Read on Twitter
In normal times, connections between students on a college campus are an asset. During a pandemic, they are not.
My paper with Ben Cornwell on the structure of course enrollment networks @Cornell now out @SociologicalSci ( #OA): https://www.sociologicalscience.com/articles-v7-9-222/
1/n
My paper with Ben Cornwell on the structure of course enrollment networks @Cornell now out @SociologicalSci ( #OA): https://www.sociologicalscience.com/articles-v7-9-222/
1/n
I've tweeted about this paper before, but thought it'd be useful to update now that paper is peer-reviewed & published.
Teaser for regular readers: last tweets in today's thread reflect on findings in light of "reopening" conversation of last week+. 2/ https://twitter.com/WeedenKim/status/1258894522127396866
Teaser for regular readers: last tweets in today's thread reflect on findings in light of "reopening" conversation of last week+. 2/ https://twitter.com/WeedenKim/status/1258894522127396866
Key findings: the average Cornell student shares courses with a max of 529 other students over a week of classes. (Actual will be lower, because attendance is not 100%.)
This increases to 600 students if just look at undergrads, because grad courses typically smaller. 3/
This increases to 600 students if just look at undergrads, because grad courses typically smaller. 3/
In the university network including grads and undergrads, only 2% of all possible pairs of students take the same course, but 59% are connected indirectly through one other student (2 steps), and 92% are connected in 3 steps or fewer.
Move over, 6 degrees of Kevin Bacon. 4/
Move over, 6 degrees of Kevin Bacon. 4/
Connectivity is even greater in the undergrad network: 99.8% of student pairs are reachable in 3 steps, and virtually all student pairs are connected by more than one path. Student A may be connected to Student B through Student C, but also through Students D, E, and/or F.
5/
5/
Teaching courses of 100+ students online reduces the share of student pairs that can reach each other in 3 steps to 78%.
Teaching courses of 30+ online reduces share reached in 3 steps to 21%, increases average path length to 3.8.
A Rorschach result.
6/
Teaching courses of 30+ online reduces share reached in 3 steps to 21%, increases average path length to 3.8.
A Rorschach result.
6/
Networks in Cornell's liberal arts college (4.4K), UMich (30K; @ulisrael1 @TimMcKayUM), and a small SLAC (2K; Sean Tennant) have similar structure, although the average number of other students co-enrolled w/ focal student is, unsurprisingly, lower in @CornellCAS and SLAC.
7/
7/
Caveat: Course networks are just one of many sources of connections among students, faculty, and staff.
Also, Ben and I don't formally model spread, account for <100% attendance or physical distance within classrooms, assess other parameters that affect outbreak risk.
8/
Also, Ben and I don't formally model spread, account for <100% attendance or physical distance within classrooms, assess other parameters that affect outbreak risk.
8/
So. Colleges can reduce connectivity in the small world network of college campus through hybrid models of instruction (some courses on-line, others F2F), but not eliminate it.
Key questions are about relative risk under different scenarios, how else risk can be reduced.
9/
Key questions are about relative risk under different scenarios, how else risk can be reduced.
9/
Conversation about face-to-face instruction in fall needs to be upfront about the risks, recognize they aren't borne equally, debate appropriate tolerance. Hiding behind rhetoric of moral imperatives or deploying wartime analogies to justify "sacrifice" is ... disingenuous.
/fin
/fin
