Project Summary:

Fine Foraging is an Alexa cooking skill that suggests quick, simple recipes using commonly foraged ingredients. I designed it as part of a Voice User Interface Design certification course through Career Foundry.  The project required that I create ten different recipes involving at least three steps. Recipes had to be included for breakfast, lunch, dinner, and snack buckets. 

​

Methods and Tools:

• User Stories

• Sample Dialogs

• User Flow

• Miro

• User and System Personas

• Lucidchart

• Voice Script

• Usability Testing

• VoiceX Prototyping Tool

• Node.js

 

Challenge:

In order to provide a delightful user experience, I needed to find solutions for the numerous constraints in the project criteria and code template such as:

• Limiting recipe steps due to the seven second time limit built into Alexa's responses.

• Finding a unique recipe theme due to the hundreds of existing cooking skills for Alexa.

• Maintaining a light-hearted tone and voice while being informative.

• Reducing cognitive load with recipe steps.

 

​The beginning of the project brief stated that, "constraints actually encourage creativity," and I found this to be true. I was able to design a useful and unique cooking skill despite the challenges. I accomplished this by conducting careful and thorough research.

 

Research and Findings:

Since there are already so many cooking skills for Alexa, finding a novel idea was not a simple task. I began by looking through the existing 40+ pages of cooking skills available for Alexa.  It seemed like there was a cooking skill for every facet of cooking, but I was surprised to see that there wasn't anything focused on foraging food.  I have been interested in foraging food since my childhood when we would go hunting for morel mushrooms and nettles every spring. Foraging for food is ecological, and promotes a positive connection to the environment.  I had found my concept! 

 

The next step was to do some conversation mining to find engaging language and terms to use in the skill.  I started by looking at recipe blogs focusing on food foraging.  I quickly realized that the style of these sites didn't match what I wanted to accomplish in a concise yet light-hearted tone.  Often, I found myself scrolling down past repetitive and tedious exposition to get to the information I was looking for.  I did come across the phrase, "when in doubt, throw it out" on numerous sites, and it inspired the rest of the voice script that I used in my skill.  It was memorable and catchy which is ideal for a voice-only interface. This phrase was a perfect motto not only for foraging, but also for my writing and editing. 

 Before I began designing User Personas, a System Persona, and Sample Dialogs, I created a word bank based on some of the terms I saw frequently used on foraging sites such as, "environmentally- friendly, sustainable, seasonal, safe, responsible sources, wild food, eat weeds, eat invasives, family foraging, tasty, satisfying, and nutritious."

 

This word bank helped inform my User Stories, Sample Dialogs, and User Flow. I then began writing a Voice Script and creating recipes. I knew that complex recipes wouldn't be possible, so I looked for recipes that would be very simple. I started talking to friends and family in a live conversation mining process which changed my thinking about how recipes work in conversation.  I noticed that if I asked a friend who is a chef, "What would you make with ramps?" They didn't respond with an entire recipe. They would assume I had a basic command of cooking skills and would say, "put them in an omelet, or add to a frittata." I decided to use this format for my recipes since it matched the natural way people talk about using ingredients. People instinctively chunk information when talking about recipes which reduces cognitive load. I wrote recipes and reduced steps and measured ingredients wherever possible.

​

Usability Testing:

User tests were conducted remotely with Zoom. The system prompts and responses were played using the VoiceX Prototyping tool. There were five participants ranging in age from 19-47 and were all frequent voice system users as well as farmers market shoppers or foragers.  

​

Test Objectives:

• Identify any sticking points in the user flow.

• Determine if the novice prompts are clear and help users to navigate easily.

• Evaluate disconfirmation handling.

• Improve usability and navigation.

• Collect additional utterances.

​

Test Results:

• Most people wanted to search by meal type or ingredient, and they tried despite the scenario having them search by season. I decided to eliminate season as a search option and have it a default state.

• Everyone had great suggestions for expanding the skill with multi-modal options. I am planning to work with a developer to add images and recipe cards.

• Even though users knew the test wasn't with an actual Alexa device they still responded with utterances as if it was Alexa.  Everyone had different utterances for each intent, but they used that utterance repetitively. One user said, "Alexa, can you.." and another used, "Alexa, show me..." Many were not utterances I was expecting, so it was helpful to collect them.

• It was difficult for some of the users to listen to the list of ingredients on a recipe they had decided to deconfirm. It was clear that barge-in should be available for the list of ingredients.

• Once user recommended there be an option to identify some of the foraged ingredients that aren't commonly known like salsify and ramps. I will be adding identifying information for these ingredients.

​

Outcomes and Feedback:

I submitted Fine Foraging through the Alexa Developer Console, and it was approved.  A few months later I received an email from the in-house Alexa skill experts saying that they liked my skill and thanking me for publishing it.  They also offered some additional recommendations, that if implemented would make my skill eligible to be featured in marketing channels. Their recommendations were:

• ​Add Multimodal Experience

• Improve the Natural Language Understanding Accuracy

• Add Home Cards

• Add Multiple Language Capability

​

I agree with all of these suggestions especially as they were consistent with the findings of my usability testing.  I think Home Cards and Multimodal images and sounds would add much to the experience. I have only noticed one NLU accuracy issue which is with the skill invocation name, so I plan to work on improving that.  I have discussed editing the code with a developer, and look forward to implementing the suggestions from the in-house Alexa Skill Experts.