In January, 2018, the Board of Trustees set the goal of becoming a carbon-neutral campus by 2037, Mount Holyoke’s bicentennial. So far, the school has done a really good job: reduced electricity use by 16% and thermal energy by 13%. The school has been doing a great job so far, but as members of the community, we wonder: can we contribute to the process as individuals? Thus, our group came up with the idea to develop an App to help the students and faculties to participate in the mission and contribute their own effort.

What it does

It learns the user's long-term food preferences and recommends similar foods according to health-factors and eco-factors. The app can increase the usage frequency by giving feedbacks to users' food consumption, their daily carbon footprint and build a community. Here is the interaction procedure: In this part, we include a complete interaction process, including two parts. Environmental behavioural theory Theory of reasoned action (TRA) and theory of planned behaviour (TPB) are common theories used to investigate the factors associated with an individual’s environmental behaviours. Persuasive technology refers to an interactive product that is designed to evaluate changes in attitudes, behaviours and their interactions

The first part is a pre-survey for first time users, we use environmental behavioral questions and carbon footprint awareness questionnaire (See Appendix 1) to generate personalized eco-friendly scores that’s based on the average carbon footprint of the Moho community. A five-point Likert scale ranging from one to five is used to score responses to each item in the constructs. Each item is constructed according to the national mean and standard deviation. There are two types of questions: a),behavioral questions: “On average, how long does it take you to take a shower? 1, less than 3 mins 2, 3-6 mins 3, 6-9 mins 4, 9-12 mins 5, 12 mins and above We will score this question reversely, add 1 to the total score if the user select ‘12 mins and above. ’ b) carbon footprint awareness: “Global warming problems have consequences for my life.” 1 Strongly Disagree 2 Disagree 3 Neutral 4 Agree 5 Strongly Agree We will add 5 to the total score if the user select ‘Strongly Agree’ The app will map the score to the distribution of all users’ scores and present a ranking. After the new user finished the first part, they will join the other users to the second part: Daily log and how to reduce your carbon footprint. It begins the long-term (stable) food preference elicitation. This is to collect what the user generally likes to eat. The user can browse a full catalog of recipes from Mount Holyoke College’s daily menu, and mark those that he/she has eaten or would like to eat that day. This is also the part where the user can specify their diet restrictions.

In order to find food recommendations which both appeal the user’s appetite, help them to decrease their carbon dioxide emission and have balanced meals, an active learning technique, which is called User Utility Multidimensional Prediction, is used. We modify existing nutrient-based food recommender algorithm to include additional parameters that are used for modelling the dependencies between the environmental component and and the food preference and nutrition component.

How we built it

Our first step is to calculate the average daily carbon footprint per capita. We scraped the numbers of annual water, electricity and fuel usage from the school website, multiply them by the carbon emission factor. There are approximately 2,335 students and 312 staff and faculties at Mount Holyoke last year, so we get the daily average carbon footprint per capita: 36012122.87/365/2647 = 37.27 ton/per day/ per person We also want to take school’s recycling programs into consideration. The following chart is the annual resources recycled. We calculate the offset to be 4 tons/per capita/ per day.

Challenges we ran into

How to enact our privacy policy? How to integrate existing algorithms with nutritional factor database. How to do UI/UX design and use Adobe XD

Accomplishments that we're proud of

Organize and analyze three datasets with 20+ explanatory variables focusing on the energy usage at Mount Holyoke College in recent 10 years Source: Dataset information 1997 -2018 Annual Energy Use in BTU FY 2017-18 Recycling Report 2010-2018 Water Annual Designed a questionnaire to understand a user’s personal habits and diet preference (the design methodology is based on two social theories Environmental Behavior Theory and Persuasive Theory) Provide an assessment score on the user’s engagement in carbon footprint reduction based on a self-designed Environmental Engagement Index (EEI) including 7 indicators Based on the aforementioned index (EEI), EcoCal calculates a user’s average amount of carbon footprint consumption and sets it as the base case (see details of algorithm design in Part III Implementation) For each user, EcoCal stores this user’s data over time,
Recommend to user with daily nutritious diet plan which produces low carbon footprint based on our self-designed formula: User Utility Multidimensional Prediction UUMP Formula: util(u, f) = wp ∗ pref(u, f) + wc ∗ carbon(u, f) +wh * health(u,f) Explanation: We define a user’s utility with three components: personal preference (based on the frequency of a food), nutrition, carbon footprint production EcoCal defines the optimal plan as the one with highest user utility Prioritized the diet recommendation that is easy for user to achieve a smooth transition from previous diet habits Allow user to invite friends to participate in online group challenge and games in order to expand our customer base through the network effect

What we learned

We learned a manual procedure to produce a product, psychological environmental theories and persuasive theory to encourage user usage, and UI/UX design

What's next for Ecolerdar

train/adjustment How to lead the users to improve their score gradually

Built With

Share this project: