← CHI 2026 map

CHI '26 · Honorable mention · full-paper review · confidence medium-high

QuerySwitch: Supporting the Design Process by Balancing Vagueness through Large Language Models

Myungjin Kim , Bogoan Kim , Kyungsik Han

DOI PDF Program page Page JSON

QuerySwitch is a thoughtful CHI design contribution because it does not just add more control to LLM use; it reframes vagueness as something to be balanced across creative stages. The novelty is credible and the evaluation is appropriate for an early-stage prototype, though the evidence remains domain-specific and small-scale.

Axes Lens

Rare contribution shape, typical evidence profile. The point here is not a score. It is to show what kind of claim the paper makes, and whether the evidence pattern is unusual or baseline in this 268 -review set.

Review Summary

QuerySwitch stands out because it challenges a common assumption in AI-assisted design: that better support means making prompts and outputs more precise. Instead, the paper treats vagueness as a resource that designers intentionally move between, and it operationalizes that idea through two distinct modes—hierarchical keywords for divergence and combinational images for convergence. That is a meaningful interaction contribution because it connects a conceptual claim about creativity to a concrete mechanism that designers can actually use during moodboard and sketch work. The paper’s strongest point is the alignment between the problem framing, the prototype, and the evaluation: the authors do not claim a universal LLM solution, but rather show how a specific interaction pattern can scaffold designer agency and creative exploration in fashion design. The validation is solid for a CHI prototype paper: a within-subject study with 10 fashion professionals and a ChatGPT+DALL-E 3 baseline gives a reasonable first look at whether the mechanism helps. At the same time, the limitations are important and well stated. The sample is small, the session is short, and the domain is narrow. The authors also acknowledge that generated images may not reflect current social trends or cultural context, and that the system depends heavily on prompting. So the paper’s contribution is best read as a design principle and interaction technique for creativity-driven workflows, not as evidence that vagueness balancing is solved in general. Overall, this is a credible honorable-mention-level contribution: conceptually sharp, well scoped, and empirically supported at the level expected for an early-stage HCI system paper.

What Changed

Canon before

Prior CHI work on LLM-supported design typically emphasizes prompt refinement, controllability, or output quality; this paper instead centers vagueness as a design resource to be actively balanced across stages of creative work.

Departure from common sense

The paper argues against the intuitive impulse to make AI prompts and outputs as concrete as possible. Instead, it treats vagueness as something designers should manage deliberately by switching between modes, so the system supports creative exploration rather than simply maximizing specificity.

Actual novelty

QuerySwitch introduces a two-mode interaction for balancing vagueness in fashion design: Abstract Query-Hierarchical Keywords for divergence and Parallel Query-Combination Images for convergence, applied across moodboard and sketch stages to scaffold iterative design exploration.

Evidence

The paper presents an interactive prototype and evaluates it in a 90-minute within-subjects study with 10 fashion professionals. The study compares QuerySwitch against a ChatGPT plus DALL-E 3 baseline and reports that the system supports vagueness management, usability, and creative exploration, while also documenting concrete limitations around trends, sample size, duration, and reliance on prompting.

Limits

Method limits

Evidence comes from a small within-subject study with 10 fashion professionals over 90 minutes, so the findings are indicative rather than broadly generalizable. The evaluation also compares against a specific baseline, which constrains what can be inferred about relative performance across other LLM tools or design domains.

Deployment limits

The authors note that combinational images may not sufficiently reflect social trends, and the system may miss cultural context. Because the interaction relies primarily on prompting, deployment depends on the quality and behavior of underlying generative models and may require adaptation for other creative workflows.

Boundary conditions

The approach is grounded in fashion design practice and in an iterative divergent-convergent process. Its benefits are most plausible when designers need to preserve ambiguity while still steering generation; it may be less effective where trend fidelity, cultural specificity, or long-horizon workflow integration are central.

Manual check: false

Position in field

This is a CHI design-system contribution that reframes vagueness as a productive resource in human-AI creativity. It sits between interaction technique and design-method work, offering a concrete prototype plus design principles for creativity-driven domains rather than a general-purpose LLM platform.

Abstract

Designers often regard vagueness as an essential aspect of creative work, as it fosters diverse interpretations and helps prevent fixation. Although large language models (LLMs) are increasingly viewed as a promising creative partner, designers struggle to productively incorporate vagueness into AI-supported workflows. To address this challenge, we present QuerySwitch, an interactive prototype that enables fashion designers to manage vagueness by flexibly switching between two distinct query-output modes. Findings from a user study show that QuerySwitch helps fashion designers balance vagueness, enhances the usability of LLMs in design tasks, and promotes creative exploration. This work contributes to HCI by (1) foregrounding a critical construct in human–AI collaboration, (2) demonstrating how interaction mechanisms can scaffold designer agency in LLMs use, and (3) articulating design principles—structuring exploration and preserving key query formulations—that extend to creativity-driven domains.