LayoutAI

Modern analytics platforms rely heavily on manual dashboard construction, requiring users, often non-designers, to make repeated low-level layout decisions involving alignment, spacing, hierarchy, and visual balance. While these decisions are essential for readability, they contribute little to analytical insight and impose significant cognitive and temporal overhead.

Recent advances in AI-assisted design raise an important research question:

How can AI assist with repetitive interface layout tasks while preserving user agency, trust, and interpretability?

This project explores that question through the lens of Human-Centered AI (HCAI) and human–AI collaboration, focusing specifically on dashboard layout authoring.

Research Framing (HCI Perspective)

This project sits at the intersection of:

Human–AI Interaction (HAI)
Cognitive Load Theory
Sensemaking in Data Visualization
Mixed-Initiative Systems
Explainable & Trustworthy AI

Rather than treating AI as an autonomous decision-maker, this work frames AI as a collaborative assistant that supports, but does not replace, human judgment.

Context & Problem Definition

The Existing Workflow

In typical dashboard authoring tools, users begin with a blank canvas and manually:

Place charts and tables
Adjust alignment and spacing
Iterate through trial-and-error to achieve visual balance

This workflow disproportionately affects:

Non-designers
First-time or infrequent users
Analysts focused on insight rather than presentation

Observed Challenges

Repetitive manual placement of components
Difficulty producing visually coherent layouts
Lack of starting guidance, increasing time-to-first-insight
High extraneous cognitive load during early stages

Research Methods

Although exploratory, this project was grounded in established research methods:

1. Literature Review

Human-Centered AI design guidelines
Explainable AI (XAI)
Mixed-initiative interaction models
Cognitive load in interface authoring

2. Task Decomposition

I decomposed dashboard layout authoring into:

High-level analytical intent (what insight is needed?)
Low-level layout mechanics (spacing, alignment, ordering)

This revealed a key insight:

Layout mechanics are repetitive, low-risk, and ideal candidates for AI assistance.

3. Pattern Analysis

Based on repeated dashboard examples, I identified common layout patterns:

Time-series prioritized in the upper-left regions
Summary KPIs are placed above detailed charts
Consistent grouping based on data relationships

4. Conceptual Prototyping

I designed LayoutAI as a conceptual AI assistant, focusing on:

Interaction flow
User control points
Transparency mechanisms

Design Principles (Human-Centered AI)

The system was explicitly aligned with HCAI best practices:

Human Control: Users can accept, modify, or reject AI suggestions
Transparency: AI explains why a layout was suggested
Low-Risk Automation: AI targets repetitive tasks, not analytical reasoning
Learning Through Feedback: System adapts to user corrections over time
Ethical Awareness: Acknowledges bias risks in learned layout preferences

Conceptual System Design (Layout AI)

System Role

LayoutAI functions as a mixed-initiative assistant, embedded within the Host Layout canvas.

It does not finalize layouts autonomously; instead, it:

Proposes starting configurations
Explains its reasoning
Learns from user interaction

Interaction Flow

Explainability & Trust Design

A key research contribution of this project is the explanation layer.

Each layout suggestion includes:

Visual callouts highlighting key placements
Plain-language explanations (e.g., “Time-based trends are placed in the upper-left for faster scanning”)
Confidence indicators to calibrate trust

This design explicitly supports:

Trust calibration
Learning
Reduced automation anxiety

Key Features (Research-Oriented Framing)

Layout AI: Let the AI arrange your dashboard while you focus on insights

LayoutAI is an AI-powered assistant embedded into Host Layout (A blank canvas to begin creating a report dashboard). It helps users by generating smart starting layouts that can be customized and improved over time

Automated Starting Layouts

Reduces blank-canvas paralysis and accelerates onboarding

Adjustable Parameters

Allows users to steer AI behavior rather than accept fixed outputs

Learning Feedback Loop
Captures user corrections to refine future suggestions

Transparency Panel

Explains decision rationale and surfaces system intent

Evaluation Strategy (Proposed)

Although not deployed, I defined a clear evaluation plan:

Quantitative Metrics

Time-to-first-layout
Number of manual adjustments
Task completion time

Qualitative Metrics

Perceived trust in AI suggestions
Sense of control
Cognitive workload (NASA-TLX style assessment)

Research Contributions

This project contributes a design-oriented investigation into:

Human-centered automation in interface design
Explainable AI for creative tasks
Balancing efficiency and user agency in AI-assisted tools

Reflections & Learning

Transparency is essential for user confidence in AI systems
Automation is most effective when scoped narrowly
AI should support thinking, not replace it
Designing for non-experts requires explicit scaffolding and explanation

Future Research Directions

Longitudinal studies on AI-assisted layout learning
Comparative evaluation of AI vs template-based starting points
Exploration of bias accumulation in adaptive layout systems

Why This Project Matters

LayoutAI demonstrates my ability to:

Translate HCI theory into system design
Design AI features responsibly
Frame design work as a research inquiry
Communicate complex ideas clearly

This project is not a speculative feature; it is a research artifact exploring the future of human–AI collaboration in analytical systems.