Academic Rigor at SIPA in the Age of AI
Shockwaves in Academia
From the Covid pandemic to the rise of generative Artificial Intelligence, this generation has gone through radical revolutions in the modes and means of teaching.
Once constrained to the realm of fiction, LLM tools now afford one and all access to instant proof-readers, virtual teaching assistants, meeting note-takers, and beyond.
Faced with this new challenge, the administration at Columbia SIPA has adopted a laissez-faire policy of experimentation, leaving it up to the professors to adapt the content of their teachings, methods of examinations, and policies on the use of AI for coursework.
Rethinking Exams
In this context, Dr. Stephen Biddle, director of the international security & diplomacy concentration, has introduced a new examination mechanism that steers clear of the traditional academic “blue-book” written exams. As you would with your team or your boss, so now are students expected to provide briefings on given subjects during one-on-one examination sessions.
The use of AI tools is welcome in their preparation, but students are expected to sustain a 40-minute critical interrogation of their analysis immediately following their briefing. Outside notes and digital supports, of course, are not allowed. One must make it out alive with none other than a printout of their deck, acquired knowledge from the course, and both the magnificence and the quirks of their own brain.
The onus of the learning outcomes are therefore less about writing skills or the use of AI tools, but placed entirely on the student’s capacity for understanding the subject and using analytical frameworks. It also reflects a closer alignment with the realities of the work environment, where one’s own conclusions must be defended when reviewed by superiors, colleagues, or editors.
Students have generally welcomed this shift, as it fosters deeper exchanges with their professor who is able to identify and immediately consolidate gaps in the student’s mastery of class concepts. Exams have thus become a better opportunity for growth and learning.
Other Approaches
Others have approached this evolution differently. Leave it to the cybersecurity professors to create an LLM of their own, specifically trained on cybersecurity issues of all kinds including the contents found in their course syllabus. This is exactly what Professor Matt Devost – who teaches a Cybersecurity & Business Risk course – crafted, through the Devost.ai LLM. He rejects outright bans on such technologies acknowledging their inevitable use in the workplace, and chooses instead to harness their power to enhance the learning experience.
Some intrepid professors have even dared to force the use of AI models as a supplementary tool with which to complement (not formulate) strategic thinking. Having written the first half of an assignment themselves, students enrolled in the Strategic Entrepreneurship course were then asked to make use of at least two LLM models to generate insights based on their pre-existing work. The final write-up became a product of students’ own strategic blueprints, enhanced by AI-generated insights for which critical analysis skills were employed to filter the good from the bad.
Getting Artificial Intelligence Right
The use of AI by students is counterproductive when it substitutes for analysis. When used correctly, however, Dr. Biddle argues that such tools can enhance both academic performance and future career performance. This depends on the ability to compartmentalize their use to clerical, repetitive, and time-consuming tasks, while preserving and sharpening one’s analytical capacity.
That capacity constitutes a student’s true comparative advantage on the job market. The ability to synthesize information, apply frameworks in context, and produce defensible conclusions is slow to develop but far more durable than rapidly evolving AI-tool proficiency. Moreover, as digital tools generate increasingly large volumes of information, the value of such judgment will only increase.
In the era of instant digital access the decisive skill is no longer finding information, but making critical sense of it.