AIDED: Augmenting Interior Design with Human Experience Data for Designer–AI Co-Design

Paper Title

AIDED: Augmenting Interior Design with Human Experience Data for Designer–AI Co-Design

Publication Info

• Topic area: Human–AI collaboration in interior design using multimodal client data.
• Keywords: Interior design, generative AI, human–data interaction, multimodal data, gaze heatmaps, questionnaires, AI-predicted overlays, co-design, interpretability, authenticity.

Background and Problem

• Problem / challenge: Conventional interior design tools fail to capture nuanced client experiences, relying primarily on demographic and stylistic inputs while neglecting affective and perceptual data. Generative AI systems produce rapid outputs but lack integration with client-specific experiential signals.
• Significance: Incorporating client experience data into design workflows can improve alignment between design outcomes and client needs, enhancing satisfaction and usability.
• Motivation and related work: Prior research has explored evidence-based spatial design, perceptual modeling, and generative AI tools, but these approaches often overlook direct client inputs like gaze patterns or emotional responses. This study builds on these efforts by treating client experience as a primary design material.

Solution

• Proposed approach: AIDED (Augmenting Interior Design with Human Experience Data), a workflow integrating multimodal client data—demographics, gaze heatmaps, questionnaire visualizations, and AI-predicted overlays—into generative AI-assisted design processes.
• Novelty:
  1. Development of a system that incorporates experiential client signals into generative AI workflows.
  2. Empirical evaluation of how different client data modalities affect design outcomes, decision-making, and collaboration.
  3. Identification of an authenticity–interpretability trade-off in multimodal data integration.
• Procedure and key techniques:
  • Four experimental conditions (baseline demographics, gaze heatmaps, questionnaire visualizations, AI-predicted overlays) tested with professional designers.
  • Iterative design refinement using generative AI tools, informed by multimodal client data.
  • Post-task evaluations, novice surveys, and interviews to assess satisfaction, appropriateness, and usability.

Results

• Concrete findings:
  • Questionnaire visualizations were rated as the most reliable and actionable modality for decision-making (M = 6.00, SD = 0.95).
  • AI-predicted overlays improved communication with generative AI but required natural language explanations for trust and usability.
  • Gaze heatmaps were visually clear but lacked interpretability, increasing cognitive load.
• Advantage over baselines:
  • Designs created with richer client data (Conditions 3 and 4) were judged more appropriate by novices (59.3% and 54.9% preference rates, respectively) compared to baseline demographics (36.8% preference).
  • Structured questionnaire data enabled more precise and client-aligned modifications than qualitative gaze data or minimal demographics.
• Experiments / evaluation:
  • Mixed-method study with 12 professional designers and 30 novice evaluators.
  • Within-subjects design across four conditions, assessing cognitive load, trust, creativity, and satisfaction.
  • Statistical analyses (Friedman tests, Wilcoxon signed-rank tests) revealed significant differences in decision-making efficacy and design appropriateness across modalities.
• Limitations and future work:
  • Limited sample size (12 designers) and constrained task contexts (residential interiors).
  • Need for improved interpretability of AI-predicted overlays and broader client demographic representation.
  • Future research should explore longitudinal data collection, real-world deployments, and expanded design typologies.

Summary

AIDED integrates multimodal client data into generative AI workflows, addressing gaps in conventional interior design tools by balancing authenticity and interpretability. Empirical findings show that structured questionnaire data and AI-mediated overlays enhance decision-making and client alignment, while gaze heatmaps require interpretive support. The system preserves professional autonomy, enabling designers to use client data as scaffolding rather than prescriptive input. Future work should refine AI-predicted overlays, expand participant diversity, and situate research within authentic project phases to further advance human–AI co-design methodologies.