COVID-19 on Campus: How Should Schools Be Redesigned?
Redesigning academic institutions so that they can re-open and function safely in the aftermath of the COVID-19 pandemic is a challenging problem. Part of the value from a traditional education in an elite institution is socialization as a member of the future elites in finance, business more broadly, law, government, and industry. You meet and become friendly with people who will work with you for decades to come, whose trust you likewise will need for decades. So how should these institutions be redesigned in ways that preserve the advantages of on-campus learning while also incorporating online instruction?
Wharton’s Eric K. Clemons, a professor of operations, information and decisions, focuses on the redesign of academic institutions in this opinion piece. As he writes, “We do this in part because it is a pressing problem that requires a solution now, if universities, faculty, and students are to plan effectively for an event that is less than three months away. In addition, we study these institutions as examples that other industries may need to consider as they approach their own redesign if they cannot implement full virtualization.” (Clemons acknowledges the contributions of his colleagues Talal Shamoon and Kezia Wright of Intertrust.)
Every institution that provides goods and services to the public needs to consider carefully how it will redesign itself to function in the environment created by COVID-19. Long-term planning will need to wait until we have greater certainty about the environment we will face. (See my previous articles in Knowledge@Wharton on alternative scenarios created by the virus and alternative scenarios created by public response to the virus.) But we are beginning phased reopening now, and organizations need a design for providing goods and services to the public now. We can already see that improper design, even suboptimal design, can be catastrophic.
Optimal design involves complex tradeoffs. For some industries there is no successful redesign and for others the redesign is relatively straightforward. We will focus on elite academic institutions like Harvard College, The Wharton School, Peking University Law School, and Ecole National d’Administration in Strasbourg, because some continuation of their face-to-face operation is essential to their entire value proposition, and because redesign to safely achieve this is most complex.
Optimal design involves trading off two competing objectives. The first is to maximize safety and public health. If that were our only objective, we would continue the lockdown and prevent all indoor public spaces from reopening until a vaccine were available and had been proven effective even in the face of continuing mutation of the COVID-19 Virus. That is impossible. Likewise, if we were to choose to maximize speed of return to the world before the virus, we would drop all restrictions immediately and allow businesses to open immediately without redesign. That is also clearly impossible. Even if public officials were to embrace this objective, most of the public would not, and those that did would contribute to a resurgence of the virus. We need to balance protecting public health and the speed of reopening the economy.
We need to determine to what extent we can deliver services adequately through virtual online alternatives. We need to determine to what extent we can deliver traditional services effectively through safe redesign, which manages public risks and to convince the public that risks have indeed been managed. And we need to find the optimal balance.
In some industries there is no redesign that saves it from total economic collapse. Movie theaters represent the most extreme example. Services can effectively be delivered virtually through streaming video, and the industry is already moving in that direction. Netflix and Home Box Office are thriving under lockdown, and even before the virus they were already developing their own proprietary content. And most theaters cannot survive with 30% occupancy. Seating in every other seat, and in every other row, creates effective social distancing. Theaters with adequate spacing already exist, like the VIP Theaters in Toronto. However, these seats are enormously more expensive, and profits come as much from selling craft beer and good wine and dinners as they come from selling tickets. This cannot be scaled up. The vast majority of movie theaters will never reopen.
In other industries, effective redesign is more complex but already appears successful. Retailing will be transformed but that industry already is adapting. Amazon delivers most shelf-stable food items and dominates online delivery of most other products. Specialty retailers are achieving a balance of in-store shopping, curb-side pickup, and next day delivery. I get most of my shelf-stable items liked canned goods and jarred goods delivered by Amazon. I get my imported cheeses and dry aged prime meats delivered by DiBruno’s.
Redesigning elite academic institutions provides a concrete example of the most difficult design problem. Part of the value from a traditional education in an elite institution is socialization as a member of the future elites in finance, business more broadly, law, government, and industry. You meet and become friendly with people who will work with you for decades to come, whose trust you likewise will need for decades. And you will meet more peers at other elite institutions through events like inter-school rugby tournaments. Military academies understand that their missions are complex, and they must include discipline and trust in your future cohort of peer officers as much as they include tactics, strategy, and military history.
In the rest of this article, we focus on redesign of academic institutions. We do this in part because it is a pressing problem that requires a solution now, if universities, faculty, and students are to plan effectively for an event that is less than three months away. In addition, we study these institutions as examples that other industries may need to consider as they approach their own redesign if they cannot implement full virtualization.
Implications for Reopening Campuses in Fall 2020
As noted, a significant portion of the value of attending an elite institution is the informal time spent with faculty and with other students, and much of this value can be destroyed by improper redesign of the campus before reopening. Design needs to be guided by estimates of the state of the virus and the state of the economy in August and September.
Although I have done a more comprehensive study of the alternative futures that will face academic institutions over the coming years, we can accurately predict the most important elements of the environment that all organizations will face in September. We can assume that virus will still be with us, but under some degree of control due to social distancing. We cannot assume that the virus will have been eliminated, because we are already experiencing the impact of imperfect designs employed as many states rush to reopen. We can also assume that state, local, and federal governments will still be dealing with the massive debt created by the virus over the past several months. Further, we can assume that many families are still dealing with the financial hardship created by the lockdown.
I started considering alternative strategies for reopening by constructing a simulation model and running it under a wide range of conditions. Let us not attempt to predict a single outcome with certainty; let us do sensitivity analysis under a wide range of assumptions and examine the implications for prudent risk management. The actual simulation was constructed rapidly by students working under my direction as a term project and we cannot trust the specific values predicted, like the speed of the spread of the virus, or the percentage of instruction that would need to be virtual in order make opening a campus safe. But the results do suggest alternative strategies.
Let us assume that all returning students and all returning faculty and staff are virus-free at the start of the semester, which will allow us to create the most optimistic assessment of the risks of reopening. However, we cannot assume that the campus remains virus-free indefinitely. The Hyderabad campus of the Indian School of Business was designed as a clean bubble, a place that is sealed, where it is safe for a Western visitor to drink the water and to order lettuce, tomato, and cucumber on a Subway sandwich. No American university was designed as a bubble, and we can assume with certainty that some individuals will contract the virus off-campus, visiting, dining out, or traveling. We need to assess whether the inevitable undetected arrival of disease carriers on campus can be managed, or whether it will lead to unmanageable spread of the disease across an entire campus.
Structure of the Model
We start with a Healthy population in three age groups, corresponding to Students, most Faculty and Staff, and the Oldest faculty and staff. We assume that individuals on a campus meet principally as a result of three activities: class, dining, and dormitory. We assume that essentially all non-academic activities, from sports to drama and music, will have been canceled. We do not base our model on traditional estimates of R0, the average number of people infected by each carrier of the disease. Rather, we model the probability that an encounter with an infected person will lead to a new infection (NewInfectionProb), the number of people each individual encounters in a day (NumberOfContacts), and the probability that any individual encounter will be with a person who is already a carrier (InfectedProb). InfectedProb is computed by determining the percentage of individuals on campus who are carriers of the virus. R0, the average number of people infected as the virus progresses through a campus, can be derived from these three terms. We model the transition of individuals from the Healthy population, into Exposed, Infectious Asymptomatic, Infectious Sick, Recovered Immune, or Deceased.
We explore the impact of various interventions. First is full PPE or other protection for faculty and staff. Second is mandatory social distancing for students in the classroom. Third is mandatory social distancing for students in dining and student housing. Fourth is social contact tracing without testing. Fifth is social contact tracing with frequent and regular testing. These interventions can be combined to yield various strategies for reopening a campus.
Strategy 1: Do Nothing and Strive for Herd Immunity
A university campus is not and cannot be perfectly sealed and perfectly isolated. Students will encounter infected individuals while shopping, dining, traveling, or socializing. Not surprisingly, we note that a disease-free campus is inherently unstable, because the daily number of contacts each individual has is so high. With no intervention to limit the spread, once the virus has been introduced infection would soon overwhelm the entire campus, as it would do without intervention in other places where people are tightly confined, like a prison. Fatalities among elderly faculty or staff would certainly occur. Deaths among students with pre-existing medical conditions is highly likely as well. With the speed of infection of the entire campus, education would certainly be disrupted, and the university might be legally liable for its inability to provide a safe operating environment. Reopen and change nothing is not a viable strategy.
Strategy 2: Protect the Vulnerable and Strive for Herd Immunity Among Students
The first active intervention would be to protect the faculty and staff while making no other active interventions. This would reduce or eliminate death among these groups. It would eliminate only one source of contagion, that due to exposure from infected faculty. It would not slow the spread of the disease sufficiently. We could not be certain that the disease would not become rampant within a campus, and we could not ensure that normal university operations could continue. Additionally, providing a safe environment for faculty without providing similar protection for students would raise serious ethical and legal concerns. This analysis is equally valid for any organization that has senior management and line employees, like a meat processing plant or a manufacturing line with staff who work with little physical separation. Protecting only the senior staff and hoping for herd immunity among the rest of the organization is not a viable strategy.
Strategy 3: Protect the Vulnerable and Slow the Spread of the Virus through Social Distancing Combined with Virtual Instruction
This strategy combines two interventions, protection for faculty and staff and enforced social distancing in the classroom and in the corridors. Perhaps one-third of students would attend class on any given day, while two-thirds would attend virtually. The actual ratio would be determined by modeling for each campus, for each building, or for each classroom where necessary. This would ensure that classes could be held without changing the rooms that would have been assigned prior to the virus. This would still ensure that existing rooms would be large enough to permit social distancing. Students would sit in the same assigned seats each time they attended class, to limit their total number of contacts over time and to facilitate testing and tracing. Faculty would either wear PPE or would teach behind a clear protective barrier. Hallway congestion would be the greatest remaining source of disease transmission from teaching-related activities. All students and faculty would be required to wear N95 masks in the hallways. Once again, protecting faculty and staff without protecting students is fraught with both ethical and legal issues; any student would have the right to declare himself or herself medically vulnerable, and any self-declared vulnerable student would be entitled to remain off campus and to participate through distance learning.
This slows the spread of the virus, but perhaps not sufficiently to ensure that education can proceed without disruption. Careful monitoring is required at all times. This strategy could also be employed in industries where some work can continue to be performed by traditional labor, while other work is automated, reducing the physical proximity of employees. Unfortunately, there are very few industries where mixing virtualization and traditional labor is effective. It is easier to automate an entire production facility than it is to reengineer it to increase spacing between traditional employees. It is difficult to double the spacing between employees in a meat processing facility without doubling the size of the building or reducing output by half.
The combination of protecting senior staff and blending virtualization with traditional service delivery is possible in some industries but not others. Moreover, it is not sufficient in settings where employees or students also live together in close quarters after work, like a university or an aircraft carrier.
Strategy 4: Protect the Vulnerable and Slow the Spread of the Virus through Virtual Instruction and Redesigned Dormitory Experience
This intervention would continue full protection for faculty and staff and a mixture of in-class and virtual instruction and would change the dormitory experience to limit student exposure in living and dining spaces. Mostly dormitories have been designed to increase social interaction, with students sharing rooms, rooms divided among suites, and suites sharing lounges. It is probably not economically feasible to reduce occupancy levels for dormitories. But dormitories would need to be subdivided into discrete units, reducing the number of contacts between units. Dining schedules would need be designed around these same discrete units, again to reduce the numbers of contacts each student would have. The changes would also greatly facilitate contact tracing when an individual student did contract the virus. Additionally, shared facilities would require high levels of cleaning and disinfection.
These three interventions would greatly slow the explosive exponential spread of the virus on a campus, but the spread would still remain exponential. If the growth is sufficiently slowed, then illness will not interfere with the university’s mission. Herd immunity will emerge, either naturally as the virus slowly works its way through a campus, or more rapidly, with the introduction of a vaccine. And of course, these interventions are not appropriate in all settings where students or employees live together after work. It is impossible to redesign an aircraft carrier, or worse yet a submarine, to achieve perfectly safe living accommodations.
The combination of protecting senior staff and blending virtualization with traditional service delivery is possible in some industries but not others. Moreover, while redesign of living accommodations will be essential where students or employees live together after work, this will not always be possible.
Strategy 5: Protect the Vulnerable, Slow the Spread of the Virus through Social Distancing in the Classroom and Dorms, and Remove the Contagious through Test and Trace
The only way to eliminate exponential growth would be the nearly instantaneous removal of infected individuals from the campus and placing them in quarantine. Quarantine would not need to be onerous; students could be housed in the hotel-like facilities that many universities maintain for executive education or visiting faculty, often with its own computers in every room and its dedicated dining and exercise facilities.
Unfortunately for the purposes of disease control, many individuals in the student pool are going to be asymptomatic even when infected and contagious. Testing only when symptomatic individuals have been detected may give each infected individual many days to infect dozens of other students, many of whom will likewise be asymptomatic. Even in the presence of all forms of intervention described above, testing will need to be constant and universal, with effective contact tracing after the identification of infected individuals. Daily testing across campus will not be possible. And when a significant portion of infected individuals are asymptomatic, hot spots are likely to have grown quite large before they are detected.
One possible testing strategy has been suggested that may make frequent testing economically feasible, even if the inconvenience factor would remain quite high. The idea would be to take samples from entire living groups of maybe 100 or 200 students. The expectation is that almost all tests would be negative; a single test could then “clear” as many as 200 students. If there was even a single infection in a block of 200 students, then ten blocks of 20 students could be tested, with ever smaller test groups until the infected individuals were identified.
The optimal strategy for reopening is a combination of protecting senior staff and blending virtualization with traditional service delivery where possible. When employees live together after work this will also require redesign of living accommodations. However, this will never be entirely effective except in sealed environments like naval vessels at sea, so aggressive programs of testing and contact tracing will also be required.
Recommendations
For the remainder of this document, I will focus on academic institutions, since we are facing critical time pressure as we decide how to reopen in less than three months. Our objective, at least for elite institutions, is to preserve the quality of instruction and equally to preserve the quality of face-to-face socialization and interaction.
I would not be comfortable making a recommendation for any single campus, even my own, before completing additional refinements to the model and calibration of its results. But it is possible to offer some tentative guidance. If a university does choose to open in the Fall, it will need to adopt the following practices, all of which are contained in Strategy 5.
It will be necessary to provide protection to faculty, who because of their age are more vulnerable and more likely to experience severe health consequences.
It will be possible to operate the university with a mixture of in-class and virtual instruction, provided that social distancing can be maintained in dormitories as well. The degree of social distancing will need to be carefully calibrated to ensure that the level of illness on campus does not become so high that it makes instruction impossible.
For legal and ethical reasons students must be permitted to self-identify as vulnerable. Those students will remain off campus and will participate in only virtual instruction.
Frequent testing and rapid social tracing will be essential to slow outbreaks of the virus, which inevitably will occur.
For some schools within a university, like a medical school, a veterinary school, or a dental school, physical operations with social distancing will be required. This is likewise going to be necessary for most graduate students in most experimental disciplines that require laboratory work.
Implementation
Implementation strategies will be slightly different for every campus, but for instruction currently based on lecture formats, like undergraduate programs and MBA programs, the following will be necessary:
Optimal redesign of classrooms and of passage between classes;
Optimal redesign of dormitories and dining experiences;
Provision of safe dining options for students who eat in class or between classes;
A clear explanation for students of what their options truly are — help make the campus safe or revert to lockdown. That means that compliance is in best interests of all students;
And design of the least intrusive enforcement mechanism.
All college faculty and administrators were once undergraduates. We all remember drinking before we reached legal drinking age. We all remember at least some use of recreational drugs, at least among some of our friends. We all remember ignoring rules about being in a boyfriend’s dorm room after hours or sleeping in a girlfriend’s dorm room when her roommate was conveniently out. We did not follow rules. We especially did not follow rules when we thought we would not be caught, or we thought the penalty for violating the rules was not really severe.
Students are not going to see the risk of COVID-19 as severe for them. They will not be certain that violating a social distancing norm will expose them, or that exposure will lead to illness, or that illness would be severe for them. They are not going to give up seeing friends or lovers. Redesign of living on-campus units will be the most ineffective part of the transformation, and campus living will be the most vulnerable part of the redesigned campus. And students will — quite reasonably — resist tracking through their phones. Where they sleep, and who they sleep with, and how late they are up drinking, and who they drink with, are all data they will not wish to share with their universities or with anyone else.
So how do universities gain any cooperation from students at all? I suggest the following:
We start with a clear explanation of the university’s model, its assumptions, and the resulting design criteria. Students need to know what their school is doing and why.
We provide a clear explanation of the results of the optimization model, the implications for on-campus transmission and illness, and the implications for keeping campus open vs. having to deal with campus closure and lockdown again. The students need to trust the university’s design, and they must understand that we are trying to provide the most intimate and safest environment possible, with minimum possible intrusion.
Enforcement must not be based on students’ phones! Every student would have a unique RFD token. Without the token they would not be admitted to any classroom building. That would ensure that students only attended class on days that they were permitted to attend. Cellphone data will not be collected.
Students will be encouraged to remain within their dormitory dining cohort and their dormitory living cohort. We understand that this cannot and probably should not be rigorously enforced. There will be times when students meet friends and dates. There will be leakage between cohorts. Our models and our response to entry of the virus must account for leakage. Super-spreader events should be banned; there probably will not be giant on-campus parties in the foreseeable future.
Contact tracing will be essential and contact tracing must be automatic, secure, and private. I think we need an app that lets students record their contacts with other students automatically and that keeps that information secure. Information should capture who was met but not when or where or for how long, other than to be able to identify the day. When students are contacted for contact tracing, they will be expected to share this information voluntarily, but they cannot be compelled to do so and even this limited data cannot be accessed by anyone without a warrant legally authorizing access to their phones.
Testing must be frequent. Students who have contracted the virus, whether they are symptomatic or not, must be quarantined, and quarantine must be without stigma. Ideally, quarantine of students who reside in dormitories would be in facilities nicer than existing dormitory facilities. Students who live off campus will be expected to quarantine at home. Students who are quarantined will have their RFD tags noted and will automatically be denied access to classroom buildings and dormitories.
Contact tracing must be immediate, compassionate, and anonymous. Students who have been exposed must be notified. They must be told as gently as possible. And they must not be told the name of the individual who may have exposed them.
The entire process must be transparent. Students must know the reason for each measure that has been imposed. And they must be able to know the state of the disease on campus at any time, including within their individual school and within their discrete living and dining cohorts.