Juan E. Gilbert, Ph.D. Associate Professor Auburn University Computer Science & Software Engineering Human Centered Computing Lab [email protected] http://www.juangilbert.com/ http://www.HumanCenteredComputing.org/
Mar 27, 2015
Juan E. Gilbert, Ph.D.Associate Professor
Auburn UniversityComputer Science & Software Engineering
Human Centered Computing [email protected]
http://www.juangilbert.com/http://www.HumanCenteredComputing.org/
US Supreme Court vs. UMich
Two land mark cases challenged the University of Michigan admissions policies Grutter v. Bollinger, 539 U.S. 306 (2003) Gratz v. Bollinger, 539 U.S. 244 (2003)
US Supreme Court vs. UMich
When the Law School denied admission to petitioner Grutter, a white Michigan resident with a 3.8 GPA and 161 LSAT score, she filed suit, alleging that respondents had discriminated against her on the basis of race.
US Supreme Court vs. UMich
She felt that she had been discriminated against because Michigan gave preference to minorities by weighting race.
US Supreme Court vs. UMich
This problem occurred because Michigan understands the benefits of
racial/ethnic diversity. The number of qualified applicants exceeded
the number of admissions slots.
Hence, some qualified applicants had to be turned away.
US Supreme Court vs. UMich
The Supreme Court issued clear, strong, unequivocal language endorsing Justice Powell’s opinion in Bakke, holding that the promotion of student body diversity is a compelling interest when necessary to achieve a school’s educational mission, and can justify the use of race as a “plus” factor in a competitive admissions process where all applicants are “on the same footing” for consideration.
Malcolm, S.M., Chubin, D.E. & Jesse, J.K., Standing Our Ground: A Guide for STEM Educators in the Post-Michigan Era, AAAS (2004)
US Supreme Court vs. UMich
A university has discretion, grounded in the First Amendment, in matters of academic judgment and does not have to exhaust every conceivable race-neutral alternative before considering race as one of many factors in order to satisfy the “narrowly tailored” definition, as long as the university engaged in a “serious, good faith consideration of workable race-neutral alternatives that will achieve the diversity.” A university is not required to actually adopt those “alternatives” if the university deems them inappropriate or unworkable.
Malcolm, S.M., Chubin, D.E. & Jesse, J.K., Standing Our Ground: A Guide for STEM Educators in the Post-Michigan Era, AAAS (2004)
US Supreme Court vs. UMich
A program implementing a flexible, holistic, individualized consideration of each applicant where race is only one of several relevant factors considered is likely to satisfy the “narrowly tailored” definition, whereas a program implementing a rigid, numerical value to each applicant based, even only in part, on race is less likely to be upheld and the automatic awarding of points based only on race is not permissible.
Malcolm, S.M., Chubin, D.E. & Jesse, J.K., Standing Our Ground: A Guide for STEM Educators in the Post-Michigan Era, AAAS (2004)
“I strongly support diversity of all kinds, including racial diversity in higher education. But the method used by the University of Michigan to achieve this important goal is fundamentally flawed. America is a diverse country, racially, economically, and ethnically. And our institutions of higher education should reflect our diversity.
US Supreme Court vs. UMich
Remarks by President George W. Bush on the Michigan Affirmative Action Case,The Roosevelt Room, January 15, 2003.
A college education should teach respect and understanding and goodwill. And these values are strengthened when students live and learn with people from many backgrounds. Yet quota systems that use race to include or exclude people from higher education and the opportunities it offers are divisive, unfair and impossible to square with the Constitution.”
US Supreme Court vs. UMich
Remarks by President George W. Bush on the Michigan Affirmative Action Case,The Roosevelt Room, January 15, 2003.
US Supreme Court vs. UMich
The Court decided that race could be considered in admission’s decision, but could not be the deciding factor.
Although this decision appears to support affirmative action efforts, it limits how race can be used to achieve diversity goals.
In sum, the Supreme Court ruled that diversity could be used in university-based admissions, but did not specify how diversity should be used.
US Supreme Court vs. UMich
As a result, several academic institutions have spent large sums of money to holistically evaluate admissions applications. University of Michigan, $1.8 Million Time intensive Point systems minus race/ethnicity
When university-based admission offices holistically evaluate applications. How does this translate into practice? What techniques could be employed to compare large
volumes of applications?
US Supreme Court vs. UMich
Ideally, a holistic evaluation would involve all the variables on an admissions application, such that no single variable (e.g. race/ethnicity) is the sole determinant of the admissions decision.
Rather, a collection of those variables are used to determine the admissions decision.
US Supreme Court vs. UMich
Imagine trying to holistically evaluate thousands of qualified applications?
Not possible, for humans.
Computer Aided Holistic Evaluation
What if every application was compared to every other application?
We would find similar applications in groups or clusters.
Summer 2003; The idea Fall 2003; Meeting with Drs. Walker & Wilson Spring 2004; Applications Quest Fall 2005; Open Source
What is Applications Quest?
Applications Quest
Applications Quest is a software tool that clusters admission’s applications based on holistic comparisons.
This software uses clustering algorithms to automatically compare thousands of applications to each other and place them into groups, based upon a holistic view of their similarity (i.e., similar applications appear within the same cluster).
The clusters represent diverse applicant pools.
Applications Quest
Applications Quest uses attribute-values on an application to determine similarity. GPA, GRE, GMAT, Essay, Race/Ethnicity,
City, State, Citizenship, Gender, Major, Degree
The more attributes in common, the more similar the applications.
SimilarApplications
How Similar?
The Applications Quest Approach
Applications Quest Approach
Numeric Attributes GPA, GRE, GMAT, TOEFL, etc.
Opinion Attributes Essays, Personal Statements, etc.
Nominal Attributes Race/Ethnicity, Gender, Major, Degree
Applications Quest Clustering
Applications are translated into an n-dimensional space, e.g. plotted as points on an x,y,z axis
Divisive Assume all points belong to one cluster At each step, select most different points Split the cluster around the two points Eventually each point is a cluster When do you stop?
When the specified number of clusters have been met.
Applications Quest Clustering
Applications Quest Clustering
Applications Quest Clustering
Applications Quest Clustering
Applications Quest Clustering
US Supreme Court vs. UMich Revisited
If Michigan had been using Applications Quest, what would have happened in the Gratz and Grutter cases?
Progress Report
Progress Report - Funding
NSF ITWF: Scholars of the Future: An Implementation Model for Increasing Diversity in Information Technology #0420485 Funds used to support Applications Quest
Auburn University Outreach Grant
Progress Report in the Press
Diverse (formerly Black Issues in Higher Education)
The Chronicle of Higher Education
Contra Cost Times – San Francisco
Berkeley Planet
Recruitment and Retention in Higher Education
University Business
Progress Report
Investigated licensing deal with 4 companies Admissions application processing Government hiring Ideal candidate application processing
Progress Report
Applied for patent on the application of this clustering algorithm.
Applications Quest prototype using AU Graduate School Applications.
Multi-University Pilot Study
Progress Report
Pilot Study Protocol Privacy Agreement Submit sample applications Identify admitted students Process the applications with Applications Quest Identify the Difference Index for Both Populations
Progress Report – Best Practices
Establish Minimum Admissions Requirements Test Scores, GPA, etc.
The Admission Bar
Progress Report – Best Practices
Two Part Admissions Process
First Part Typical applications attributes Test scores, GPA, Recommendation Letters, etc.
Do you meet “The Admissions Bar”? Do you have 2 bars? One for automatic admits and one for prospective admits
Progress Report – Best Practices
Second Part For those that met or exceed “The Admissions Bar” Collect Diversity information Race/Ethnicity, Gender, Political Affiliation, Memberships,
Legacy, Athlete, Views, etc.
Views Age, Religion, etc. The hard stuff that enriches the educational experience
Progress Report – Best Practices
Two Part Admissions Process Use Applications Quest to process the applications that
meet or exceed “The Admissions Bar”
Progress Report – Best Practices
University Wide Diversity Use the existing university students as part of the
admissions process On going collaboration with Dr. Richard Tapia at Rice
University (Naomi Reed, M.S. Student in Computational Mathematics)
Rice University - http://www.rice.edu/
Progress Report – Best Practices
Targeted Hires Submit an “ideal” application and interview members of the
“ideal” applicant’s cluster.
Progress Report – Best Practices
Usability Studies What’s the ideal N size?
Can Applications Quest find N? Studies are underway.
Progress Report – What’s Next?
February 27, 2006 Communications of the ACM Article
Ongoing pilot studies
Publish Best Practices Journal of College Admission
Press Release Fall 2007
Applications Quest Summary
Holistic Evaluations are Humanly Impossible.
Diversity is a worth while cause. And has been for a very long time.
Applications Quest Summary
Aristotle was open to the ideal of diversity, acknowledging that it could be useful for political discussion. He believed that conflict was inevitable and that multiple points of view served to make democracy stronger. For Aristotle, the state was better described as a plurality (made of many) than a unity (made of one); he understood the polity as requiring difference rather than homogeneity.
Frank, J. (2005). A democracy of distinction: Aristotle and the work of politics.Chicago: University of Chicago Press.
Applications Quest Summary
Mill’s Market Place of Ideas
[T]here are many truths of which the full meaning cannot be realized until personal experience has brought it home. But much more of the meaning even of these would have been understood, and what was understood would have been far more deeply impressed on the mind, if the man had been accustomed to hear it argued pro and con by people who did understand it. (p. 105)
Mill, J. S. (1974). On liberty. London: Penguin Books. (Original work published 1859)
Applications Quest Summary
People want transparent, fair and equitable admissions policies.
Applications Quest meets all of these requirements and it’s free! The 2 phase admissions process works!
Applications Quest Summary
What if every academic institution in the nation used Applications Quest? How would students apply to college?
Apply to increase their diversity/uniqueness So what would be the national outcome?
Uniform diversity? Something to think about
Thank You
Juan E. Gilbert, Ph.D.Associate Professor
Auburn UniversityComputer Science & Software Engineering
Human Centered Computing [email protected]
http://www.juangilbert.com/http://www.HumanCenteredComputing.org/