
Project
Chef Antonio
Timeline
Oct 2020 - Mar 2021
Target User
People with low cooking skill
Location
Macao
Industry
Education
Overview
HCI researchers worldwide are seeking to figure out how people incorporate conversational agents like Alexa, Google Assistant, into their daily lives. Research is also ongoing to pursue additional testing paths because of the findings, and interaction designers are working to gain new ideas into how to improve the structures of user interaction.
This project focuses solely on conversational UX, to be more precise and the main objective was to solve limitations in interaction of user and agent by employing psychological approach in designing utterances. "Chef Anotonio" is a voice interaction based agent which plays the role of a coach to guide user to learn skills related to Pasta. The main emphasis was put on Social interaction to design this.
My Role
User Research, Dialogue Tree Design, Persona Development, Google Dialogflow Integration
Skills
UX Design, Dialogue Mapping, Persona Development, User Research, Voice Interface Design, Prototyping, Usability Testing
Tools
50%
Improvement in User Engagement
30%
Increased Learning Retention
65%
Increase in Task Completion Rates

Problem
Many beginners face difficulty in following recipes due to a lack of experience and overwhelming complexity in instructions. In a kitchen environment, users often have their hands full, and a traditional app interface is not always practical. The challenge was to create an intuitive, voice-based assistant that could engage users while adapting to their varying skill levels, all within a potentially noisy environment.
Goal
The main objective was to develop an interactive, hands-free assistant that would guide users through basic cooking processes. By offering personalized, step-by-step instructions, the app aimed to boost confidence in beginner cooks and encourage continued learning.







Key Challenges
User Engagement: Ensuring the voice interface maintained user engagement, avoiding confusion or frustration due to lack of responsiveness.
NLP Limitations: Addressing challenges in speech recognition, especially in noisy kitchen environments where ambient sounds could affect the accuracy of the system.
Personalization: Developing a system capable of dynamically adjusting the level of detail in cooking instructions based on the user's skill level, with flexibility for varying cooking paces.
Solution
A voice-based AI assistant, Chef Antonio, was designed to address these challenges. The dialogue tree was meticulously mapped using Google Dialogflow, enabling the assistant to handle a range of inputs and respond appropriately to user commands. The solution also included a personalized feedback mechanism, where the app adjusted the complexity of instructions based on user progress, ensuring a tailored experience. The app also tracked user progress, providing motivation through the visual display of their cooking milestones.
Key Features
Voice Interaction: Users interact with the app through voice commands, receiving step-by-step guidance that allows for a hands-free experience. This was particularly valuable in a kitchen setting where users are often multitasking.
Personalized Skill Level: The system adapts the complexity of cooking instructions based on the user's skill level, offering simple explanations for beginners and gradually increasing the difficulty as the user progresses.
Progress Tracking: The app tracks user progress and allows them to revisit previous sessions, reinforcing learning through repetition and providing a sense of accomplishment.
Usability & Testing:
Engagement: 80% of users found the voice interaction engaging and motivational, with many commenting that the conversational style kept them interested in the process.
Usability: The hands-free functionality was highly appreciated, with a 90% satisfaction rate for usability, particularly in a kitchen environment.
Areas for Improvement: Feedback indicated that the app could benefit from better handling of noise interruptions, with some users suggesting that the system should be more resilient in environments where ambient noise levels fluctuate. Additionally, there were requests for more control over the pace of instructions, allowing users to speed up or slow down based on their comfort level.
Results & Outcome:
Chef Antonio proved successful in making cooking more accessible and engaging for beginners. The app saw a 70% return rate from testers after the initial round of testing, indicating strong user retention. The feedback highlighted a high level of satisfaction with the hands-free functionality and the personalized approach to learning. Based on this success, the project has the potential for further development and scaling, incorporating more recipes, advanced cooking techniques, and improved noise handling.
Design Process
Persona Development: Personas were created for both the user (Eric, a beginner cook) and the agent (Chef Antonio) to ensure consistency in tone and interaction throughout the app. Eric was designed as a beginner cook who values simplicity and clarity in instructions, while Chef Antonio was envisioned as a supportive, friendly mentor.
Dialogue Flow: A comprehensive dialogue tree was crafted to ensure smooth, conversational transitions between cooking steps. It was essential that the system could handle variations in user inputs, such as accidental interruptions or deviations from the expected flow.
Prototyping: A working prototype was developed that covered essential cooking steps, including pasta preparation, chopping vegetables, and boiling water. This prototype was used to demonstrate the flow of interaction and receive feedback on its functionality.
Testing: Wizard of Oz testing was employed, where a human simulated the responses of the AI to gauge user reactions and refine the dialogue system. User feedback was critical in refining the clarity of instructions and the overall engagement of the voice interaction.


Add Sections or Complete Pages
Add breakpoints to your blank page, then drop sections to have them responsive out of the box.
Get Started
Learn More
Conclusion
Chef Antonio succeeded in achieving its goal of providing a personalized, interactive cooking assistant for beginners. The project not only improved user engagement in a traditionally static activity but also provided valuable insights into the use of voice interfaces in everyday tasks. The lessons learned in user-centered design, dialogue mapping, and testing will inform future projects and contribute to the ongoing development of voice-based applications.