Blab AI
Harvard College Dean Deming Signals Shift Toward Guided AI Use and New Assessment Models
Harvard College has unveiled a fresh strategy for AI in its classrooms, spearheaded by Dean David J. Deming. In a Wednesday morning email to students, he detailed a plan that promotes AI where it can enrich learning, yet imposes stricter limits on AI‑dependent assignments. He also urged a shift from take‑home exams to proctored assessments.
The email, dispatched on the first day of the fall term, outlined Deming’s goal of crafting advisory guidelines for AI across Harvard courses. He emphasized that the college would prioritize assignments that cannot be handled by AI and argued that take‑home exams and problem sets are increasingly unsuited for gauging mastery. He likened the challenge of policing AI use on these tasks to a mythic temptation students must overcome.
Within the same note, Deming proposed a contrasting approach for writing‑heavy courses. He wrote that papers and other substantial assignments should welcome AI—or even encourage its use—rather than be subjected to intensified detection. He warned that an overemphasis on AI detection erodes trust between faculty and students.
A short time later, Deming published a critique of AI detection tools on his Substack, labeling them as part of a race that merely speeds the demand for swifter detection techniques.
Deming clarified that the email’s ideas are exploratory, not policy. He noted that developing concrete guidelines would demand broad consultation with faculty, staff, and students, and that the AI landscape is evolving rapidly, leaving no one‑size‑fits‑all solution.
Earlier this summer, Harvard’s Bok Center for Teaching and Learning issued guidance on how faculty could adapt courses to the growing use of AI by students. The center’s recommendations echo many of Deming’s points, particularly the emphasis on in‑person assessments.
Since taking the dean’s office last year, Deming has made AI a priority, consistently eschewing a blanket policy in favor of course‑specific approaches. In an April open forum, he likened the introductory CS course CS50 to a small English seminar, noting that identical guardrails would be ill‑suited for both.
At the university’s Convocation ceremony, Deming underscored the need for a new era in higher education—one that prioritizes building learning communities over merely imparting expertise. He linked this shift to the broader trend of technology democratizing knowledge and the waning trust in experts.
In the same email, Deming reaffirmed community as Harvard’s guiding purpose, announcing several initiatives designed to boost student‑faculty interaction, including dining‑hall meal swipes for staff and faculty and additional social events scheduled for the fall.
The dean’s remarks arrive amid a nationwide scramble for universities to integrate AI tools into teaching and assessment. According to Deming, Harvard’s strategy is to act rather than merely observe, while still acknowledging the necessity of careful, course‑specific guidelines.
Harvard’s next steps will involve crafting detailed guidelines in consultation with stakeholders. The resulting framework is expected to shape how the college’s courses balance AI use, assessment integrity, and community building in the coming academic year.
This policy shift echoes a wider debate about AI in education, where institutions must weigh the learning benefits of AI against the risks of cheating and eroding trust. Harvard’s plan to encourage AI in some contexts while tightening controls in others reflects a nuanced approach to these challenges.
Educators and policymakers will closely monitor the university’s new initiatives and forthcoming guidelines as they seek to integrate AI tools into higher education while preserving academic standards and fostering collaborative learning environments.
The email, dispatched on the first day of the fall term, outlined Deming’s goal of crafting advisory guidelines for AI across Harvard courses. He emphasized that the college would prioritize assignments that cannot be handled by AI and argued that take‑home exams and problem sets are increasingly unsuited for gauging mastery. He likened the challenge of policing AI use on these tasks to a mythic temptation students must overcome.
Within the same note, Deming proposed a contrasting approach for writing‑heavy courses. He wrote that papers and other substantial assignments should welcome AI—or even encourage its use—rather than be subjected to intensified detection. He warned that an overemphasis on AI detection erodes trust between faculty and students.
A short time later, Deming published a critique of AI detection tools on his Substack, labeling them as part of a race that merely speeds the demand for swifter detection techniques.
Deming clarified that the email’s ideas are exploratory, not policy. He noted that developing concrete guidelines would demand broad consultation with faculty, staff, and students, and that the AI landscape is evolving rapidly, leaving no one‑size‑fits‑all solution.
Earlier this summer, Harvard’s Bok Center for Teaching and Learning issued guidance on how faculty could adapt courses to the growing use of AI by students. The center’s recommendations echo many of Deming’s points, particularly the emphasis on in‑person assessments.
Since taking the dean’s office last year, Deming has made AI a priority, consistently eschewing a blanket policy in favor of course‑specific approaches. In an April open forum, he likened the introductory CS course CS50 to a small English seminar, noting that identical guardrails would be ill‑suited for both.
At the university’s Convocation ceremony, Deming underscored the need for a new era in higher education—one that prioritizes building learning communities over merely imparting expertise. He linked this shift to the broader trend of technology democratizing knowledge and the waning trust in experts.
In the same email, Deming reaffirmed community as Harvard’s guiding purpose, announcing several initiatives designed to boost student‑faculty interaction, including dining‑hall meal swipes for staff and faculty and additional social events scheduled for the fall.
The dean’s remarks arrive amid a nationwide scramble for universities to integrate AI tools into teaching and assessment. According to Deming, Harvard’s strategy is to act rather than merely observe, while still acknowledging the necessity of careful, course‑specific guidelines.
Harvard’s next steps will involve crafting detailed guidelines in consultation with stakeholders. The resulting framework is expected to shape how the college’s courses balance AI use, assessment integrity, and community building in the coming academic year.
This policy shift echoes a wider debate about AI in education, where institutions must weigh the learning benefits of AI against the risks of cheating and eroding trust. Harvard’s plan to encourage AI in some contexts while tightening controls in others reflects a nuanced approach to these challenges.
Educators and policymakers will closely monitor the university’s new initiatives and forthcoming guidelines as they seek to integrate AI tools into higher education while preserving academic standards and fostering collaborative learning environments.
