fCrit : A Visual Explanation System for Furniture Design Creative Support Conference: Explainable AI for the Arts Workshop 2025; June 23, 2025; Online

Vuong Nguyen and Gabriel Vigliensoni · 2025

Read fCrit : A Visual Explanation System for Furniture Design Creative Support Conference: Explainable AI for the Arts Workshop 2025; June 23, 2025; Online by Vuong Nguyen and Gabriel Vigliensoni (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-12416.txt.

Introduction

Vuong Nguyen and Gabriel Vigliensoni introduce fCrit, an AI system that critiques furniture designs through dialogue, in a 2025 paper. The work was presented at a workshop on explainable AI for the arts. Creative tools that explain themselves must do more than expose their logic: they should speak the designer's own language. The system, fCrit, is built on that idea.

Several cooperating agents draw on a structured knowledge base of formal design analysis, in the spirit of reflective learning. The system tailors its criticism to how each user frames and talks about a piece, grounding comments in the design's visual features. The authors present fCrit as a step toward human-centered explainable AI for creative practice: dialogic, situated, and shaped around the user's thinking rather than the model's.

Catalog record

CREATORVuong Nguyen and Gabriel Vigliensoni
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE18 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-12416

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

fCrit : A Visual Explanation System for Furniture Design Creative Support Conference: Explainable AI for the Arts Workshop 2025; June 23, 2025; Online Vuong Nguyen and Gabriel Vigliensoni Abstract . We introduce fCrit , a dialogue-based AI system designed to critique furniture design with a focus on explainability. Grounded in reflective learning and formal analysis, fCrit employs a multi-agent architecture informed by a structured design knowledge base. We argue that explainability in the arts should not only make AI reasoning transparent but also adapt to the ways users think and talk about their designs. We demonstrate how fCrit scaffolds this process by adapting explanations to users’ design language and awareness levels through a functional prototype. This work contributes to Human-Centred Explainable AI ( HCXAI ) in creative practice, advancing domain-specific methods for situated, dialogic, and visually grounded AI support. Additional Key Words and Phrases: human-centred explainable AI, mixed-initiative interaction, chatbot, domain-specific LLM, knowledge base, design critique, reflective learning, furniture design 1. Introduction Explainability is intrinsic to design critique ( Uluoǧlu 2000 ) . Furniture designers commonly employ Formal Critique —an exercise focussing on formal elements to develop formal awareness ( Hannah 2002 ) . This involves articulating tacit formal elements and analysing their relationships with design intent, often identifying both successful formal relationships and contradictions as opportunities for refinement. Based on this understanding, effective explainability must not only make formal concepts and patterns transparent, but also scaffold the designer’s ability to identify and articulate these elements themselves. As recent work in Human-Centred Explainable AI ( HCXAI ) suggests ( Bryan-Kinns 2024 ; Ridley 2025 ) , explanations should align with the user’s awareness, language, and intents. In this paper, we present a functional prototype of fCrit , an AI critiquing system for furniture design that supports visual thinking through reflective and adaptive feedback. Our system is based on the premise that explainability in the arts requires interpretive alignment, dialogic responsiveness, and domain knowledge structured from established formal critique methodology ( Hannah 2002 ; Wong 1993 ) . Our contributions are twofold: (1) a theoretical framework combining reflective learning, formal critique, and a mixed-initiative creative interface that adapts mixed-initiative interaction principles for design practice; and (2) a system architecture enabling adaptive explainability grounded in visual design concepts and patterns. 2. Related Work Evaluating design solutions and enhancing designer’s task performance are key functions of critiquing systems ( Robbins and Redmiles 2000 ; Silverman 1992 ) . Their architecture comprises an analyser module that matches user input against reference knowledge base, and a dialogue module for error detection ( Silverman 1992 ) . While these systems have found widespread application in design—from design sketching ( Oh et al. 2004 ) to architectural layout planning ( Woodbury and Drogemuller 2000 ) , essential aspects of creativity support—feedback modality ( Oh and Oh 2017 ) , cognitive ergonomics ( Tianfield and Wang 2004 ) , and contextual awareness ( Fischer 2012 ) —remain neglected in their development. Tutoring systems emerged from the need for personal support system for learners, with capabilities enhanced by large language models (LLMs) ( Ait Baha et al. 2024 ) . Nevertheless, design critique demands more sophisticated knowledge structures than conventional error-detection approach, especially when translating designers’ informal expressions and visual language. This limitation highlights the need for design-oriented learning systems that accommodate designers’ natural thought processes, communication patterns, and domain-specific vocabulary. Adept design critique systems must entail competencies on adaptive design thinking ( Sharp and Macklin 2019 ) , deep reflection ( Redström 2017 ) , and problem framing ( Cross 2006 ) . In the following sections, we demonstrate how fCrit addresses these shortcomings through a structured knowledge base and a responsive dialogic architecture, using Gebrüder Thonet’s M - 209 armchair (Fig. 1 ) as our illustrative case study. Figure 1. Gebrüder Thonet’s M-209 bentwood armchair (circa 1900). The dialogue demonstration in Section 6 shows how fCrit interprets informal descriptions of its curvilinear elements (“noodle-y,” “playful”) and scaffolds formal design analysis. https://www.thonet.de/en/all-products/detail/209 . An image of Gebrüder Thonet's M-209 armchair, circa 1900, highlighting its elegantly curved wooden frame and flowing design. The armchair serves as a case study for our dialogue demonstration. 3. Theoretical Framework Our framework encompasses recent calls for XAI approaches that support sensemaking in creative domains ( Bryan-Kinns 2024 ; Ridley 2025 ) . A system must adapt its language to match the user’s awareness and choice of words, and encourage deeper reflection. We integrate three frameworks that support explainability in furniture formal critique: Reflective Learning Following Schön 1992 and Moon 2013 , critique should enhance reflection-in-action and support the articulation of tacit knowledge. Rather than prescribing solutions, fCrit scaffolds design reflection by mirroring insights, prompting elaboration, and constructing visual analogies. Mixed-Initiative Creative Interface: Inspired by ( Deterding et al. 2017 ) , rather than a unidirectional exchange, we frame and model the interaction as a co-constructive dialogue where both human and AI can take initiative towards insight discovery—the system dynamically adjusts its level of intervention and critiquing tone based on the designer’s need and evolving observations. Formal Critique Terminology: fCrit uses a domain-specific vocabulary derived from prominent formal critique literature ( Hannah 2002 ; Wong 1993 ) . The knowledge base encodes visual concepts (e.g., line, shape, form) and patterns (e.g., balance, contrast, unity), with multiple levels of abstraction and formal awareness. Lastly, formal critique is unconcerned with regional-stylistic or period-based labelling (e.g., Scandinavian or Victorian). This approach enables the system to interpret colloquial language and make formal connections without constraining creative freedom through stylistic designation. 4. Critique Knowledge Base Inspired by the key principle of tailored explanation ( Bryan-Kinns 2024 ; Ridley 2025 ) and structured after formal critique methodology ( Hannah 2002 ; Wong 1993 ) , a knowledge base for formal critique is built to capture visual concepts and patterns as a structured embedding containing multiple ontological perspectives. Each embedding includes formal definitions, perceptual effects, applications, examples, and awareness-specific terminology. This structure supports adaptive explanations while ensuring design principle consistency. For instance, Figure 2 shows a hierarchical representation of properties for Curvilinear Line , which could be used to describe some aspects of the chair in Figure 1 . This case example is demonstrated in Section 6 to illustrate how the system connects user descriptions like “noodle-y” and “playful” to formal concepts such as curvilinearity and rhythmic repetition —exemplifying how the knowledge base promotes tacit insight interpretation and visual literacy. A visual representation of the concept `Curvilinear Line' as organised within fCrit's knowledge base. The hierarchy includes formal definitions, perceptual effects, tailored language for different awareness levels, domain-specific furniture design applications, and examples such as rounded armchair backs and curved table edges. Figure 2. Hierarchical representation of Curvilinear Line in fCrit ’s knowledge base, showing how visual concepts are structured with formal definitions, perceptual effects, awareness-tailored language, applications, and examples to support adaptive design critique. A visual representation of the concept `Curvilinear Line' as organised within fCrit's knowledge base. The hierarchy includes formal definitions, perceptual effects, tailored language for different awareness levels, domain-specific furniture design applications, and examples such as rounded armchair backs and curved table edges. 5. System Architecture fCrit is implemented on n8n, 1 1 1 https://github.com/n8n-io/n8n a workflow

Read the complete text