How AI Understands Client Experience in Design

Explore how AI integrates human experience data into design workflows to improve decision-making and create more client-aligned spaces.

The Core Problem: Translating Client Experience into Design

The AIDED framework showing how different types of client data impact design decisions and the trade-off between authenticity and interpretability.

In interior design and broadly across AEC there’s a persistent gap:

• Clients experience spaces emotionally

• Designers work with structured inputs

That mismatch creates friction.

The paper AIDED focuses on this exact issue:

How can client experience be captured and used effectively in AI-assisted design workflows?

---

What the System Actually Does

Four types of client data used in the study, ranging from basic demographics to AI-predicted attention overlays.

The researchers built a system called AIDED, designed to integrate client experience data into generative AI workflows.

Instead of relying only on traditional briefs, the system introduces four types of client information:

---

1. Demographic Briefs (Baseline)

• Age, occupation, budget, lifestyle

• Represents typical early-stage inputs

👉 This serves as the control condition

---

2. Gaze Heatmaps

• Tracks where users look in interior scenes

• Visualized as attention maps over space

👉 Captures implicit attention, not verbalized preferences

Finding:

• High authenticity

• But increased cognitive load and ambiguity

---

3. Questionnaire Visualizations

• Structured ratings across 15 Architectural Experience (AE) dimensions

  • e.g., comfort, naturalness, complexity

• Presented as charts

👉 Converts subjective experience into structured data

Finding:

• Most trusted and actionable format

• Improved satisfaction and creativity

---

4. AI-Predicted Attention Overlays

• AI predicts which spatial regions matter based on questionnaire data

• Generated using a multimodal model + Grad-CAM

👉 Acts as a mediated interpretation layer

Finding:

• Improved communication with AI

• Required natural language explanations to build trust

---

The Study Setup (Why This Matters)

The AIDED interface where designers interact with client data and iteratively refine designs using AI.

This isn’t a conceptual framework, it was tested.

• 12 professional designers

• 4 conditions (C1–C4) using different data types

• Each designer:

  • iteratively modified designs

  • evaluated trust, effort, and satisfaction

👉 This gives weight to the findings, it’s based on actual workflow behavior

---

The Most Important Insight

The Authenticity–Interpretability Trade-Off

The paper identifies a core tension:

• Authentic data (like gaze)

  • rich, close to real experience

  • harder to interpret

• Structured data (like questionnaires)

  • easier to use

  • less nuanced

This trade-off directly affects:

• cognitive effort

• trust

• design decisions

---

What Designers Actually Did Differently

The iterative workflow showing how designers use client data and AI prompts to refine design outcomes.

1. Structured data led to better decisions

Designers:

• trusted it more

• used it more confidently

• produced more satisfying outcomes

👉 Because it was clear and actionable

---

2. Raw experiential data wasn’t enough

Even though gaze data reflects real behavior:

• designers struggled to translate it into design actions

👉 Insight:

More data doesn’t automatically improve design, it needs interpretation

---

3. AI’s role shifts from generator to mediator

A key contribution of the system:

• AI doesn’t just generate images

• It helps translate client signals into design-relevant insights

This becomes especially clear when:

• overlays are paired with LLM explanations

• turning visual signals into actionable suggestions

---

What This Means for AEC Workflows

1. Client data needs structuring to be useful

• Designers already rely on:

  • questionnaires

  • checklists

• The study confirms why:

  • they reduce ambiguity

  • support better decisions

---

2. More data ≠ better outcomes

• Raw signals (like gaze) can:

  • increase cognitive load

  • slow down interpretation

👉 Tools need to translate, not just collect

---

3. Explainability is critical for adoption

• AI-generated insights alone are not enough

• Designers need:

  • reasoning

  • context

👉 This directly affects trust

---

4. Designers remain central to decision-making

The system:

• supports designers

• does not replace them

Designers still:

• interpret trade-offs

• balance feasibility

• maintain control over final decisions

---

Limitations Highlighted by the Study

• Small sample size (12 designers)

• Focused on residential interior design

• AI outputs still require:

  • human validation

  • interpretation

👉 This is an early-stage exploration

---

Final Takeaway

The contribution of this work is not about better image generation.

It’s about something more specific:

How to make human experience usable inside AI-driven design workflows

And the key lesson is simple:

• Raw human data is valuable

• But only when it becomes interpretable and actionable

---

This blog post is based on research by Yang Chen Lin, Chen-Ying Chien, Kai-Hsin Hou, Hung-Yu Chen, and Po-Chih Kuo, published in the paper “AIDED: Augmenting Interior Design with Human Experience Data for Designer–AI Co-Design.”

---

If you are exploring ai in construction or need support in GTM for your construction tech startup, book a discovery call using this link :

https://calendly.com/mayur-mistry7/consultancy-discovery-call