible, scalable psychological support through emotionally intelligent interfaces (Ni & Jia, 2025). However, ethical transparency, emotional authenticity, and clearly defined boundaries emerged as crucial for responsible and effective mental health interventions. This underlines the critical need for ethically sound AI development in sensitive application areas (Zhang & Wang, 2024).
Finally, Ethical Considerations emerged prominently as a fundamental pillar underpinning psychologically informed AI design. The review emphasized that comprehensive ethical frameworks, transparent decision-making processes, and bias mitigation strategies are essential for maintaining user trust and ethical integrity in AI technologies. Regulatory guidelines, such as the European Union’s AI Act, offer essential governance frameworks for responsible AI deployment (Oesterling et al., 2024).
Collectively, the review highlights the significant advantages of integrating psychological theories into AI systems, providing practical guidance for designers (Schmager et al., 2025) and researchers. However, it also identified gaps requiring further research, particularly regarding real-world validation, longitudinal studies, and cross-cultural applicability. Future research should focus on empirical validation in diverse, real-world contexts, ensuring that psychological principles translate effectively into practical AI functionalities (Azzam & Charles, 2024).
6. Conclusion
This narrative review underscores the importance of integrating psychological theories into AI design to enhance AI systems’ usability, trustworthiness, and ethical integrity. By critically synthesizing current literature, it becomes evident that psychologically informed AI significantly enhances user interactions, facilitating more intuitive, efficient, and trustworthy engagements (Bach et al., 2022). Each identified theme—Theory of Mind, trust and emotional intelligence, cognitive load management, human-centred design principles, mental health applications, and ethical considerations—provides valuable insights and practical implications for future AI developments (Xu & Gao, 2023).
The review confirms that aligning AI systems closely with human psychological processes, cognitive capacities, emotional needs, and ethical standards is not merely beneficial but essential for effective technology adoption and sustained user engagement. By embedding Theory of Mind, emotional intelligence, cognitive load principles, and user-centred design approaches, AI developers can create systems that intuitively resonate with users, reducing barriers to acceptance and promoting meaningful interactions (Virvou, 2023).
Moreover, the ethical dimension of psychological AI integration cannot be overstated. Maintaining ethical transparency, fairness, accountability, and clearly communicated boundaries ensures AI technologies are responsible, trustworthy, and accepted by users. Regulatory frameworks and ethical standards will continue to be vital for guiding responsible AI deployment and safeguarding against potential misuse or unintended consequences (Nastoska et al., 2025).
Despite significant advances, the systematic review highlights areas requiring additional attention and research. Specifically, future research must prioritize longitudinal studies and real-world validations to ensure psychological theories are effectively implemented and beneficial across diverse contexts. Additionally, cross-cultural research should explore the applicability and adaptation of psychological theories in varied global settings, ensuring AI systems meet diverse cultural, social, and psychological needs (Peters & Carman, 2024).
In conclusion, integrating psychological theories into AI design represents a critical pathway toward creating more effective, trustworthy, and ethically sound AI technologies. Continued research and practical implementation of these principles are essential for the future of AI, facilitating deeper psychological alignment, user trust, and broad societal acceptance of advanced technologies (Poole, 2025).