Why AI Didn't Break Econiful’s Assessments
Last year, Princeton reversed a 133-year-old no-proctor exam policy because the tools available to cheat had gotten so good that the honor system stopped being a workable assumption. According to articles like The New York Times' "How A.I. Killed Student Writing (and Revived It)" and The Tufts Daily's "Professors Shift Many Essay-based Assessments to In-person Exams Amid AI Concerns," educators at Tufts, at community colleges in Brooklyn, in AP classrooms in Atlanta and Philadelphia, and beyond have been arriving at the same conclusion about out-of-class writing assignments: as a reliable measure of what a student actually knows, they've become very hard to defend.
Between May and December of 2025, the share of students regularly using AI for homework climbed from 48 to 62 percent, according to More Students Use AI for Homework, a RAND research report. A third reported using it specifically to draft or revise writing. A study from the Higher Education Policy Institute, as reported by The Tufts Daily, found that 88% of students surveyed were using ChatGPT or similar tools for assessments, with 8% submitting AI-generated work without editing it at all.
For economics teachers, this isn't just an integrity problem worth lamenting. It's an economic problem, and it has an economic answer.
The Case Against Out-of-Class Assessments in 2026
The benefits of take-home work used to outweigh the costs. AI has flipped that equation.
The opportunity cost of an in-class performance-based assessment is the instructional minutes you'd otherwise spend teaching new content. For years, that tradeoff made sense to avoid. The benefits of out-of-class work, including saved instructional time and room for deeper individual effort, clearly outweighed the costs: occasional plagiarism, manageable with a detection tool. But in this day and age, that balance has flipped.
On the benefit side: if 80% of an essay or other writing-intensive assessment can be AI-generated and you can't detect it, you're not measuring what you think you're measuring. The validity of the assessment has collapsed.
On the cost side: suspected AI use is nearly impossible to prove. Humanizers rewrite AI-generated text to sound more like a student; autotypers simulate the typing process in real time, complete with fabricated typos and deletions, so version history looks authentic. A June 2026 investigation by New York Times education reporter Dana Goldstein, "Student Cheating Is Becoming Impossible to Detect in an A.I. Era," found that these tools are being aggressively marketed to students on TikTok and YouTube, sometimes labeled as ads, sometimes not. Jenny Maxwell, head of education at Superhuman (the company behind Grammarly), told The New York Times that the race between detection and evasion is "ultimately, a dead end." Educators trying to keep up are, in her words, looking at a "bigger cat, bigger mouse" situation with no finish line.
Add the common pushback, "I'll be able to use AI at work, so what's the problem?", and the cost side expands even further. That argument misses something foundational though: students who lean on AI for everything never build the underlying content knowledge and skills. Without that foundation, they can't fact-check AI output or judge when it's actually helping them. Nicholas Anderson, a former lecturer in political science at Tufts, put it this way in an interview with The Tufts Daily: "You have to really have the experience of being uncomfortable and not knowing the answer, and really struggling with it...This slow, patient thinking, which is one important experience of a good liberal arts education, is undermined by AI and its instantaneous gratification."
Catching AI Cheating Is Getting Harder
The marginal cost of catching one more AI cheater is rising. The marginal benefit is falling toward zero.
The marginal cost of catching one more instance of AI-assisted cheating on an out-of-class assignment, including detection software subscriptions, the humanizer-detection arms race, student appeals, and lack of at-home support, is high and rising. The marginal benefit of catching that instance, meaning one more data point on whether a student actually knows the material, is falling toward zero, since detection is failing. Our in-class assessments sidestep this by asking students to demonstrate their learning in class using paper-and-pencil assessments.
How Teaching Strategies Adapt when External Conditions Change
The right policy five years ago isn't the right policy now. External conditions changed, and assessment strategy has to follow.
The interdependence principle tells us the best policy choice, in part, depends on external conditions and expectations about the future, and those conditions have changed. Sending performance-based assessments home and relying on tools like Turnitin was the right call five years ago because it protected instructional time. It isn't the right call now. Princeton reversing a 133-year-old no-proctor policy is the clearest possible proof that external conditions dictate the optimal choice, not tradition or preference. And Princeton isn't an outlier. As The New York Times reported, the shift toward mandatory in-classroom writing is happening across the educational landscape, from suburban districts and urban charter schools to community colleges and the Ivy League.
Econiful's Approach to Assessments
Econiful's assessments were designed for the classroom from day one, not as a response to AI, but because that's where real learning happens.
While folks strive to rethink their assessment strategies, Econiful's assessments were designed for in-class implementation from day one.
Each summative assessment pairs a rubric with an in-class introduction and a dedicated instructional day to complete the assessment. Most especially, this includes lessons like Lesson 1.12, Lesson 2.9, Lesson 3.8, Lesson 3.13, Lesson 3.19, Lesson 3.20, Lesson 4.7, Lesson 4.14, and Lesson 4.15.
Beyond summative assessments, Econiful's curriculum was built around discussion-based learning, productive struggle, and handwritten in-class work. Not as workarounds for an AI problem, but because that's solid economics pedagogy.
That also means the common objection, "We can't afford to lose an instruction day to an assessment," becomes less of a pain point. When the prep for an assessment and the assessment itself are both meaningful and mastery-building, the day isn't lost. It's instruction in a different form. The perceived cost of giving up a day shrinks accordingly.
The Bottom Line
The conversation about AI in education is moving fast, and it should be. But for economics teachers, the analytical tools to navigate it have been there all along. Opportunity cost. Marginal analysis. Interdependence. The same principles we teach students to apply to the world apply here too.
Econiful's curriculum was built to meet students where the learning actually happens: in the room and in the conversation.
Want to see these assessments in action?
Become an Econiful member for free access to all 80 lesson plans, including every in-class assessment and rubric. Everything you need to teach economics with confidence.