Personalized meal planning system incorporating ChatGPT

AI Design ・ Prompting Engineering

Overview

The AI-Powered Meal Planner project highlights my design process from initial concept to iterative improvement. Starting with a simple text box for dietary preferences, I identified user challenges and evolved the interface to include selection options and a custom text box for special needs. This approach improved user guidance and ensured practical AI-generated meal plans.

 

This project demonstrates my commitment to user-centric design, leveraging AI to create a more dynamic and personalized meal planning experience.

Timeline

Completed in 3 weeks (2024.05)

Role

UX/UI Designer
AI Engineer

Tool

ChatGPT

Context

Addressing limitations of existing planners and
Deepening AI knowledge

From my previous project—a planning assistance app—I discovered that existing meal planners often rely on pre-existing recipes, which limits selection and personalization.


As a designer, I also wanted to deepen my understanding of artificial intelligence and its capabilities, and to explore how I could incorporate AI into my products.

How might we
Utilize AI to create a flexible and adaptive meal planning system that caters to diverse user needs with minimal limitations?

How might we
Utilize AI to create a flexible and adaptive meal planning system that caters to diverse user needs with minimal limitations?

Design Process:

My AI Journey: Beginning with ChatGPT

To begin, I started by asking ChatGPT to guide me and help write the code necessary for my project. Through this process, I learned how to build my system step-by-step, leveraging the capabilities of AI to streamline development.

 

While using ChatGPT 4.0, I encountered some bugs in the code. To resolve these issues, I sought out resources such as YouTube tutorials, Velo documentation, and advice from friends.

Understanding how AI is integrated into the website

Optimizing prompting to achieve better results from AI

To guide ChatGPT to yield better results, a couple prompt engineering techniques were implemented:


  • Use system message to adopt to a persona

  • include details in query

  • provide examples

Design iteration 1

Simple text box for users to input their dietary preferences

#1 Allows users the flexibility to comprehensively articulate their specific needs and preferences.


#2 The format of text box leverages users' familiarity with text prompts, clearly indicating that this is an AI-powered meal planner.

Key Issues Identified through User Testing

#1 Entering all information in a plain text box is challenging

Meal planning involves various parameters such as goals, dietary preferences, and allergies. It was challenging for users to enter all this information at once in a plain text box.

Key Issues Identified through User Testing

#2 AI's responses were overly creative and impractical

The AI's responses tended to be overly creative and sometimes impractical. For example, if a user expressed a love for chocolate, the program might suggest incorporating chocolate into every meal, which is not realistic.​

Design iteration 2

Improved User Interaction and Practicality

To address these problems, I added basic selection options while keeping a custom text box at the end for special needs.

This approach solved the issues identified in the first prototype:​

#1

​​​Enhanced User Guidance: Using selection options to guide users reduces the cognitive load, making it easier and more comfortable for them to start.


#2

Practical Constraints: This method ensures the AI's output remains within practical constraints, providing more realistic and useful meal plans.

Next Step

This is an ongoing project that I'm continuously iterating on.

Here are the next steps I would like to elaborate on:

Fine-Tuning the model

Streamlined Design Integration

Use ChatGPT's fine-tuning feature to improve the accuracy and relevance of meal suggestions, making them more personalized and practical.

Improve Processing Time

Personalized and Automated Interaction

Optimize server performance, streamline code, and enhance API efficiency to reduce processing time, ensuring a smoother user experience.

Add Feedback Loop

AI-Enhanced Methodologies

Add a chat box below where users can prompt for revisions after reviewing the recipe. This feedback will be incorporated into the model to continually improve the AI's meal planning output.