Zero is a smart home system that helps you reduce food waste by learning about your consumption habits and generating shopping lists based off of wasteful habits in the kitchen. It is an accessible way to lessen food waste in everyday life, making combating food waste as easy as recycling.

Team: Hanyuan Zhang ↗, Meredith Newman ↗
Role: User Interface, User Research

Zero is devised for a class that prompted us to apply artificial intelligence towards a domain of our interest. I was paired up with two other designers from Carnegie Mellon to work on this project. Given the prompt, we were immediately drawn to the idea of applying artificial intelligence to the topic of wellness. We considered physical and mental wellness, considering the anxiety of traveling and seeing cooking as a social experience. By considering food as wellness, we ran into interesting discussions about distribution of food and its consumption. 

Our focus on food and artificial intelligence led us to research done around attitudes and behaviors of home cooking. A lot of interesting research has been done around food consumption at home and its implications on cultural and social norms. For example, the tradition of home cooking is much more revered for the French than it was for the British. Furthermore, the introduction of the Swanson TV dinner in 1953 started the trend for convenience foods, leading a decline in home cooking. As prepared foods become more accessible, there is a general perception that cooking is a social activity. People find that cooking requires extensive planning and the waste that it generates is not worth it. From there, we came to the topic of exploring the kitchen and how AI could encourage home cooking. This would entail changing perceptions about cooking at home. However, this focus was too narrow and we needed to explore the whole ecosystem of the food industry to pinpoint where our intervention would take place.

In an effort to better understand the extent of our scope and identify what processes would be optimized with the introduction of artificial intelligence, we set out to better understand the people involved in the production of food. We reached out to Market Street Grocery, a grocery store downtown that sells a small selection of produce as well as breakfast, lunch, and dinner, to talk about their practices in consumption and repurposing food. Because of the size of the grocery store, Rachel was able to manage waste by repurposing produce that are close to its expiration dates into ingredients for prepared foods. The way that Rachel stocks her store is also largely dependent on past experience, what’s seasonal, and customer requests.

We also corresponded with a local food co-op, East End Co-Op, to learn about their habits, and we found that unlike Market Street Grocery, unsalable produce (blemished, etc.) either get composed or given to staff, presumably due to the larger size of the East End Co-Op. He also had an interesting insight about the fact that although a higher percentage of those who shop at East End Co-Op have a focus on sustainability, it is very difficult issue for people to say that they’re sustainable and do it in practice. Fads around sustainability come and go very quickly.

Beyond understanding food systems, we also wanted to get a better understanding of the technology that goes behind machine learning and artificial intelligence. We talked to Amanda Coston, a PhD candidate in Machine Learning & Public Policy joint PhD program at CMU. Her focus area is applying machine learning to areas of social impact, including healthcare, human rights and social justice. From talking to her, we realized that machine learning is very good at seeing patterns and predicting behavior but it's bad at deciphering why people act in certain ways. The implications of behavior is still the job of humans, and it’s difficult to change people’s behavior but what can be done is to use machine learning to heighten awareness.

Through talking to various specialists on food waste and consumption through the lens of grocery stores, we realized that because sustainability and waste has a direct correlation on profit models in grocery stores, there is more incentive to not be wasteful. Most grocery stores have a better understanding of consumption patterns and are aware of their stock, unlike customers. Instead of creating a product that would be grocery store-facing, we realized that there is more value in changing the grocery store experience for customers to initiate better habits in waste and sustainability.

After gaining a better understanding of how grocery stores are run, we wanted to understand how grocery store shopping is like for the consumer. We sent out a survey through Mechanical Turk to understand the frequency by which grocery shoppers go to the store, their spending habits, and what their habits are like when it comes to food waste.

Through this research we came to the understanding that people who shop less often buy less produce and that the more frequent people shop, the higher possibility that they would be spending more on a trip to trip basis. A majority of participants whose produce lasts more than a week (50% of the responses) still go to the grocery store at least once every week. From this understanding, we created personas that that matched the needs and wants of different groups.

Through synthesizing our research, we set out to create something that is customer-facing and focuses on bringing habits in the kitchen into consideration when shopping at grocery stores. Users want to make the grocery shopping experience for the household more transparent and have a better understanding of what goes in and out without needing to do too much. We found that machine learning could benefit this process the most by allowing for a way to track waste produced by the household. Many participants identified problems dealing with roommates or family members and the lack of transparency between each other's consumption and waste habits. A significant amount of people also depend on shopping lists as the main determinant of what they buy at the grocery store, though few had any idea about items that were consistently not been eaten or consumed in the household.

To address concerns about cataloguing intake, one of Zero's product features is an in-app scanner that catalogs items based on what the user buys. Alternatively, they may also add grocery store memberships to sync bought items in app and be notified of promotions or coupons. Sensors are placed in the kitchen both over the kitchen counter and above trash cans to collect data based on activity, using machine learning to recognize gestures and recognize food items that had been previously scanned. Zero also has a dashboard that gives insight on household consumption and waste. An easily accessible 'add' button allows users to add food items or scan receipts, making manual input as straight forward as possible. The collaborator mode in Zero allows household members to have equal say in what gets bought and also allow for their intake to be tracked. The dashboard will be shared by the household and used to generated suggested list of items to buy when in the grocery store. Those contributing to the app are also able to add wish list items, though in the case of a child contributor, wish list items will be sent to be confirmed by the main app holder. We also want to give user incentives by creating milestones that encourage more sustainable practices.

Through doing this project, we gained a better understanding of the limitations and reaches of artificial intelligence. Design’s role in AI is in facilitating communication and data to affect changes in behavior. Zero strives to do that with a presence in the home as well as a constant reminder through the app. In using sensors throughout the house, the product allows for better accuracy and a changed way of perceiving interaction within the household. By instituting AI, designers are given the opportunity to better make sense of its user base and visualize information to impact the user for better.