Is It Time for End-of-College ‘Board’ Exams?
With AI use rising in higher ed, faculty members need new ways to ensure their students are actually learning
By Colleen Eren and Eric Torres
A recent Wall Street Journal headline proclaimed, “There’s a Good Chance Your Kid Uses AI to Cheat.” That article focused on the growing use of artificial intelligence to skirt the onerous tasks of writing, reading, critical thinking and analysis among middle and high schoolers. But epidemic levels of cheating among college students with generative AI are now an open secret among faculty. In our and our colleagues’ experience teaching at both an Ivy League university and a mid-tier regional public university, a sizeable proportion of student papers, reflections, homework assignments and exams are written wholesale by generative AI and then tweaked through more AI like StealthGPT, which unabashedly markets itself as being able to avoid detection as it “humanizes” language. Or the assignments are expertly rewritten, sometimes from drafts evidencing functional illiteracy, by AI tools such as Grammarly.
Doubtless there will be the usual slow march through peer review of studies demonstrating that AI is linked to faculty demoralization and showing just how pervasive is the inappropriate usage of AI in colleges. But now, when undergraduate degrees have already been dramatically devalued due to a crisis of confidence in higher education, cheating with AI may bring colleges to the brink.
It’s time for educational stakeholders to devise—quickly—methods of demonstrating students’ development of core knowledge and proficiencies. We offer one straightforward but controversial intervention that could preserve the meaning and value of a college degree, while providing benefits linked to transparency of outcomes: national, major-specific “board” exams and a generalized undergraduate “board” exam that would be required for any undergraduate degree.
The College Bubble
Accessible and often undetectable AI tools for avoiding many conventional forms of university-level coursework have come at a vulnerable time for higher education. For the past 75 years, colleges and universities have enjoyed growth in enrollment and social prestige that has allowed serious weaknesses in the function and culture of higher education to go uncorrected. To the degree that this growth has outpaced the value that ostensibly justifies it, a burstable bubble looms.
This bubble was precipitated by developments that made higher education more accessible (such as the 1944 G.I. Bill, the expansion of community college systems and government-subsidized aid and loans), but also by the logic of a degree as an imperative for social mobility and signal of meritocratic achievement. As a consequence of these developments, enrollment in college swelled in the U.S., going from about 1% of the total population in 1940 to peak at 6.8% in 2010, declining slightly to 5.8% in 2024. Those over the age of 25 with a bachelor’s degree or higher went from 4.6% in 1940 to 38.3% in 2023. The number of colleges and universities in the U.S. rose from 950 in 1920 to 3,500 in 2024.
While doubts about the value of college are not new, the emergence of AI comes in the midst of plummeting public trust in a system that has long enjoyed near-universal public confidence. According to a Gallup survey conducted last June, the proportion of Americans who report at least some trust in U.S. higher education sits at 68%, down an astonishing 22 percentage points from 2015. While the loss of trust is much more pronounced among Republicans, over half of whom now express little to no confidence in higher education, the general pattern of declining trust is consistent across political groups.
Part of this story is undoubtedly that college has become much more expensive, attenuating the wealth premium for degree holders and forcing a higher level of attention to the risks and tradeoffs involved in attending. But it isn’t just rising costs that are shaking public confidence. According to the same Gallup poll, the two most widespread, and interrelated, reasons for mistrust in higher education are the perceptions that colleges are pushing a political agenda and that students are not being properly educated.
These perceptions are not without grounding. In addition to a growing body of research suggesting a stark ideological imbalance on college campuses and self-censorship among students, public media coverage of higher education has regularly highlighted controversies about the seriousness of the educational experience. Revelations of the degree to which elite universities applied dramatically different standards for admission on the basis of student race or legacy status, findings of pervasive grade inflation, accounts of growing safetyism and reports from employers about the underpreparedness of graduates all fuel the perception that academic rigor has been supplanted by ideological socialization.
Screens and Smokescreens
The newest and perhaps most significant factor threatening the meaning of college education in an age of AI has been the rise of entirely online degrees and of conventional degree programs that allow students to fulfill a large portion of their coursework online, particularly in the postpandemic period. To give a sense of the scope of this phenomenon, between 2017 and 2024, online college education enrollment increased by 90%, while overall enrollment declined by 5%. There are now almost a million students in fully online programs. Many of these programs appear predatory, and not only among private, for-profit colleges that are readily identified as such online, such as University of Phoenix. Affixed to even reputable colleges and universities, these programs have been criticized, and in some cases legally challenged, for a lack of quality and rigor and poor student outcomes.
While cheating in academic pursuits is probably as old as the academy itself, entirely online programs targeting often vulnerable and busy student populations made it exceptionally easy, even before generative AI became widespread. Accounts of rampant cheating are now commonplace. Inside Higher Ed’s 2024 survey of faculty and students on cheating showed that almost half of students thought that cheating was easier than the previous year due to generative AI (probably higher than this, because the survey chose to separate out “use of ChatGPT” as a distinct reason for cheating apart from its generative AI counterparts), and a quarter thought it was easier due to online classes.
When the educational failures of these programs inevitably come to light, through either students reporting their experiences or employers discovering the uselessness of degrees as signals of certain capacities, they are likely to reflect on the degree-granting colleges and higher education as a whole, rather than simply discredit the online programs. This is because the diplomas are often the same for students who attend online-only, online-heavy and brick-and-mortar. Their resumes will be identical.
Especially for smaller schools and institutions whose budgets are threatened by dramatic reductions in federal research funding, international student tuition and declining enrollment (not only due to shifting choices away from higher ed, but also to an impending demographic cliff), mainstreaming of the perception that degrees are being earned through duplicitous use of AI may be an existential threat to most institutions of higher learning.
Bring in the Boards?
Degrees obtained after 2020 will very soon be viewed as suspect in their capacity to reflect a collectively intuited level of attainment, due to generative AI and other factors. This perception is also playing out during a period when the Trump administration is challenging higher ed’s accreditation process for allegedly being riddled with ideological content. In the face of these challenges, we propose a requirement by accrediting bodies that students completing associate and bachelor’s degrees take in-person, national, cumulative, general and major-specific “board” exams at the end of their college careers. These exams should cover a breadth of information and/or applied skills that college graduates are expected to possess.
While those who did not pass would still be able to obtain their degree, those who passed the exams would earn an additional “certification” credential in their degrees and majors. This certification would signal competency to potential employers, incentivizing students not to rely on AI to take their courses. These exams would be analogous to the postgraduation, professional certifying exams ensuring that students are actually qualified to practice in their field, such as the exams for doctors, nurses, lawyers, accountants and teachers.
The exercise of identifying the content for boards would be helpful in its own right. By engaging academic professional organizations with democratic governance structures and the ability for all affiliated academics to theoretically join, devising board exam content could stimulate a valuable conversation about what learning in the disciplines requires at a time of tremendous social, vocational and technological upheaval.
The exercise of academic organizations thinking through what curricular requirements in their discipline should look like is not a new endeavor. In 1951, the American Political Science Association published their report, rooted in surveys and interview data from instructors, on the teaching of political science in hundreds of colleges nationally. Among the questions were: “What form will [a political science] curriculum finally take, and what will be its spirit and methods?” “How can one deal with improvements in teaching without dealing with goals?” Its authors advised instructors to “guard against excessive proliferation of courses and a tendency toward disintegration,” and suggested courses that should be standard across colleges. A similar endeavor among other professional organizations—to form task forces and committees to ask what the goal of teaching in the discipline should be and what the pillars of the discipline are—is an important one.
Even in the absence of board exams, it is an important period for academics to think through what it would mean for students to attain a degree in their discipline. For sociologists, the American Sociology Association might create committees tasked with identifying what core precepts, theories and methodological skills a sociology major should possess. Should he or she be fluent in the writings of Émile Durkheim, W.E.B. DuBois and Max Weber? What types of methods should form an undergraduate level of understanding in the social sciences? What types of data literacy?
A general college board exam poses more complicated questions in terms of leadership and ultimate authority in adjudicating its final form, and quotidian issues like how frequently, and by whom, it is updated. We imagine that a consortium composed of the American Association of University Professors, disciplinary associations, accrediting bodies and university leadership associations like the Association of American Colleges and Universities and the American Council on Education would need to convene and adopt balanced, democratic frameworks for deliberation and decision-making.
This proposal may seem radical. Yet as Stephanie Davidson, the vice chancellor for academic affairs at the university system of Ohio, observed, “[T]here is a groundswell from the public about whether a college degree is worth what people are paying for it. ... [P]eople are asking for tangible demonstrations of what students know.” Correspondingly, there has been a nascent parent- and legislature-driven push in the United States toward exit exams, but it has not registered in any meaningful sense. Few colleges and universities are even attempting to test their students in writing and quantitative literacy prior to graduation, and then only for internal assessment purposes.
Several existing exams attempt to evaluate cumulative learning in college. The Collegiate Learning Assessment is used internally by some colleges to gain “information about their students’ performance on tasks that require them to think critically, reason analytically, solve realistic problems, and write clearly.” However, the goal of this test is “to provide a summative assessment of the value-added by the school’s instructional and other programs (taken as a whole) with respect to certain important learning outcomes,” and its scores have not been validated for making inferences about individual students’ abilities.
The Educational Testing Service’s Major Field Tests are also deployed by some departments to internally, quietly assess discipline-specific outcomes. These provide comparative data to departments about how their students fare versus all other senior-level students from other institutions. However, the ETS-authored exams rely exclusively on multiple-choice questions, and the data are not accessible to the public.
Requiring national board exams for accreditation would ensure transparency among colleges and departments. It would also force students, administrators and faculty to come to terms with the fact that the use of AI, grade inflation and loss of traditional metrics useful in predicting college performance in admissions are self- and institution-defeating practices.
Reservations
While instituting boards might help rekindle public confidence that degrees are an indicator of serious learning, a turn toward standardized exit evaluations would bring with it a variety of serious challenges that should not be taken lightly.
For one, there are philosophical arguments that standardized examinations are not a good fit for every discipline or even for undergraduate education as a whole. A major justification for the bar exams, medical boards and teacher licensure exams is that effective practice within each respective profession requires specific knowledge, the absence of which could result in serious harm to the clientele. But in the U.S., the relationship between one’s college major and postgraduate employment is often loose, and many of the specific expectations for professional knowledge tend to be handled through on-the-job training. Insofar as testing is driven largely by an interest in signaling professional preparedness, it risks misunderstanding the relationship between schooling and work.
Given the flexibility that many colleges offer students in constructing their course of study, identifying the right content to test students on would be a challenge. The imposition of narrowly focused tests could discourage intellectual omnivores, enshrine the values and priorities of the test-makers and effectively curtail the academic freedom of faculty by forcing them to teach to the tests. Domain-general tests, meanwhile, risk conflating general intellectual aptitude with evidence of specific learning throughout college.
High-stakes exams also create direct incentives to do well on the exam as opposed to doing well on what the exam aspires to measure. While much effort by psychometricians goes into developing tests where the impact of irrelevant factors on student scores is minimal, evidence of score inflation on high-stakes tests in the K-12 context demonstrates how incentives to perform well can distort the information tests provide. Students from wealthier backgrounds also have better resources to learn how to “game” the tests, thus making boards potential stratification-multipliers at the margins.
Since test results have an appealing simplicity, a serious risk of overreliance on scores at the expense of other important information about graduates could make these and other problems worse. To the degree that tests become a major stressor—casting a long shadow over students’ experiences on campus, absorbing time that might be spent socializing or engaging in extracurriculars, or otherwise stultifying the college experience—they could also impose cultural harms on higher education that sap the experience of some of its vitality.
One of the most serious problems with having to devise a disciplinary board exam is that academic professional societies and accrediting bodies would be responsible for setting standards and modalities. If the public already distrusts academic disciplines, the content viewed as exam-worthy may be, consciously or not, similarly distrusted. If the accrediting body is operating under the threat of an administration intent on expunging what is perceived as politicized content, that too will inspire public distrust.
There are, in theory, ways to mitigate many of these concerns. Exams as we’ve proposed them would be constructed by experts in the disciplines as opposed to potential employers, reducing the worry that these tests would hand control of college learning over to the vocational sector. Using a pass/fail model based on cut scores as opposed to offering stratified results could reduce the advantage that students with extra resources possess. A pass/fail model would also discourage a hyperfocus on scores at the expense of other meaningful elements of a graduate’s profile.
Thoughtful partnership with schools, departments and individual faculty in working toward implementation could mitigate risks of inhibiting faculty academic freedom or imposing unreasonable administrative burdens. Focusing on assessing baseline competence by selecting core knowledge and abilities that are spread throughout the curriculum and widely seen as essential in disciplinary thinking could further reduce worries of boards crowding out students’ opportunities to compose diverse courses of study or inhibit faculty’s ability to teach in their niches.
But even if undertaken with care, widespread institution of exit assessments would require the creation of an assessment leviathan whose long-run impacts are difficult to predict.
The academy may ultimately prove too divided and sclerotic to make exit exams a reality. And additional strategies may be necessary to confront the challenges that AI poses to teaching and learning. But division, distraction and intransigency cannot be excuses for a failure to confront the crisis of legitimacy facing academia. ChatGPT and its kin are already here, and signs suggest that they are earning degrees around the country. Absent serious, systemic efforts to assure that the opportunities of the university are not reduced to free training data for AI firms, higher education could crumble, and society at large may soon have to reckon with a future in which we have outsourced the gifts of learning to systems we scarcely understand.