human-ai interface design

Human-AI conversational interfaces face design challenges stemming from ambiguous user goals, users’ limited awareness of AI capabilities, and the transient, ephemeral nature of interactions. Method: This study proposes an AI-agent-driven CHAI (Conversational Human-AI Interaction) workflow, positioning AI as a collaborative design partner to support real-time prompt generation, user-agent simulation testing, and goal-clarification scaffolding. Developed through four iterative cycles and empirically validated with 10 users, the workflow integrates design probes, thematic analysis, and contextualized prompt engineering. Contribution/Results: We deliver an interpretable prototype suite and introduce the first reusable CHAI agent workflow paradigm under a “research-through-design” framework. Evaluation shows significant improvements in user goal articulation (+42%) and feedback depth. Additionally, we establish a curated, annotated design asset library comprising 12 canonical interaction patterns, enabling high-fidelity early-stage CHAI experience validation.

As AI agents become deeply integrated into core enterprise workflows, designing effective human-AI interaction to enhance user experience, foster adoption, and support user-centered decision-making has emerged as a critical challenge. This study addresses this issue through a mixed-methods approach, combining qualitative interviews and quantitative experiments to systematically investigate interaction patterns between humans and AI agents in business contexts and identify key design elements that shape user experience. Grounded in empirical findings, the work proposes a set of human-AI interaction design guidelines tailored for commercial environments, along with a quantifiable evaluation framework. These contributions offer both theoretical grounding and practical guidance for development teams seeking to optimize and deploy large-scale human-AI collaborative systems.

Current AI tools exhibit fragmentation in professional domains, supporting only isolated task augmentation without enabling sustained, adaptive human-AI collaboration. To address this, we propose a unified human-AI collaboration architecture centered on *process as the first-class citizen*, jointly modeling interaction, process logic, and infrastructure. This enables explicit, verifiable, and dynamically adaptive process representation—the first such realization. Our novel paradigm—verifiable, intervenable, and evolvable human-AI processes—supports cross-task and longitudinal goal alignment and co-evolution. We overcome tool silos via Process-as-a-Service (PaaS), observable collaborative state design, and multi-granularity intent alignment. The resulting architecture provides a scalable foundation and empirically grounded design principles for collaborative AI platforms. (136 words)

This work addresses the challenge that current AI agents struggle to interpret users’ concurrent interaction intents on shared artifacts, thereby limiting dynamic co-creation. To overcome this, we propose CLEO—a collaborative intelligent agent grounded in mixed-initiative interaction principles—that dynamically switches among delegation, guidance, and collaboration modes by recognizing user concurrent behaviors in real time. We introduce the first collaborative model capable of real-time intent interpretation, identifying five behavioral patterns, six triggering mechanisms, and four enabling factors, and implement a decision framework comprising six interactive loops. Based on 214 rounds of interactions with professional designers, we quantitatively analyze mode usage (70.1% delegation, 28.5% guidance, 31.8% collaboration) and release design guidelines alongside a labeled dataset to support future research.

To address the lack of a systematic framework for agent design in mixed-initiative visual analytics, this study proposes the first comprehensive, lifecycle-spanning six-dimensional agent design space—encompassing perception, environment understanding, action capability, communication strategy, role dynamics, and human–agent collaborative reasoning. Grounded in a systematic literature review and cross-case coding analysis of 90 visual analytics systems and 207 agents, we develop an extensible, reusable classification framework. This framework supports both design decisions for new systems and systematic positioning and evaluation of existing ones. It explicitly identifies critical research gaps—including dynamic role switching and formal modeling of collaborative reasoning—thereby providing theoretical foundations and practical guidance for agent-driven visual analytics. (136 words)

As AI systems become increasingly agentic, user interactions shift from frequent dialogue toward greater reliance on explanations and oversight. This study addresses this transition by introducing source perception theory from communication studies into the design of agentic AI interfaces—a novel integration in this domain. It proposes three key types of explanations: action process, uncertainty, and coordination, and embeds them within an explainable AI (XAI) framework to enable user-customizable explanation delivery. By doing so, the work enhances users’ trust and sense of control while carefully balancing AI autonomy with human agency. The research thus offers both theoretical grounding and actionable design principles for achieving transparency and supervisability in highly autonomous AI systems.

This study addresses the prevalent conflation of human–AI interaction with genuine collaboration, noting that most current systems operate through consultation, instruction, or delegation rather than exhibiting core collaborative features such as symmetry, shared goals, and mutual regulation. Drawing on theories of collaborative learning, the work proposes a novel five-tier taxonomy of human–AI diagnostic collaboration—ranging from transactional to truly collaborative—and rigorously distinguishes pseudo-collaboration from authentic forms. It further identifies the critical functionalities and affordances necessary for achieving higher-order collaboration. Through process-sensitive empirical analysis of interaction data from educational writing and problem-solving tasks, the research demonstrates that mainstream AI systems predominantly remain at lower tiers, with only the highest level meeting established criteria for true collaboration. This framework offers a theoretical foundation, evaluative metric, and design guidance for responsible human–AI collaboration in education.

This study addresses coordination inefficiencies in human–AI collaboration within shared workspaces, where the absence of effective coordination mechanisms often incurs process losses and can even reduce team performance when new collaborators are introduced. To mitigate these issues, this work proposes a scaffolding mechanism that integrates shared group memory with human-in-the-loop (HITL) approval gating, employing structured coordination strategies to optimize responsibility allocation and expert knowledge scheduling. Evaluated across 1,482 experimental sessions using the Collaborative Gym environment and DiscoveryBench tasks, the approach significantly enhances joint decision-making performance in three-person teams, sharpens the clarity of responsibility signaling, and more effectively channels expert knowledge to guide collective action.

Existing multimodal human-AI interaction systems often treat alignment, explainability, and user agency in isolation, leading to poor user understanding of AI intent and diminished trust and sense of control. This work proposes a unified collaborative framework that, for the first time, co-designs multimodal alignment, real-time multimodal explainable feedback (encompassing visual, textual, and spoken modalities), and user intervention mechanisms within a continuous interaction paradigm. By establishing a closed-loop architecture for multimodal intent recognition and responsive feedback, the framework significantly enhances users’ comprehension of system behavior, perceived control, and overall transparency. Empirical validation in two high-stakes, time-sensitive scenarios—collaborative design and warehouse robotics—demonstrates the efficacy of this approach in fostering effective and trustworthy human-AI collaboration.

This work addresses the paradox that, despite the growing prevalence of graphical user interfaces, terminal-based interactions remain dominant in AI collaboration. Drawing on human-computer interaction (HCI) theory, the paper proposes that effective human–AI–interface collaboration hinges on three core attributes: representational compatibility between the agent and the interface, interactional transparency of agent behavior, and low barriers to human participation. Through theoretical analysis, design principle derivation, and cross-modal comparison, the study systematically articulates and formalizes the terminal’s inherent advantages as an interaction paradigm for AI agents, demonstrating how it naturally fulfills these attributes. The research not only clarifies the irreplaceable role of terminals in AI collaboration but also distills generalizable design principles applicable to graphical and spatial interfaces, offering a theoretical foundation and practical framework for future multimodal human–AI collaborative systems.