As artificial intelligence (AI) becomes deeply embedded in daily life, the conversation around human-centered AI (HCAI) has gained significant attention. AI is no longer just about technological performance – it must also align with human values, usability, fairness, and ethical responsibility. However, despite the growing emphasis on HCAI, there is no clear consensus on what constitutes human-centeredness, leading to fragmented approaches in AI development.
A recent study, “What is Human-Centeredness in Human-Centered AI? Development of Human-Centeredness Framework and AI Practitioners’ Perspectives”, authored by Aung Pyae from Chulalongkorn University, Thailand, addresses this gap. Submitted in arXiv (2025), the study develops a hierarchical framework of 26 attributes that define human-centeredness in AI, validated through practitioner input and empirical analysis. This framework prioritizes ethical foundations, usability, emotional intelligence, and personalization, offering actionable guidance for designing AI systems that truly serve human needs while upholding societal values.
Understanding Human-Centered AI and Its Evolution
Human-Centered AI (HCAI) is an approach that prioritizes human well-being, aligns AI systems with ethical standards, and enhances user experience. It draws from interdisciplinary fields such as Human-Computer Interaction (HCI), Human-Centered Design (HCD), and User Experience Design (UxD) to ensure that AI systems are not just technologically efficient but also socially responsible.
The concept of HCAI has its roots in early HCI theories, where technology was designed to augment human capabilities rather than replace them. Pioneers like Engelbart (1962) and Licklider (1960) emphasized human-technology collaboration, which later evolved into user-centered design frameworks. However, with AI’s growing societal impact, the need for clear ethical guidelines and inclusive design principles became more pressing. Organizations such as Stanford’s HAI, the AI Now Institute, and the Partnership on AI have since contributed to the research on responsible AI practices.
Despite these advancements, there has been no unified framework that systematically defines what it means for AI to be “human-centered.” Existing definitions focus on specific aspects like transparency, user control, or fairness, but fail to integrate them into a cohesive model. This study bridges that gap by synthesizing insights from AI practitioners, academic literature, and industry best practices to develop a practical and empirically validated framework for HCAI.
Building the human-centeredness framework
To develop a robust definition of human-centeredness, the study followed a multi-step research process. It began with a systematic review of 81 definitions of HCAI from academic papers, international organizations, and enterprise reports. Through thematic analysis, 78 attributes of human-centeredness were identified. These were then refined through frequency analysis, expert validation, and a practitioner survey to determine the 26 most critical attributes.
These 26 attributes were categorized into four hierarchical tiers:
-
Primary Attributes: Ethical Foundations and Core ValuesThe highest priority attributes in the framework focus on ethics, trust, and human values. AI practitioners rated fairness, transparency, user trust, and ethical data usage as essential. These attributes ensure AI aligns with societal values, respects privacy, and provides unbiased decision-making.
-
Secondary Attributes: Usability and Human AutonomyThe second tier emphasizes user-friendliness, control, and decision-making support. AI systems should be intuitive, easy to interact with, and respect user autonomy by allowing individuals to make informed choices without manipulation or coercion.
-
Tertiary Attributes: Emotional Intelligence and User-Centered InteractionsAI should be empathetic and responsive to human emotions, enhancing its ability to support users in sensitive contexts. Practitioners highlighted human well-being, empathy, and personalized user experiences as important but challenging to implement in AI systems.
-
Quaternary Attributes: Behavioral and Adaptive IntelligenceThe final tier includes attributes related to human behavior adaptation, cognitive support, and stakeholder engagement. While these factors contribute to personalized AI experiences, they were rated lower in priority compared to ethical and usability concerns.
Findings: The priorities of AI practitioners
The study surveyed 120 AI practitioners from diverse backgrounds, including software engineers, UX designers, data scientists, and machine learning researchers. Participants provided insights into how they define and implement human-centered AI in real-world applications.
The results showed that ethical considerations were ranked as the most critical aspect of HCAI, with human values, fairness, and transparency receiving the highest scores. Practitioners agreed that AI must be designed to respect user privacy, avoid harm, and promote trust. Surprisingly, attributes related to emotional intelligence and user adaptation were considered important but secondary, indicating that practical usability concerns take precedence over AI’s ability to understand human emotions.
Additionally, while human control and decision-making autonomy were emphasized, there was a lack of consensus on the extent to which AI should allow user intervention. Some practitioners argued for full user oversight, while others supported semi-autonomous systems that require minimal human supervision. This reflects ongoing debates in AI ethics regarding automation versus human control.
Future implications and the path forward
This research provides a structured and validated framework that can help AI designers and policymakers prioritize human-centered attributes in AI development. However, there are several challenges and areas for future research:
First, cross-industry variations in human-centered AI need further study. HCAI principles may vary significantly across domains such as healthcare, finance, education, and autonomous systems. Tailoring the framework for specific industries will be essential for practical implementation.
Second, the evolution of human-centeredness in AI must be tracked over time. As AI becomes more advanced and deeply integrated into society, the definition of human-centered AI may shift. Future research should explore how AI practitioners’ priorities change as technologies and societal expectations evolve.
Lastly, regulatory frameworks must incorporate these findings to ensure AI systems are built with human well-being in mind. Governments and industry leaders should use this framework to create policies that mandate ethical AI design, promoting transparency, fairness, and accountability.
