Guidance by Assessment Type

Each section identifies 1) some example categories of learning objectives for assignments of that type, 2) where students could use AI to substitute for their own performance, 3) ways to adapt the assignment for AI resilience, and 4) some sample assignment-specific AI policy language you can adapt for your assignment prompt. The adaptations are options and ideas among which you can choose or use as a jumping off point for your own ideas.

1. Argumentative essays and research papers

Example learning objectives: Taking a position and defending it; selecting evidence, anticipating objections; sustaining a line of reasoning; framing a research question; finding and evaluating sources; synthesizing sources; analyzing data; formulating an argument in text.

AI substitutes by: producing fluent, well-organized arguments based on sources provided by the student or sources it finds.

AI-resilience strategies

Adapt this assignment with scaffolded, mostly in-person touchpoints.

Oral topic proposal. The student states the position they intend to defend, in person, in office hours or section, in about five minutes, and answers two or three questions about it. You find out what they plan to argue before any text exists.
In-class thesis paragraph or outline. Students write it in class, on paper, with devices away.
Staged draft. One section or one paragraph, submitted early enough that your feedback reaches the student while they are still writing. Having students conduct peer reviews of drafts will also provide them with feedback and encourage them to engage meaningfully with the work because they know their peers will see it.
Short structured oral exchange after submission. The student articulates the argument they committed to in writing and answers questions about how they arrived at it.
Reflective statement submitted with the essay. The student describes their process and the reasoning behind their choices.

Anchor the prompt in course-specific material. A text you read together, a claim made in lecture, a position argued in section. The evidentiary base of the essay is then something the student personally engaged with.

Provide an AI-free writing space.

Hold quiet writing sessions with instructor support. Devote some instructor or TF office hour time to holding writing sessions in the library or other spaces. Have students close everything but their word processing applications while they are present or ask them to work with pen and paper.
Form peer writing groups where students get together for writing sessions with their AI and possibly all Internet access turned off. Have students complete short logs of what they worked on and what they discussed with peers during the sessions.

Sample AI policy

Option 1

This assignment develops and assesses your skills in analyzing texts and constructing a well-supported argument. If you rely on AI to summarize the readings, develop your argument, or draft your paper, it will interfere with your learning and your success on the in-person elements of the assignment. For that reason, AI should not be used to perform these or other substantive tasks for this assignment. Generative AI search results, grammar suggestions, or other built-in generative AI features may be used when thisuse is incidental.

Option 2

This assignment develops and assesses your skills in analyzing texts and constructing a well-supported argument. If you rely on AI to summarize the readings, develop your argument, or draft your paper it will interfere with your learning and your success on the in-person elements of the assignment. You may choose to use AI tools as you work on this assignment but should not use them to substitute for your own thinking. You must include an AI statement with each submitted element of the assignment indicating whether you used any AI tools, which tools you used, and how you used the tools.

2. Reading and reading responses

Example learning objectives: tracking an argument across pages of text, close reading, disciplinary analysis.

AI substitutes by: delivering a summary, writing cogent short responses.

AI-resilience strategies

Adapt this assignment with scaffolded, mostly in-person touchpoints.

Oral exchange in section. Well structured questioning in section that probes student understanding.
In-class writing. Move reading response composition to the first few minutes of section. Pose specific questions or draw student attention to particular points. Ask students to compare and contrast what they have just read with a prior reading.

Adapt this assignment by turning it into practice for an in-class assessment.

Assessed in-class reading. Assess reading and analysis in class so that students are incentivized to practice these skills in their out-of-class assignments.
Reading skills tested on in-class quizzes or exams. Assess reading skills and comprehension in proctored quizzes or tests. Ask the student to identify an author or era from a text, apply a concept from the reading, argue for or against a position based on the reading, or restate an idea in their own words. Avoid quizzes that check simple recall or non-pedagogically significant facts. The follow-up should be pedagogically useful in itself.

Adapt this assignment by making the use of AI unproductive relative to doing the work directly.

Paper annotation or collaborative annotation. Have students annotate on a paper copy of the reading or using an annotation app like Perusall. Though AI may still be used for summary and ideas, submitted annotations in these forms make doing the work directly similarly efficient to getting help from AI.

Provide an AI-free reading space.

Hold quiet reading sessions with instructor support. Devote some instructor or TF office hour time to holding reading sessions in the library or other spaces. Instruct students to purchase/rent hard copies of the book or provide course packs or hard copies of the readings.

Sample AI policy

The objective of our reading assignments is for you to learn to read and interpret texts. Using an AI summary will interfere with your development of this skill, your access to subtle or original perspectives, and your experience of structural or stylistic aspects of the text. For this reason, you are discouraged from using AI to summarize the text. The responses you write are an exercise in shaping your thinking and analysis through writing. Using AI to complete them completely defeats their purpose. For this reason, AI should not be used to assist in writing your responses. You may use AI as an interlocutor to ask questions about the text or about sources or ideas the text references that are not among the assigned readings.

3. Problem sets

Example learning objectives: application of concepts to new problems. Clear presentation of an argument or series of steps that lead to a solution. Exploration of a possible solution. Experimentation, data collection, and analysis. Production of code or prototypes.

AI substitutes by: producing correct, well-presented solutions.

AI-resilience strategies

Adapt this assignment with scaffolded, mostly in-person touchpoints.

Paper follow-up in class. A few minutes, on paper, after the pset is handed in. Ask the student to explain a step in their own solution, extend a result to a variant, explain choices they made and why they selected them over alternatives, or debug or explain a related piece of code. It takes a few minutes to write and a few minutes to grade. Avoid quizzes that check simple recall or non-pedagogically significant facts. The follow-up should be pedagogically useful in itself.
Oral follow-up in class, section, or a peer group. Ask students to present or defend some aspect of their work orally.

Adapt this assignment by turning it into practice for an in-class assessment.

Switch the problem set to practice sets with the grade based on an in-class test. This removes the advantage of a solution the student did not produce. It also lets you be permissive about how students get through the set. Consider offering students AI-backed tutor bots and other resources to help them get practice that increases in difficulty as they gain mastery.

Provide an AI-free problem-solving space.

Hold homework sessions with instructor support. Devote some instructor or TF office hour time to problem set sessions. Make these sessions device-free and provide course materials for reference in hard copy.

Sample AI policy

Option 1

The purpose of the problem sets is for you to go through the productive struggle of discovering how to apply the course concepts to the problems. This cannot be achieved if you seek answers from peers or generative AI without making progress yourself or with collaborators who are at the same point. You may use generative AI or other help sparingly if you get stuck. You should write your problem set solutions entirely in your own words. You will be expected to fully understand the work you submit and be able to explain it.

Option 2

The purpose of the problem sets is for you to go through the productive struggle of discovering how to apply the course concepts to the problems. This cannot be achieved if you seek answers from peers or generative AI without making progress yourself or with collaborators who are at the same point. You may use generative AI and other supports as you choose, but you will be expected to fully understand and be able to explain the work you submit.

4. Capstone and group projects

Example learning objectives: Typically 1) executing the full stack of tasks involved in producing real-world work in the discipline, or 2) exercising disciplinary learning objectives in a non-traditional modality on a topic selected by the student or group.

AI substitutes by: assisting with idea generation, research, scripting, design, editing, production, coding. Note that, separate from AI use, creative projects and multimodal projects in particular were already notoriously difficult to assess.

AI-resilience strategies

Adapt this assignment with scaffolded, mostly in-person touchpoints.

Oral proposal before project work begins. The student describes what they intend to make and what they intend their creation to argue or demonstrate.
Work-in-progress critique with structured peer feedback. Students show a rough version to peers and answer questions about it. Short, in-class presentations of rough drafts are also opportunities to have students share not just what their project is about, but what inspired them to pick a topic, in what class sources they are grounding their research, and with what elements they are stuck on or struggling.
Staged deadlines with a short check-in at each one. Students submit artifacts of learning throughout the term (proposals, outlines, source reviews, drafts, revisions, etc.) . The check-in is a conversation about the work in progress. The artifacts of learning could also be produced in section or in another proctored environment.
Statement of Goals and Choices, submitted with the final capstone. The student explains why this sequence, this audio, this framing. They provide evidence of deliberate decision-making, making the intellectual work legible even when the medium is unfamiliar to the grader. If this is done as part of an in-person touchpoint, it can be part of verifying that the student has done the decision-making themselves.
Live defense after an in-class screening or presentation. The student answers questions about the artifact in front of you and their peers.

Anchor the project in material that is not part of LLM training data.

Use the Harvard Collections. Have students center their projects on objects in the Harvard libraries or collections.
Generate data for analysis. Have students collect oral histories, interviews, experimental data, or other data they generate themselves for use in the project.
Incorporate elements of the in-class experience into the project. Ask students to use examples, discussions, or other aspects of the classroom experience as the basis for part of the project.

Hold projects to as high a standard as conventional written work. Ensure that your grading methodology for the project is robust and fair.

A rubric that scores intellectual content separately from technical execution. Decide which one you are assessing in each category of the rubric. If technical skill such as art or performance ability or media production is part of the goal, identify this skill and teach it rather than assuming students arrive with it.
Ungraded time to explore the tools before graded work begins. This keeps prior technical experience from becoming an advantage unrelated to course content.
For group projects, an individual process memo. Each student explains their own contribution. Discuss the expectation before the project begins.

Sample AI policy

AI may be used in the completion of this project but uses must be fully cited and disclosed in a statement that indicates what tools were used and how they were used. For tools that can produce a log of interactions, the log must be submitted together with the work.

5. Take-home exams

Example learning objectives: take-home exams encompass a broad range of learning objectives, most of which are compromised by access to generative AI.

AI substitutes by: completing the exam highly competently in a fraction of the time, with or without student editing and contribution.

AI-resilience strategies

Move it into the classroom. A seated exam gives reliable evidence of individual learning and tends to differentiate grades well. Seated exams are not a direct substitute for take-home exams, so some objectives of take-home exams will have to be moved into other assessments such as projects, problem sets, or papers. We suggest you give the exam on standard printer paper so student exams can be scanned for grading (standard paper is easier to scan than blue books because standard sheets of paper can be scanned as a large stack) or that you use a Canvas Quiz with Respondus Lockdown Browser enabled.
If you maintain a take-home exam, follow it with an oral defense or live interview that counts for a considerable part of the grade.

Sample policy

You may not use generative AI and other resources as you work on this exam. The exam assesses your own command of the course material, and you will be asked to explain and defend your answers in person after you submit. Answers you cannot reconstruct and discuss will not earn credit. You must include an AI statement indicating whether you used any AI tools, which tools you used, and how you used the tools.