CogInstrument: Modeling Cognitive Processes for Bidirectional Human-LLM Alignment in Planning Tasks

Anqi Wang, Dongyijie Primo Pan, Xin Tong, Pan Hui | ArXiv

• Introducing cognitive motifs as fundamental units for human–LLM alignment, formalizing user reasoning as causal, compositional, and revisable structures rather than flat intents, and reframing alignment as operating over user-grounded cognitive processes.
• Implementing CogInstrument, a cognition-centric interaction system that extracts motifs from user interactions, externalizes them as editable causal graphs, and maintains a shared reasoning space for iterative, bidirectional negotiation between users and LLMs.
• A within-subject study (𝑁=12) shows that motif-based externalization improves articulation and diagnosis of complex reasoning, reduces repair effort, and enhances user trust and agency, providing a new analytical lens on human–AI alignment.

NexusAI: Enabling Design Space Exploration of Ideas through Cognitive Abstraction and Functional Decomposition

Anqi Wang, Bingqian Wang, Huiyang Chen, Keqing Jiao, Lei Han, Xin Tong, Pan Hui | ArXiv

• Proposing Cognitive Abstraction (CA) as an intermediate computational layer that transforms monolithic LLM outputs into manipulable, multi-dimensional design spaces, addressing the problem of compositional opacity in existing ideation tools.
• Developing NexusAI, an LLM-powered interface that operationalizes CA through What–How–Value decomposition, multi-level abstraction operators (L1–L4), and typed cross-dimensional recombination, enabling fine-grained manipulation of idea components beyond “whole ideas.”
• A within-subject study (𝑁=14) demonstrates that externalizing these primitives reduces cognitive load, mitigates design fixation, and supports multi-directional exploration, while preserving the cognitive benefits of expert-level synthesis in AI-assisted workflows.

Reflexa: Exploring the Large Language Models as Artifacts to Elicit Reflection in Creative Coding

Anqi Wang, Zhengyi Li, Lan Luo, Xin Tong, Pan Hui | ArXiv

• Proposing LLM-elicited creative reflection strategies as scaffoloding for creative coding tasks, including reflective suggestions, lightweight inspiration, and non-linear version visualization.
• Developing Reflexa, an LLM-powered interface that operationalizes creative reflection strategies.
• A within-subject study (𝑁=20) demonstrates that the effectiveness of this reflection scaffoloding in supporting reflection, creativity, and creative outcomes.

DesignerlyLoop: Forming Design Intent through Curated Reasoning for Human-LLM Alignment

Anqi Wang, Zhengyi Li, Xin Tong, Pan Hui

ArXiv | Proceedings of the 2026 ACM Designing Interactive Systems Conference (DIS’26), Singapore, June 2026 (Date of conditional acceptance: 18 Mar 2026)

• Reframing human–LLM misalignment in design thinking: Identifies a core gap between iterative intent formulation in design and LLMs’ assumption of complete inputs, limiting exploratory reasoning.
• Curated reasoning as an interaction paradigm: Introduces curated reasoning, enabling designers to inspect, reorganize, and selectively regenerate LLM reasoning via a two-layer structure separating intent and model reasoning.
• Empirical evidence for creativity and design quality gains: Studies with designers show that curated reasoning enhances creative exploration and output quality, repositioning LLMs as structured reasoning partners rather than generators.

Dream the Dream: Futuring Communication between Lgbtq+ and Cisgender Groups in Metaverse Space

Anqi Wang, Lei Han, Jiahua Dong, Muzhi Zhou, David Yip, Yuyang Wang, Pan Hui

ArXiv | Proceedings of the 2026 ACM Designing Interactive Systems Conference (DIS’26), Singapore, June 2026 (Date of conditional acceptance: 18 Mar 2026)

• Revealing barriers in cross-group digital communication: Identifies how heteronormative norms and platform biases shape interactions between LGBTQ+ and cisgender users in virtual environments.
• Futuring participatory design for inclusive Metaverse interactions: Engages participants in speculative workshops to co-create socio-spatial-technical solutions across activity, interaction, scene, and space layers.
• Designing for power-aware and equitable virtual spaces: Proposes principles addressing spatial cues and power geometry, positioning futuring as a method for reimagining inclusive communication infrastructures.

Eye2Recall: Exploring the Design of Enhancing Reminiscence Activities via Eye Tracking-Based LLM-Powered Interaction Experience for Older Adults

Han Lei, Mingnan Wei, Qiongyan Chen, Anqi Wang, Rong Pang, David Yip

PDF | ArXiv | Proceedings of the 2026 ACM Conference on Intelligent User Interfaces (IUI 2026), March, 2026, Paphos, Cyprus (Date of acceptance: 12 December 2025)

• Designing for reflective reminiscence: Derived design considerations from expert interviews, focusing on accessibility, cultural sensitivity, and emotionally safe engagement for older adults.
• Gaze-to-prompt interaction mechanism: Developed Eye2Recall, a mixed-initiative conversational system that transforms eye gaze on photos into adaptive LLM prompts, reducing user effort while aligning with users’ attention.
• Empirical evidence of well-being and recall support: Pilot study with older adults shows improved mood, low interaction burden, and enhanced retrieval of concrete memories through gaze-driven reflective storytelling.

Echoing Streams: Constellating Calligraphic Ritual, Place and Data through Interactive Experience

Lei Han, Anqi Wang, Ziqi Lin, David Yip

International Symposium on Electronic / Emerging Art (ISEA 2026), Dubai, United Arab Emirates (Date of acceptance: 27 Feb 2026)

• Reimagining cultural ritual through interactive systems: Translates Qu Shui Liu Shang and calligraphy tracing into a multisensory, data-driven installation that bridges historical practice and contemporary interaction.
• Integrating embodied interaction with generative media: Combines remote-sensed terrain, real-time handwriting capture, and algorithmic audiovisual feedback to create a shared, participatory calligraphy environment.
• Exploring identity through collective and mediated practice: Preliminary study shows high fluency and aesthetic engagement, highlighting how technologically mediated participation supports cultural learning and identity construction.

Pinning “Reflection” on the Agenda: Investigating Reflection in Human–LLM Co-Creation for Creative Coding

Anqi Wang, Zhizhuo Yin, Yulu Hu, Yuanyuan Mao, Lei Han, Xin Tong, Keqin Jiao, Pan Hui

PDF | ArXiv | Proceedings of the ACM Conference on Computer-Supported Cooperative Work & Social Computing (CSCW ‘25), Bergen, Norway, October 2025 (Date of acceptance: 19 June 2025)

• We demonstrate how Baumer’s reflection typology applies and adapts in the context of LLM-based creative coding;
• We conduct an empirical evidence showing how task-decomposed prompting fosters deeper, more strategic reflection;
• We propose design implications for rethinking reflective prompting beyond productivity scaffolding, but as a catalyst for creative reasoning and adaptive goal-setting.

Towards AI-Architecture Liberty: A Comprehensive Survey on Design and Generation of Virtual Architecture by Deep Learning

Anqi Wang, Jiahua Dong, Lik-Hang Lee, Jiachuan Shen, Pan Hui

PDF | ArXiv | ACM Computing Surveys (CSUR) (Date of acceptance: 9 July 2024)

PoEmotion: Generative Calligraphy Visualizing Emotion in Poems

Tiancheng Liu, Anqi Wang, Xinda Chen, Jing Yan, Yan Li, Pan Hui, Kang Zhang

ArXiv | International Symposium on Electronic Art 2024 (ISEA 2024), Brisbane, Australia, June 2024

Decentralized, not Dehumanized in the Metaverse: Bringing Utility to NFTs through Multimodal Interaction

Anqi Wang, Ze Gao, Lik-Hang Lee, Tristan Braud, Pan Hui

PDF | ArXiv | ICMI 2022