# mk-ai-activities-list

-

# August AI Sessions

![alt text][image1]

## Teaching with AI: 10 Activities to Get You Started

### read your course’s policy (and discuss it\!)

**goal:** understand how AI can help (or hinder) student achievement of learning objectives (and understand what the course head is wanting them to do).

If they already have their course’s policy, they can read it, but we will have multiple sample policies for them to work with for the purposes of the activity.

* read your policy
* discuss with a partner
* share out with the group

### get connected to the Harvard supported tools\!

**goal**: make sure they all get over the hump of getting connected (for later activities, but also for their work this term)

* Bok staff with map out the available tech stack (i.e. Sandbox, gemini)
* Participants log into gemini for activities for the day

### ask the AI a simple question about your course’s material

**goal**: ask Gemini the most basic question (to get started) about the course they are teaching. understand what the LLM gets right and wrong when prompted in a simplistic way, and reflection on how this intersects with what they want students to be learning

### level up your AI intuitions

* Multiplication activity:
* Ask students to multiply a large number (\>5 digits) by another large number– without code.
* Then ask it to solve the same equation with code
* Compare outputs; discuss why a purely statistical engine can be mostly right yet occasionally wrong.
* “Where would a next-token predictor excel or fail in your discipline?”

### How are your students using AI?

**goal**: XX

* Have people turn in short poll (mix of multi-choice and written short answers) on Poll Everywhere.
* Card sorting activity with these statistics:

[https://www.chronicle.com/article/how-are-students-really-using-ai](https://www.chronicle.com/article/how-are-students-really-using-ai)
[https://www.grammarly.com/blog/company/student-ai-adoption-and-readiness/](https://www.grammarly.com/blog/company/student-ai-adoption-and-readiness/)

* According to Grammarly Jul 2025 study, 2000 students in US higher ed, 87% use AI for school, and even more (90%) use it for general life admin
* Numbers vary by study, but rapid surge is noteworthy; according to HEPI 2025 study, approx. 1000 students in UK higher ed, in 2025, the student use of AI has surged in the last year, with almost all students (92%) now using AI in some form, up from 66% in 2024, and some 88% having used GenAI for assessments, up from 53% in 2024
* Why use AI? According to Grammarly, Helping to brainstorm ideas (49%), checking grammar and spelling (42%) and understanding difficult concepts from classes (41%) are the top three ways students are already putting AI to use
* HEPI: Most commonly cited reason to use AI was to save time (51%), followed by improving one’s work (50%) and getting instant support (40%)
* CDT report: school \- worrying about falling behind; general life \- searching for info, advice seeking
* Need for guidance: According to Grammarly, 55% of students admit that they feel like they’re navigating this territory without proper guidance
* 62% believe that learning how to use AI responsibly is essential for their future career and success; half of students believe that learning how to use AI is the most important skill they’ll gain in college

* Take csv from first poll, analyze it with a python notebook for trends/patterns, related to the stats we did a card sorting activity for.

### How to level up your students’ AI intuitions

**Goal**: some students think this is a “magic black box”-- for students to use it better, they need to have a better understanding of how LLMs work

* [TikTokenizer](https://tiktokenizer.vercel.app/)
* Semantle, [embedding projector](https://projector.tensorflow.org/)
* [How transformers work](https://ig.ft.com/generative-ai/)

### try these prompting practices

**goal**: Level up prompting skills and deepen understanding of how AI can both help and hinder.

* Begin with a discussion of basic prompting strategies
* Introduce Karpathy’s idea of “context engineering”
* Share strategies for saving prompt snippets, reusable text, and reference structures
* Hands-on annotation activity:
* Distribute printed copies of a deliberately poor prompt
* Participants annotate the prompt with pens
* Collect annotated sheets for photo/CV analysis of the edits
* Use their collective feedback to create a “group edit” version of the prompt, which will be revisited when we do the “complex” question activity, right after we toggle tools on and off.

### toggle on and off the “tools” in the Harvard supported interface

**goal**: develop intuitions about what each of these things does, and how it relates to what students are learning

* Playground:
* Models
* Thinking vs. Regular
- 2.5 vs 2.5 Pro
* Tools:
* Deep Research
* File Upload (RAG)
* Modalities:
* Text vs. Voice
* Canvas

### ask the AI a more COMPLEX question about your course

**goal**: returning to when they initially prompted the LLM at the start of the session to test new techniques and tools. Will now move to NotebookLM.

* BEFORE we begin, take the “group edit” prompt from a previous activity, and show side by side output. The original “bad” prompt versus the group edit.
* First notebookLM:
* Seminal paper in their field they know well– query
* Second notebookLM
* A few documents from their course canvas
* Ask a complex question and test this output vs the initial output from plain gemini

### try out some more advanced moves

**goal**: allow interested students to push into more advanced tools/capabilities

* Data Visualization
* ask ChatGPT to create a graph
* 4o
* Then ask Claude to generate a graph
* Claude artifact
* Explore the world beyond LLMs, and generate some other materials:
* VEO/Sora
* ElevenLabs
* SD and ChatGPT 4o image generation
* notebookLM podcast
* Preview of APIs via Colab

### AI drawbacks/ethical/pedagogical concerns

**goal**: bias, environment, etc. highlighting reasons you may not want to use it, before getting into AI proofing.

* Activity ideas:
* put their writing in an AI detection tool
* bias/image generation-- source prompts from the audience
* [Simulated LLM training](https://www.nytimes.com/interactive/2023/04/26/upshot/gpt-from-scratch.html)
Hallucinations \- Bender (LLMs are trained on form, not meaning; deep sea octopus analogy: [https://nymag.com/intelligencer/article/ai-artificial-intelligence-chatbots-emily-m-bender.html](https://nymag.com/intelligencer/article/ai-artificial-intelligence-chatbots-emily-m-bender.html))
Pedagogical concerns \- [https://arxiv.org/abs/2506.08872](https://arxiv.org/abs/2506.08872)
* LLM vs. search engine vs. brain-only comparison on cognitive activity for essay writing task, found lower levels of engagement with LLM only
Cheating
* Conflicting info on whether cheating has increased with the introduction of ChatGPT ([https://www.sciencedirect.com/science/article/pii/S2666920X24000560?via%3Dihub](https://www.sciencedirect.com/science/article/pii/S2666920X24000560?via%3Dihub) vs. [https://www.theguardian.com/education/2025/jun/15/thousands-of-uk-university-students-caught-cheating-using-ai-artificial-intelligence-survey](https://www.theguardian.com/education/2025/jun/15/thousands-of-uk-university-students-caught-cheating-using-ai-artificial-intelligence-survey))
* But unreliability of AI detection tools, inconsistent application of academic integrity policies ([https://www.intelligent.com/4-in-10-college-students-are-using-chatgpt-on-assignments/](https://www.intelligent.com/4-in-10-college-students-are-using-chatgpt-on-assignments/))
Environment
* [https://news.mit.edu/2025/explained-generative-ai-environmental-impact-0117](https://news.mit.edu/2025/explained-generative-ai-environmental-impact-0117)

### AI-proof a section activity

**goal**: a return to the ai-policy starting activity, focused on their own agency within their sections. Driving home the importance of in-person, oral discussion within section. Transitioning– not just a way to ensure students don’t use AI, but also a way to give students who refuse to use AI another option.

### get connected to the Bok Center team

**goal**: for graduate students looking for extra help, or to work with the Bok Center this year on AI-related projects (or to connect their faculty with us) we’ll outline what we’ll be doing this year in the Bok AI Lab.

## takeaways