ADD FOOD: Leveraging the Edamam API for Tailored Dietary Solutions for ADD Individuals

Project Diary: Jim Verheijen, MA Data Driven Design Student 1712516

 

Premise:

An adaptive digital solution designed to assist individuals, especially those with ADD, in making healthier dietary choices. The project leverages the capabilities of the Edamam API (Edamam, 2023) to provide recipe recommendations based on available kitchen ingredients.

 

Synopsis:

In a world faced with dietary challenges, individuals with ADD confront unique obstacles, like impulse-driven choices leading to unhealthy diets and potential food waste. Addressing this, our project introduces a tailored digital platform using the Edamam API, offering recipe recommendations based on available ingredients. This isn’t just a meal suggestion tool; it’s designed with ADD individuals’ specific needs in mind. By merging with Edamam’s vast recipe database, the platform promotes healthier eating habits, reduces food wastage, and eases decision-making. The system stands as a beacon of how technology, when aligned with deep empathy and understanding, can transform dietary choices, especially for those grappling with ADD’s complexities.

 

Research & Problem Space:

The research addresses the intersection of technological advancements and socio-cultural implications, highlighting the critical role of ethical considerations in technology use.

The pain points that are identified:

  • Dietary Challenges: ADD individuals often face impulse-driven eating habits.
  • Decision Fatigue: The vast array of online recipes can be overwhelming, making meal planning taxing.
  • Unhealthy Convenience: The difficulty in decision-making might lead to unhealthy, quick food choices.
  • Potential Food Waste: Impulsive buying and decision fatigue can result in unused ingredients.

 

The central research question is: “How can the Edamam API assist individuals with ADD in developing healthier dietary habits?”

 

APIs (Application Programming Interfaces): APIs are tools that allow software applications to communicate and share data (Redhat, 2022). They enable the integration of specific functions or data sources from third parties. For this project, the Edamam API will be utilized.

 

ADD (Attention Deficit Disorder): ADD is a neurological condition characterized by attention difficulties and hyperactivity and one’s acting on impulses (Janssens, 2022). It’s a subtype of ADHD, where the individual experiences inattention without notable hyperactivity.

 

The Double Diamond process, comprising the four phases of Discover, Define, Develop, and Deliver, was chosen to navigate the project’s complexities. Grounded in critical-making principles, this approach guarantees designs remain user-centric and reflective throughout  (Ratto, 2011).

 

Discover:

In the “Discover” phase, understanding ADD individuals’ dietary challenges was paramount. They often struggle with impulse control, affecting their eating habits (Margues, 2020; Pinto, 2022). With rising ADD diagnoses (Kazda, 2023) and economic repercussions (Bisset, 2023), AI and data-driven tools stood out as a solution to minimize food waste. Such tools can forecast dietary needs and suggest personalized meals, supporting ADD diets (Rahman, 2023). Despite increased screen time (Rodgers, 23), many recognize the value of digital aids (Kenter, 2022).

 

Define:

Edamam API Exploration & Integration:

Initially, supermarket data scraping was planned for real-time ingredient updates, but the Edamam API largely filled that role. However, the API sometimes provided outdated or missing recipe links, leading to user disruptions.

  • Dove into the Edamam API documentation, aiming to effectively utilize its capabilities and synchronize them with the project’s requirements.
  • Prioritized fast data fetches to ensure timely responses, vital for ADD users. Such optimization was necessary to cater to their quick interaction needs.
  • Merged the API to harness the extensive recipe database
  • Navigated hurdles such as API rate limits.

 

Script Functionality:

  • Modules: Uses the requests module to interact with the Edamam APIs.
  • Recipe API: The code fetches recipes, printing ingredients and related details. This is achieved by interfacing with Edamam’s recipes API.
  • Food Database API: A separate segment fetches data about various foods using the food-database API, revealing food IDs, labels, and categories.
  • Error Handling: If expected data is not received from either API, appropriate error messages are displayed.

 

Lo-Fi Prototype Creation:

A paper-based design was made to display recipes.

  • Swiping Interface: Transitioned from a catalog design to a Tinder-like swipe mechanism for intuitive UX.
  • Design: Created a minimalistic, distraction-free design tailored for ADD users, ensuring focused interaction.
  • User-Centric: The swipe design supports swift decisions, meeting the quick interaction preferences of ADD individuals.

Gathering feedback from 2 developers and 2 individuals with ADD provided pivotal insights:

  • API Constraints: Developers emphasized potential slowdowns with extensive ingredient inputs, steering away from a comprehensive recommendation engine.
  • Database Necessity: Storing user preferences necessitated a separate database, increasing the complexity.
  • User Experience: ADD individuals in an early test favored more intuitive ingredient selection rather than text input into a development environment, resulting in the use of images and potential use of technologies such as image recognitio for input.
  • Recipe Accuracy: Some pointed out mismatches between available ingredients and suggested recipes.

 

Reflection:

In reflection, striking a balance between technical constraints and user needs became evident. I saw areas for improvement in the ingredient input system. While features like bill image recognition or barcode scanning are promising, they currently exceed the project’s scope.

 

Develop:

Initially, users were directly asked, “Do you want to visit the supermarket?” Based on feedback, this question was refined for clarity. Now, users decide if they prefer to cook with available ingredients or are open to new dishes.

The script provides recipe suggestions based on user ingredients:

  1. Modules: Imports recipe_api and food_api for fetching recipe and food data.
  2. Scoring: The function score_recipes_based_on_ingredients ranks recipes by how many of the recipe’s ingredients the user possesses. It also notes what’s missing.
  3. Ingredient Checking: check_availability verifies if user ingredients are in the known food database.
  4. User Input: Users specify if they want recipes based on their ingredients (yes) or random suggestions (no).
  5. Recommendations:
    • For yes: Users input their ingredients. Recipes are then ranked by ingredient matches and displayed.
    • For no: Random recipes are showcased

Terminal

Hi-Fi Prototype Creation:

A hi-fi Figma prototype was created, spanning four screens, showcasing the process from ingredient entry to recipe recommendation.

Second feedback round with four ADD individuals, two of whom reviewed the lo-fi version, insights:

  • “I buy what I see”: Highlighting the ADD challenge with real-world impulses, leading to a more images, which can be profided by the Edamam API.
  • “I need a timer when to cook”: Underscoring the need for cooking reminders and cues to update the app with ingredients potentially keeping track of inventory.
  • “Can it also show what I still need?”: Prompting a redesign in ingredient presentation, emphasizing missing items, while showing what is already there.

 

Reflection:

Engaging with individuals with ADD offered invaluable feedback on the prototype and its features:

  • Highlighting missing recipe ingredients for shopping ease.
  • Timely notifications for cooking and ingredient inventory reminders.
  • Addressing their spontaneous shopping habits, emphasizing a user-friendly interface and potential features like on-the-spot recipe suggestions using image recognition for receipts or barcodes.

 

Deliver:

Revisions were made from user feedback on the hi-fi Figma prototype, leading to new features for improved user experience and alignment with project objectives.

  • Timed Notification: Implemented a timer for meal reminders and inventory updates, crucial for ADD individuals.
  • Smart Shopping: Introduced a structured list to guide purchases, considering ADD impulsivity, with potential future enhancements using image recognition for bills or barcodes.

 

 

Critical making & thinking:

This project, at its core, was an exercise in critical making—a method that marries hands-on production and critical reflection to produce socially beneficial technologies.

Reflective breakdown of the project:

    1. Stakeholder Engagement: Prioritized feedback from ADD individuals, turning the product from a generic tool into a tailored solution.
    2. Iterative Approach: Stayed adaptive, knowing the initial version required refinements.
    3. Focus on Core Features: Chose clarity and simplicity over multiple features, aligning with the ADD perspective.
    4. Understanding Real-world Issues: Addressed real challenges faced by ADD individuals, like impulsiveness, rather than just offering recipes.
    5. Future Image Recognition: Considered receipt scanning for immediate ingredient updates and recipe suggestions, streamlining the user experience.
    6. Wide Appeal: The app’s design can benefit those aiming for efficient meal planning or reducing food waste.
    7. Smart Appliance Syncing: Future versions may connect with smart fridges for real-time ingredient tracking.

 

Limitations:

  • API Rate Limits: The Edamam API’s rate limits could cause occasional delays during peak times, impacting the user experience.
  • Database Dependency: Relying on Edamam means limited control over data accuracy. Outdated or missing recipes could disrupt user trust.
  • Third-party Constraints: Dependency on the Edamam API reduces our system’s adaptability and binds it to a single data source.
  • Recommendation Staticity: Our system uses static data, whereas advanced recommendation engines adapt to user preferences over time.

 

Suggestions for Improvement:

  • Integrate multiple APIs to diversify recipes and reduce single-source dependency.
  • Develop a system blending static API data with a dynamic recommendation engine, tailoring suggestions based on user behavior.

 

Conclusion

In conclusion, the Intelligent Recipe Recommendation System underscored the transformative potential of API technology, specifically the Edamam API. This wasn’t just a venture into creating a platform, but an exploration of how the right technological choice can address the nuanced challenges of ADD individuals. The Edamam API exemplified how real-time, dynamic data integration can enhance user experience and efficiency. In essence, the project serves as a testament to the impactful solutions that emerge when technology is combined with deep empathy and user-centric design.

 

Bibliografie

Bisset, M. (2023, January 18). Practitioner Review: It’s time to bridge the gap – understanding the unmet needs of consumers with attention-deficit/hyperactivity disorder – a systematic review and recommendations. Retrieved from: The Association for child and adolescent mental health: https://acamh.onlinelibrary.wiley.com/doi/full/10.1111/jcpp.13752

Edamam. (2023). Edamam. Retrieved from: Edamam Home: https://www.edamam.com/

Janssens, A. (2022, June 20). Wiley Online Library. Retrieved from: Parenting roles for young people with attention-deficit/hyperactivity disorder transitioning to adult services: https://onlinelibrary.wiley.com/doi/full/10.1111/dmcn.15320

Kazda, L. (2023). Attention Deficit Hyperactivity Disorder (ADHD) Diagnosis In Children And Adolescents: Trends And Outcomes. Retrieved from: The University of Sidney: https://ses.library.usyd.edu.au/handle/2123/31065

Kenter, R. M. (2022, October 21). Internet-Delivered Self-help for Adults With ADHD (MyADHD): Usability Study. Retrieved from: JMIR Publications: https://formative.jmir.org/2022/10/e37137/

Margues, I. (2020, May 12). Effect of Impulsivity Traits on Food Choice within a Nudging Intervention. Retrieved from: National Library of Medicine: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7285079/

Pinto, S. (2022, October 16). Eating Patterns and Dietary Interventions in ADHD: A Narrative Review. Retrieved from: National Library of Medicine: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9608000/

Rahman, M. M. (2023, April 11). AI for ADHD: Opportunities and Challenges. Retrieved from: Sage Journals: https://journals.sagepub.com/doi/full/10.1177/10870547231167608

Ratto, M. (2011, July 15). Critical Making: Conceptual and Material Studies in Technology and Social Life. Retrieved from: Taylor & Francis Online: https://www.tandfonline.com/doi/abs/10.1080/01972243.2011.583819

Redhat. (2022, June 2). What is an API? Retrieved from: Redhat: https://www.redhat.com/en/topics/api/what-are-application-programming-interfaces

Rodgers, A. L. (2023, April 28). ADHD Brains on Screens: Decoding a Complicated Relationship. Retrieved from: additudemag: https://www.additudemag.com/screen-time-video-game-technology-dependence-adhd/

 

Read more "ADD FOOD: Leveraging the Edamam API for Tailored Dietary Solutions for ADD Individuals"

Utilizing Recommender Systems to Optimize Energy Usage of Dutch Households by Presenting Energy-Saving Suggestions

PREMISE  How can a Recommender System contribute to the reduction of household electricity consumption? SYNOPSIS In the midst of a regional energy crisis, exacerbated by the enduring COVID-19 pandemic and Russia’s manipulation of the European Union’s energy markets, the EU is actively exploring strategies to combat the issue. These strategies involve the implementation of digital […]

Read more "Utilizing Recommender Systems to Optimize Energy Usage of Dutch Households by Presenting Energy-Saving Suggestions"

Utilizing Recommender Systems to Optimize Energy Usage: Empowering Dutch Households with an Energy-Saving Chatbot

PREMISE  How can a chatbot be an energy saver: Using a chatbot (with a recommender system back-end) as an energy assistant for reducing electricity consumption in Dutch households.  SYNOPSIS This study aims to develop a chatbot with a recommender system back-end that can help households reduce their electricity consumption by offering usage patterns and recommending […]

Read more "Utilizing Recommender Systems to Optimize Energy Usage: Empowering Dutch Households with an Energy-Saving Chatbot"

Using AI for designing tattoos to prevent decision paralysis among young women

Premise
Using artificial intelligence technology for designing tattoos to help those who are experiencing decision paralysis when choosing tattoos.
Synopsis
Today, too many different tattoo designs can be found all over the internet and on social media; however, like any form of consumption, people who want to get tattoos can be overwhelmed by the choices available and face decision paralysis, which may take them the time and energy to overthink and cause anxiety, so it is essential to think about a solution to help them choose a design that matches their preferences. One potential solution could be creative design using artificial intelligence technology. AI has been successful in many creative areas, such as producing imaginary content and images. In addition, recommender systems are effective for finding the right item of content when individuals are overwhelmed by decision paralysis. For this project, I developed a prototype of an app that generates tattoo designs based on the user’s imagination or preferences and tested its effectiveness with young women.
Substantiation
A tattoo is a form of body modification where ink is injected into the skin to create a permanent design for various reasons, including self-expression, individuality, and association. (Sun, Z.H. et al., 2016). Tattoos are no longer considered a stigma in society, and throughout the past few decades, there has been a noticeable increase in the number of people acquiring tattoos (McCandlish et al., 2023). For instance, in Amsterdam, the number of tattoo shops has increased from 60 in 2009 to 109 in 2019 (Ozcan, 2023). However, the choice of design among too many designs, which are constantly updated and promoted on social media, can be overwhelming for people who want to get a tattoo. According to the choice overload theory, having more options to choose from may lead individuals to encounter difficulty deciding (Scheibehenne et al., 2010). Inability to decide as a result of excessive overthinking is known as decision paralysis. Some studies show that analysis paralysis in various businesses is a result of choice overload (Adriatico et al., 2022). The phenomenon of decision paralysis consists of regret anticipation, overthinking, inaction, and delay and may affect people’s decision process (Manolică et al., 2021). To design the prototype, I chose AI because AI creativity has been making a significant impact in various fields, bringing new ideas and possibilities to individuals (Wu et al., 2021). AI tattoo design tools can help to democratize the design process and enable more people to create custom tattoos that reflect their personal style and preferences. And a recommendation system (RS) would be useful to recommend an item to a user based on given information (Fayyaz et al. 2020). So, my approach involves using machine learning algorithms to identify and recommend a small set of designs that are most likely to meet a user’s preferences. Choosing the best-fitting tattoos for young people is related to some factors, including the design, the tattoo designers’ copyright rules, and the meaning and association with specific groups in society. However, my specific scope in this project is only to research and design a tool to prevent young women’s tattoo decision paralysis so, the other factors were not considered.
Iteration 1
For the first iteration, a python code was written that used an open AI API key to connect to server DALLE 2, which is an AI system that can create realistic images from a description prompt in natural language. And tested with four young women aged 19–24; one of them had tattoos and would like to have more, and the rest of them said they think about getting tattoos. They were asked to feel free about the sentences or phrases they entered into the prompt; however, they had to write the phrase ‘a tattoo design’ at the beginning of the last sentence to make sense in natural language for AI. Then they were asked about their experience.
Insights from the testing first iteration:
• They were overly excited and quickly produced image after image.
• They had to add fine-line phrases in their sentences to get a fine-line tattoo image.
• One of the users had a problem with the interface and asked for a more user-friendly interface.
• They could not produce a tattoo design with politicians and got errors (based on DALL-E 2’s content policy limitations).
• When they reversed the sentences, they got considerably different images. For instance, these two images generated considerably different.


sunset tattoo design                                               

tattoo design with sunset.png

• Tuning prompts are important. According to open AI, better prompts could improve the performance of the images to give good results about 50% to 65% of the time, but often not beyond.


Iteration 2
For the second iteration, A user-friendly interface was created. And a tattoo design with a phrase adds to the prompt as a default text in case users will forget to type it. They could even totally omit this phrase or put it at the end of the sentence.
Insights from the testing second iteration:
• Although they were free to choose one specific design, they created one based on their preferences, and it was still difficult for them to choose.
• After testing five or six images then they felt a lack of imagination and took time to think of new sentences.

Iteration 3
Since recommendation systems offer suggestions to navigate users to the items which are most likely of interest to them efficiently (Burke, 2007; Gorgoglione et al., 2019). For the third iteration, I used the non-personalized recommendation system method because the App was tested with random users so, there was not any form of collaboration between users that they could suggest together. with an API, I got a CSV file dataset of all objects of the New York Metropolitan Museum of Art. And then selected only the paintings titles (5926 titles) column from all objects and save it into a CSV file. And then a python code was written to generate random titles.



Insights from the testing second iteration:
• One of the users suggested it would be great if they are recommended by specific categories. For instance, nature, animals, and historical or cultural symbols.

Conclusion
This app was written for young women who have difficulty choosing when they have a variety of tattoo design options and tested with potential users who were faced decision paralyzed when they offered too many products. overlay, users were satisfied with the creative AI art and saved most of the tattoo design for themselves. possibly they will use them as a tattoo design. However, there is a lot of room for improvement in the prototype and the interface for the future. I will work on more machine-learning language to find better solutions to tune prompts.

References
Sun, Z. H., Baumes, J., Tunison, P., Turek, M., & Hoogs, A. (2016, December). Tattoo detection and localization using region-based deep learning. In 2016 23rd International Conference on Pattern Recognition (ICPR) (pp. 3055-3060). IEEE.
McCandlish, C., & Pearson, M. (2023). Tattoos as symbols–an exploration of the relationship between tattoos and mental health. The Journal of Mental Health Training, Education and Practice.
Ozcan, U. (2023). The health of tattoo artists. EUR J ENV PUBLIC HLT. 2023; 7 (2): em0131.
Scheibehenne, B., Greifeneder, R., & Todd, P. M. (2010). Can there ever be too many options? A meta-analytic review of choice overload. Journal of consumer research, 37(3), 409-425.
Adriatico, J. M., Cruz, A., Tiong, R. C., & Racho-Sabugo, C. R. (2022). An analysis on the impact of choice overload on consumer decision paralysis. Journal of Economics, Finance and Accounting Studies, 4(1), 55-75.
Manolică, A., Guță, A. S., Roman, T., & Dragăn, L. M. (2021). Is consumer over choice a reason for decision paralysis? Sustainability, 13(11), 5920.
Wu, Z., Ji, D., Yu, K., Zeng, X., Wu, D., & Shidujaman, M. (2021). AI creativity and the human-AI co-creation model. In Human-Computer Interaction. Theory, Methods, and Tools: Thematic Area, HCI 2021, Held as Part of the 23rd HCI International Conference, HCII 2021, Virtual Event, July 24–29, 2021, Proceedings, Part I 23 (pp. 171-190). Springer International Publishing.
Fayyaz, Z., Ebrahimian, M., Nawara, D., Ibrahim, A., & Kashef, R. (2020). Recommendation systems: Algorithms, challenges, metrics, and business opportunities. applied sciences, 10(21), 7748.

Burke, R. (2007). Hybrid web recommender systems. Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 4321 LNCS, 377–408.

Gorgoglione, M., Panniello, U., & Tuzhilin, A. (2019). Recommendation strategies in personalization applications. Information and Management, 56(6), 103143.

Read more "Using AI for designing tattoos to prevent decision paralysis among young women"

USING RECOMMENDER SYSTEMS TO HELP OVERCOME CHOICE OVERLOAD EXPERIENCED BY ASPIRING DESIGNERS WHEN SELECTING SUITABLE TYPOGRAPHY

PREMISE Using Recommender system to assist aspiring designers in font selection suitable for their project based on desired feel. SYNOPSIS With hundreds of thousands of fonts available, many designers, especially aspiring designers struggle with choice overload when selecting fonts. Choice overload can be experienced by feeling overwhelmed with the vast number of options available. Current […]

Read more "USING RECOMMENDER SYSTEMS TO HELP OVERCOME CHOICE OVERLOAD EXPERIENCED BY ASPIRING DESIGNERS WHEN SELECTING SUITABLE TYPOGRAPHY"