Research Program
This research program spans three tiers, each designed to study meal sharing, social connection, and student wellbeing at a different level of resolution. Together they offer complementary lenses on the same question, and the data from each tier strengthens the others. The specific protocols and procedures for each tier are described in greater detail below.
Daily data on meal sharing and wellbeing, collected one short survey at a time using QR codes.
Right now, daily data on meal sharing and student wellbeing does not exist. Collecting it is the first and most important step, and the challenge is to do it in a way that is never intrusive, annoying, or burdensome. Consultations this semester with Food Literacy Project fellows, Harvard University Dining Services, and academic advisors all pointed to one clear solution.
QR codes on dining hall tables. When students sit down to eat in a dining hall, a small QR code on each table leads to a brief survey. All surveys are voluntary. Anyone can decide when and where they scan a code and take a survey.
All surveys will be voluntary. This is the key advantage of this approach, but also its biggest limitation. Taking them has to be something students actually want to do. The entire effort hinges on making these surveys quick, easy, and, most of all, fun. They have to be something any student, in any dining hall, at any time of day, whether dining alone or with others, will enjoy.
More than that, the best version of this design is one that is not only enjoyable, but also shared. Taking a survey is not usually a social experience. The ultimate goal is to make these surveys something that can lead to more social connections and, hopefully, more shared meals.
Every dining hall table can have a small QR code placard like the one displayed below. Scanning the code will lead to a short 30-second survey with three questions, ending in a wildcard. Most wildcards will be joyful and meaningless. Some wildcards will lead to prizes. Prizes can only be won in pairs, sparking brief moments of social connection.
Scan the code with your phone to walk through the experience as a student would.
An interactive flow chart of every screen in the participant journey from the first scan.
Because participation is voluntary, the survey has to be worth taking. This is the greatest challenge. The first and most important incentive will be that the survey itself should be genuinely fun to take. The second will be that some surveys can lead to prizes that can only be won in pairs. Finally, all students who participate will be eligible to win an all-expense paid trip to Italy, sponsored by Barilla. Each of these is described in greater detail below.
01
Make them fun
The entire survey experience is designed to feel like a small pleasure rather than a chore. Students can pick an avatar, see a funny video, respond dynamically to questions, and be surprised by a wildcard at the end.
Every survey ends in a wildcard, drawn from a library of roughly a hundred images with a named character and funny caption. These are designed to be goofy and surprising, sparking a brief moment of delight in the middle of the day. You never know which one you will get. That unpredictability is the point and gives an important reason to keep scanning.
02
Prizes can be won
Some wildcards will also lead to prizes. These wildcards will have an image of bread with a QR code it in. To win a prize, anyone who draws a wildcard with bread will have to ask the person next to them to scan the code. In other words...they have to ask someone to break bread.
Because prizes can only ever be won in pairs, a solo survey becomes a shared experience. It also gives students dining alone a reason to engage with someone nearby. The chance of winning a prize is the alibi for social interaction.
All prizes will be random. Students dining alone or with others will always have an equal chance of winning one. The specific prizes and odds of winning each one can be arranged on an ongoing basis over the summer and throughout the year. Some potential prizes are described below.
- Coffee for two at Tatte, Pavement, or Flour
- Gift cards to Felipe's, Russell House Tavern, or Little Donkey
- Movie tickets at AMC Boston Common or the Brattle
- Sports tickets to Celtics, Bruins, or Red Sox games
- A six-student chef dinner or pasta-making class with a Barilla chef
- A personalized short video from famous Harvard alumni
Because all prizes are random, winning them will spread by word of mouth throughout the year, generating additional interest in the study. No matter what, all students can always choose to skip. If they do not want to talk to anyone, they never have to.
To make prize chances fair, all students will be limited to three surveys per day, with 90 minutes in-between surveys. All of these details will be communicated to students in the consent form when they join the study, and in an about page that will be available on the dashboard.
03
A trip to Italy
All participanting students would be entered into a raffle to win a trip to Italy for them and two friends. Unlike the prizes hidden in bread, this one would not depend on luck. It would be open to anyone who joins the study and takes five surveys or more. Every committed participant has a real shot.
This prize is still being confirmed, but could potentially be sponsored and funded by Barilla. Barilla is an Italian company built on the belief that meals are meant to be shared. Italy is a country where meal sharing is of paramount importance. This would be the marquee prize and fits squarely within the goals and mission of the study itself.
The raffle winners would be announced at a closing celebration for all participants. At the end of the year, all participants could be invited to join a massive shared meal on campus before the summer holiday, ending the study on a high note.
Because prizes can only be won in pairs, breaking bread turns a survey into a shared experience. The second question of every survey asks students if they are eating by themselves or with others. For a student already eating with someone, breaking bread verifies the shared meal and labels the specific social connection. For students dining alone, the same feature becomes a randomized prompt to engage the person sitting next to them. The chance of winning a prize becomes the alibi for social interaction.
The study will focus on Harvard students, but anyone will be able to join the study. Students and non-students simply follow slightly different paths. Students enter a Harvard email, which verifies them as students and makes them eligible for the full set of prizes, including the year-end trip to Italy. Non-students can use any email and can still win some smaller prizes, but not the Italy raffle, and they are never asked to link any Harvard records.
Nobody is asked to commit to anything before they have tried it. On a first scan, every person is taken through the whole survey. They pick an avatar, answer the three questions, and draw a wildcard, all before being asked whether they would like to join. Anyone who declines has their email and first responses discarded and is welcome to join on any later scan.
Joining the study means giving informed consent. The consent form will cover prize eligibility, data privacy, and the goals of the research. For students, this is also where they consent to link their survey responses with their Harvard records. Non-students see an abridged form that covers prizes and privacy but never touches Harvard data. The full consent form will also be available on the home dashboard for all participants throughout the study period.
Privacy will be protected by design. Once someone consents to join the study, their responses will be stored under a coded identifier rather than their name. Student emails will be verified and used to link with administrative data, but any identifying information will be kept strictly separate from the research team. This is described in detail in the Data Privacy section below.
Any participant can withdraw at any time. When someone withdraws, their data is held until the end of the study. If they rejoin during the study period, everything picks up where they left off. If they do not, their data is deleted when the study ends.
Randomly vary dining policies across Harvard houses to test effects on shared meals.
To know whether a dining policy actually promotes meal sharing, we need to test it. Harvard dining halls are the ideal setting. Each dining hall already runs its own programming. Policies rotate across the year and vary between houses. By randomly assigning policies across houses this year, we can test causal effects on meal sharing and downstream effects on student wellbeing and social connection.
Learning objectives
- Which dining policies are most likely to promote more shared meals?
- Do these policy changes improve student wellbeing and social connection?
Intervention design
Several dining policies already exist that may affect if and how often students share meals. Many of these policies are currently in rotation across dining halls. Selecting a subset to randomly vary across different houses can provide the necessary means to detect causal effects of dining policies on meal sharing, dining alone, and student wellbeing. A list of potential policies is provided below.
01
Community Night expansion
Most undergraduate houses run a programmed Community Night per week, almost always on Thursday. The hall closes to interhouse guests during this window so that residents eat together.
Randomize. A random subset of houses could add a second weekly Community Night, while other houses remain on the standard Thursday schedule.
Mechanism. A scheduled Community Night is added where one didn't exist, by construction. The treatment is identical in format to programming HUDS already runs every Thursday, this would just be an extension.
02
Interhouse guest rules
Each house has its own rules about who can eat where, varying by meal period. The rules range from very restrictive (Eliot closes lunch to interhouse from 11:30-1pm, then allows +1 with resident only 1-2pm) to nearly unrestricted (Cabot, Pforzheimer, and Mather have minimal interhouse limits).
Randomize. A random subset of houses could open a previously restricted meal period to interhouse guests for a defined window.
Mechanism. Lower friction means friends in different houses could eat together at meal periods when they previously couldn't.
03
Brain Break structure
Undergraduate houses currently run Brain Break (late night snacks) Sunday through Thursday starting at 9pm.
Randomize. A random subset of houses could formalize Brain Breaks as a seated social gathering with light programming, while other houses implement a grab-and-go format.
Mechanism. A late night moment that currently encourages solo or grab-and-go eating is converted into a social ritual at treatment houses.
04
Lunch window extension
Residential dining halls close at 2pm for lunch, while class blocks routinely run up to or past that time. Students with late schedules currently rush, skip lunch, grab Fly-By, or eat off-campus.
Randomize. A random subset of dining halls could extend lunch close to 2:30pm (matching the existing Fly-By close time) for a defined period. This would be welcomed by students according to FLP fellows, but would require quite an organizational and administrative effort.
Mechanism. The extension increases dining hall attendance during the late-lunch window, but whether those additional meals are more likely to be shared depends on whether meals eaten in dining halls are more likely to be shared in general than meals eaten outside of dining halls. In conversations with FLP fellows, this appears to be genuinely unclear. This is an important question to introduce in college surveys.
05
Fly-By variation
FLP fellows indicated that students who use Fly-By typically do so for scheduling reasons (between classes, on the way somewhere).
Randomize. Limit hours of Fly-By operations at Memorial Hall or the SEC on randomly selected days or weeks. The former may be expected to lead to more shared meals, although it is worth noting that limiting Fly-By hours would likely receive pushback from students.
Mechanism. Without Fly-By-Meal options, students may be more likely to stay and eat meals with others in dining halls. This depends on the assumption that Fly-By-Meals are less likely to be shared than meals eaten in dining halls. This is likely true, but again may be worth asking directly as a survey item in college surveys.
Measuring outcomes
Both the frequency of meal sharing and student wellbeing can be measured directly. The QR code surveys from Tier 1 capture both outcomes at high frequency, one meal at a time, across the year. A few additional items introduced into existing Residential Life surveys capture the same outcomes for the wider student body, including students who scan less often. Together the two instruments give a clear read on whether a given policy actually shifted how much students share meals and how they feel as a result.
Tradeoffs and limitations
The main limitation is the small number of houses. Although policy changes at the house level are likely the easiest to implement, the relatively small number of houses will make specific causal effects difficult to estimate. Randomizing house policies does not let us assign individual students to share more meals. We can only change the environments around them. Spillovers between houses are also a real risk, particularly for policies that affect inter-house dining. Nevertheless, this can provide a real opportunity to test whether and which dining policies promote more meal sharing.
Years of meal swipe data, linked to administrative records and analyzed at scale.
Harvard already collects detailed meal swipe data for nearly every undergraduate student. Because almost all undergraduates have meal plans and live on campus, this record is uniquely representative of the student population. It also extends back in time, spanning numerous significant changes to dining policy at Harvard.
On its own, meal swipe data tells us where and when students eat, but not whether they shared a meal. This is its fundamental limitation. As a first step, co-swipes within short time windows may be able to serve as a proxy for shared meals. Dining hall attendance itself can also serve as a rough approximation of meal sharing if students are more likely to eat with others in a dining hall than elsewhere. This can be tested directly through Residential Life surveys. Nevertheless, while potentially informative, meal swipes alone are still likely to provide only a rough approximation of meal sharing behavior.
Making meal swipes meaningful
The real power will come from merging meal swipe data with QR code survey responses and administrative data. These linkages connect each swipe to information about meal sharing behavior, student wellbeing, course schedules, grades, housing assignments, entryways, blocking groups, and mental health indicators. Each of these linkages requires a Data Use Agreement with the relevant office, described in detail in the next section. Taken together, this data can offer exceptionally valuable insights into the dynamics of meal sharing and its effects on wellbeing and social connection over time.
Survey responses can be used to infer meal sharing in historical data. QR code responses linked to meal swipes and administrative data will provide ground truth information about whether a given meal was shared or eaten alone, alongside observable features like time of day, dining hall, house, residential context, and course schedule. A prediction model trained on these labels can then project the likelihood of meal sharing back across years of historical swipe data, making a record that was never designed to capture social dining newly interpretable.
Going forward, the same QR code surveys turn meal swipes into labeled network data. Overlapping swipes between students who confirmed in a QR code survey that they ate together can be used to infer shared meals across the wider student population. This will be much more reliable than meal swipe data alone. For solo diners who received the bread wildcard, the record lets us estimate whether brief engineered encounters lead to lasting social ties by tracking student blocking groups and social connections over time.
Together, this will provide one of the largest and most informative datasets of social dining ever assembled. Using data collected across all three tiers, we can study how social dining at Harvard has changed over time, use historical policy changes as natural experiments to estimate the effects of past dining policies, and further pin down the effects of present and future policies to promote meal sharing, student wellbeing, and social connection.