AI Policy and Agreement

SOC 208, Fall 2025, Prof. Delehanty

AI chatbots, such as ChatGPT, aren’t actually intelligent. Instead, they are pattern identifiers. Here is how they work:

1. They parse your prompt to determine what kind of information you’re seeking and how you want it presented.
2. They process the information available to them, aka their training data (much of which is factually wrong, created by other AI chatbots, and/or used without permission) to see how it typically addresses the kind of query you provided. The type of output an AI interface gives you depends on what data it’s trained on.
3. Using extremely complex math and enormous processing power (at great environmental cost in terms of water and energy usage), they identify linguistic and numerical patterns, such as word sequencing, sentence structure, etc. in their training data as they pertain to your query.
4. Based on this mathematical analysis, they generate a “result,” that is, a combination of letters, numbers, and other figures that fits the patterns they found.
5. They give you that result, tailored to any style or formatting specifications you provided. This tailoring is also based on pattern recognition (if you tell a chatbot to respond in pirate language, it does that by applying patterns of “pirate-talk” it has found in its training data to the result it came up with for your query).

In other words, AI gives you the best approximation it can of what it determines you want, based on patterns it has been programmed to find in whatever data it was trained on.

This means that by definition, AI cannot produce original thoughts. It can only analyze existing information and then provide you with a combination of letters, numbers, etc. that aligns with patterns it identifies. For example, when you ask an AI chatbot for a list of citations, it doesn’t usually give you real citations. Instead, it produces text that it has determined looks like citations, based on the characteristics of whatever citations it finds in its training data. For this reason, very often, AI-generated bibliographies contain citations to books and papers that have never been written. Recently, the Trump administration provided Congress with a “Make America Healthy Again” report containing seven citations to academic publications that do not actually exist.

To put it another way, AI is a bullshit machine. I don’t mean that everything it produces is false. As philosopher Harry Franklin explains, bullshitters present information without concern for whether it is true or not, in order to impress, persuade, or advance a narrative. This is what AI does. It produces information that is plausible, but truth-neutral. It does not—cannot—know or care whether the information it gives you is true. It can only determine whether its output fits the patterns it was programmed to look for in the data it has access to.

This means that when you use AI for writing, reading, summarizing texts, distilling notes, etc., you are bullshitting. Not lying, per se, but doing your work without regard to its quality or its veracity. Generating meaningless words on a page for no other purpose than to fill blank space. Letting a machine do your thinking for you. But the machine can’t actually think.

This is a problem, because thinking is how we learn. Thinking happens through repeated intentional practices such as reading, concept mapping, summarizing, memoing, comparing, and writing. Even (especially!) when it’s hard. In my view, the two most important outcomes of a Clark education are 1) learning to think for yourself, and 2) learning how to communicate that thinking to others in speech and writing. Using AI undercuts these skills because it means you aren’t actually thinking or communicating for yourself at all. And when you are not thinking, and not learning to communicate your thoughts, you are not meeting the learning objectives of this course, developing the skills you came to Clark to get, or obtaining the education that you are paying so much for.

For these reasons (as well as others that I’m happy to discuss), students in this class must agree not to use generative AI in any substantive^[1] capacity for any course-related work. No asking it for summaries of texts. No using it to organize your notes. And certainly no using it to produce, revise, or edit the written work you submit to me.

This policy includes Grammarly and similar services (because by reformatting your writing for you, they stop you from learning how to communicate your thoughts yourself). I would much rather read your own imperfect prose than the robotic nonsense that Grammarly turns your work into. Only by editing and polishing your writing yourself can you develop your own voice, that is, learn to communicate effectively. Don’t worry, I don’t take points off for the kinds of errors that students typically use Grammarly to “fix,” and I don’t insist on the formal academic writing rules that students tend to worry about. Writing need not “sound smart” to be good.

For these same reasons, I also pledge to you that I won’t use AI for any substantive part of my work in this course. Every email you get from me will be written by me personally. Every assignment prompt, exam question, and piece of commentary on your work will be my own. I will not use AI to draft lectures, generate reading summaries, annotate PDFs, organize my notes, grade your work, or anything else. In an era of “AI-native universities,” I will be an AI-free professor.

Therefore, to promote a productive learning environment for all, students must agree to the following terms:^[2]

1. Students will not use AI in any element of the work related to this class, including but not limited to:^[3] 

• generating, organizing, revising, or refining work that you submit, including formal papers, lower stakes assignments, discussion posts, or anything else
• organizing, annotating, or summarizing notes from class
• summarizing readings

2. Students acknowledge that any first violation of this policy will result in a grade of zero on the relevant assignment^[4]
3. Students acknowledge that any second violation of this policy will result in referral to the College Board for violating Clark’s Academic Integrity Policy

Student Printed Name: _____________________________________________________

Student Signature and Date: _________________________________________________

Additional Resources: Alternatives to AI

I acknowledge that AI can be a tempting shortcut, that all your friends are using it, that you may be using it in other courses, that some professors may say it’s okay to use it, and that it’s hard to break established habits even when you want to. To help you work through these challenges, below are some alternatives to using AI. Try these out when you feel the itch to call up ChatGPT. And know that I grade work carefully and supportively. You won’t be penalized for minor imperfections in your writing or for not catching every nuance of every reading, and you’ll be rewarded for working through challenges authentically, even if that results in errors or incomplete thoughts.

For generating ideas, drafts, etc.

• Make a concept map. On paper, make a list of all the ideas/words/things/themes related to your writing topic. Don’t worry about what order they go in or other details. Just put them all down. Then, turn your list into a concept map. Put the words into bubbles and start drawing arrows between them. (i.e., thing X is related to thing Y). Then, describe those relationships (i.e., X is a necessary precondition for Y to happen, or X is a specific example of bigger pattern Y). Then make those relationships into the focus of sentences or even paragraphs. Voila – you’ve got a draft started.
• READ. When you’re struck writing, read. Even if it’s something not related to class. Reading other people’s sentences will help you craft your own. Seeing ideas presented clearly will help clarify your ideas. When I’m stuck on a writing project, I switch to reading. Almost every time, within 10 minutes spent carefully reading something else, I get an idea that helps me get back to writing again.
• Go back to something you’re written before that you liked. Read it a couple of times – first, to remind yourself that you can do this. This builds your confidence. Then read it again and try to craft a couple paragraphs on your current topic that mimic the structure of strong aspects of your previous work.

For summarizing challenging texts or passages

• Find the section breaks and think about why the author put them where they did. Read the first and last paragraphs of each section. This will give you a skeleton of the text’s meaning.
• Choose a manageable part of the text (no more than a few paragraphs). Read each paragraph carefully and write down a very few words (4-6 will do) about what that paragraph says. When you’ve finished the section, take the words you’ve written down for each paragraph and turn them into a sentence or a list of bullet points. Boom. Summary.
• Find credible summaries online. Just about any book you get assigned in a Clark class will have been professionally reviewed by other scholars or experts. Find their reviews. For mainstream books, a simple web search will turn them up (but don’t use the AI-generated summaries in google; follow links to human-produced work instead). For more niche books, a library database search may be necessary. I can show you how to do this if needed. Wikipedia also often has good summaries of well-known books, in many cases written by the author themselves.

For polishing your prose (as an alternative to Grammarly)

• Again, READ. Read whatever you want. Find a couple sentences or paragraphs you like. Think about why you like them. Then apply what they do well to your own writing.
• Cut words. Choose a chunk of your text, run a word count on it, and try to cut 20% (or more) of the words without losing any meaning. Almost every time, this will make your writing better.
• Work on transitions (beginning and end of paragraphs). The transitions are often what make your writing good or not so good.
• Ask a friend or professor for feedback.