Call for Papers · HUMANE-LLM

HUMANE-LLM provides a focused forum on human-centered, responsible, and sustainable LLM-based intelligent user interfaces. Building on work in human-AI interaction, mixed-initiative systems, user agency, explainability, responsible AI, and sustainable NLP, we invite technical and socio-technical contributions on LLM-based IUIs.

We especially welcome work connecting LLM opportunities and risks to user-facing systems such as conversational agents, recommender interfaces, decision-support tools, education, healthcare, and adaptive applications. Topics are organized into five clusters spanning two directions: system-level design, user modelling, personalization, and interaction; and risk, trust, responsibility, and evaluation.

Topic clusters

System-level design, user modelling, and interaction. This direction concerns how LLMs can be integrated into intelligent, adaptive, and user-facing systems. Relevant topics include LLM-based user modelling, preference modelling, personalization, adaptive interfaces, conversational and multimodal interaction, recommender interfaces, decision-support tools, educational and healthcare applications, and mixed-initiative systems that support user control, correction, steering, and delegation.
LLM-based agents, workflows, and intelligent interface architectures. This direction covers the design of LLM-powered assistants, tool-using agents, multi-agent systems, retrieval-augmented interfaces, long-term memory, context management, and interface architectures that combine LLMs with user profiles, external tools, knowledge bases, sensors, or domain-specific resources. We especially welcome work that studies how such systems can remain understandable, controllable, and useful in real user-facing settings.
Risk, responsibility, safety, and alignment. This direction addresses the risks that arise when LLMs influence user choices, recommendations, decisions, or actions. Relevant topics include hallucination, misinformation, unsafe tool use, prompt injection, memory risks, over-reliance, manipulation, loss of agency, inappropriate delegation, and alignment between LLM behavior, user values, interface goals, and domain constraints.
Fairness, transparency, privacy, and contestability. This direction focuses on responsible and accountable LLM-based IUIs, including fair and privacy-preserving personalization, bias and stereotyping in adaptive interfaces, transparent communication of uncertainty and limitations, explainable recommendations and decisions, auditability, and interaction mechanisms that allow users to inspect, question, correct, or contest LLM outputs and system behavior.
Human-centered evaluation, governance, and sustainable deployment. This direction concerns evaluation and governance of LLM-based IUIs in practice. Topics include user studies, field deployments, longitudinal evaluation, benchmarks, and metrics for agency, trust calibration, cognitive load, fairness, privacy, task success, and well-being, as well as resource-aware deployment strategies such as model selection, caching, prompt optimization, smaller or local models, and trade-offs among performance, usability, cost, latency, privacy, and environmental impact.

Although the core focus of the workshop is on LLM-centric intelligent user interfaces, we also solicit strong contributions that do not directly use LLMs, as long as they provide clear insights for the design, evaluation, governance, or theory of human-centered intelligent interfaces in the age of LLMs. This includes work on user modelling, adaptive systems, mixed-initiative interaction, personalization, explainability, trust calibration, fairness, privacy, and responsible interface design that can inform future LLM-based IUIs.