Swarthmore College was ranked the best liberal arts college in the United States by Academic Influence, a new college ranking method that uses artificial intelligence technology to search massive databases and measure the impact of work. people who have been affiliated with colleges and universities around the world.
Last Monday, Academic Influence released its first-ever ranking of liberal arts colleges in America – these four-year institutions that are relatively small in size, focus on bachelor’s level education, focus on direct engagement with faculty, offer an enriched residential experience and emphasize broad foundation in liberal arts with focused study in a major.
In short, here’s how the Academic Influence methodology works. This assumes that the people affiliated with a school determine its quality. To measure this quality for undergraduates, a branded metric called “Concentrated Influence” is calculated.
Using machine learning technology developed with funding from the Defense Advanced Research Projects Agency, Academic Influence searches open source data from two massive sources – Wikipedia and CrossRef – for articles, chapters, books, and citations to people around the world. Collectively, these databases contain billions of constantly updated data points on the accomplishments of millions of individuals.
Then, the concentrated influence score for a given institution is calculated by combining all the “mentions” of people who have been associated with that institution as professors, administrators or alumni. This score is then divided by the total number of undergraduates in the school. Adjusting the size in this way gives small and medium schools an equal chance to compete with large schools. A small school with a proportionately more influential faculty than a large school, whose absolute influence may be greater, will nevertheless achieve higher scores using the Concentrated Influence rankings.
In order to control for other confounding factors, the data sought is limited to the last ten years and the names of politicians, artists, performers, etc. famous are removed, solving the problem that they would otherwise have exerted a disproportionate influence on the school. ranking. (So, for example, Wellesley College won’t get a big bounce from its alumnus, Hillary Clinton.) Individuals’ influence is constantly tracked in real time, and influence scores are updated every day. quarters.
Academic influence began with around 1,000 liberal arts colleges in America meeting three criteria:
- All were non-profit institutions.
- They were all fully accredited.
- Total enrollment was at least 1,000 students.
He then ranked schools based on their concentrated influence scores. Here is his full list of the 50 best liberal arts colleges. The top 20 in order were:
- Swarthmore College
- Amherst College
- Sarah Lawrence College.
- Reed College
- Barnard College
- Wesleyan University
- Vassar College
- Hampshire College
- Wellesley College
- Williams college
- Union cooper
- Pomona College
- Oberlin College
- Bryn Mawr College
- Bard College
- Haverford College
- Smith College
- Claremont McKenna College
- Morehouse College
- Kenyon College
How does this list compare to other college rankings, like Best Liberal Arts Schools judged by American News and World Report‘s? Eleven of the top 20 Academic Influence Colleges were also among American Newsthe top 20 national liberal arts colleges; 15 were ranked in the top 40 by American News. So … considerable overlap, but clearly not redundant.
All ranking systems focus on some variables to the exclusion of others. For example, in the case of academic influence, one might object that concentrated influence scores cut too narrow a range to be a comprehensive measure of the overall quality of a college.
To complete the measure of influence, a “student desirability” score is also indicated for each school. Desirability measures the degree to which students attend one school in preference to another. Thus, for students admitted to two or more schools, desirability reflects the school they attended most often. Think of it as a measure of how well students “vote with their feet”. It acts as a proxy for a set of factors – for example, cost, location, facilities, equipment, specializations – that students take into account when deciding to attend a school relative to other competitors they are. were admitted.
In addition to its unique methodology, the core Academic Influence team is distinguished by the fact that it includes around half a dozen university professors and data scientists who began to come together in 2016. It is chaired by Jed Macosko, PhD, who is the chairman.
Macosko holds a BS from MIT and a PhD from the University of California, Berkeley. He is currently professor of physics at Wake Forest University. Other members of the team or its advisory board have graduate degrees in computer science, engineering, math, and history.
The group launched their website in August 2020, making Academic Influence the new kid on the college ranking block. In addition to liberal arts colleges, it provides rankings for universities, think tanks, and the world’s most influential individuals working in two dozen academic fields. Academic Influence’s modest operating budget is funded by an anonymous donor who wanted to create a more objective college grading system.
It also features interviews with influential academics and, starting next month, a series on influential university presidents. A recently interviewed Macosko University president, Dr. Santa Ono of the University of British Columbia (who is on two Academic Influence lists that include non-U.S. Schools) had the following ideas about college rankings:
“Although I have been critical of college rankings for a long time, I know they are influential and provide an important service to prospective students who need more information when searching for college. Current rankings suffer from over-reliance on self-reported or outdated information, subjective metrics such as reputation surveys, and input versus output data. Rankings place considerable importance on selectivity (input data) rather than the institution’s impact on students and faculty (output data). “
“I find the Academic Influence methodology innovative and refreshing as it uses a big data and machine learning approach to rank universities based on the influence / impact of its faculty and graduates. This moves away from a major flaw in existing rankings: the reliance on self-reported, outdated or reputable data and unbiased / freely available information. “
I interviewed Jed Macosko this week to find out more about the original rationale for this new approach. He said to me, “A lot of us were tired of hearing how one school ‘played’ the ranking system to dominate its competition. We thought that there should be a ranking system that no university could manipulate with incomplete or questionable reports. We knew that the enormous size of some databases does not allow us to treat them to its advantage. Wikipedia and Crossref do the trick, which is why we based our influence ranking on them. “
He went on to explain that he hoped that academic influence would provide students and families with something that other grading systems lack: “Students and their parents have grown up to expect the same schools to be in. top of almost every list. Ranking schools based on our desirability score includes these same names, because that’s where students choose to go. But our influence rankings, and Concentrated Influence ™ in particular, offer prospective students a new, more precise window into the quality of each school. We know this new window will help students make a more informed decision about where to apply – and especially where they ultimately go. Choosing a college is a life-changing decision, and we want to give people all the tools they need to make the right choice. “