Find Resources

4.4 Investigating LLMs

In this lesson, students interact directly with several LLMs to evaluate model performance and think critically about implications of model output for their work, their peers, and for society.

In this lesson, students interact directly with several LLMs to explore differences in model output and performance. Through a guided investigation of LLM responses to different prompts, students learn that LLM output is 1) a probabilistic (answers are different nearly every time) and 2) not based on ground truth. They also learn to build different prompts or prompt frameworks for different purposes.

4.2 Fine Tuning

In this lesson, students do not use LLMs. Instead, then engage in playful, hands-on/unplugged activities that focus on the crucial stage of fine-tuning LLMs, highlighting the human influence on their behavior and output.

4.1 Pre-training

In this lesson, students do not use LLMs. Instead, then engage in playful, hands-on/unplugged activities that aim to demystify how LLMs learn language by exploring concepts like tokenization, vectors, and attention mechanisms. It consists of three activities, each building upon the previous one.

4.0 Intro to LLMs Lesson

In this lesson, students ≥13 use LLMs (i.e., chatGPT, Gemini, etc.) (teachers demo for students <13) to engage in a series of investigations that reveal how different LLMs can output misleading and/or biased information. By exploring and discussing these limitations, learners realize that biased output can be the result of patterns from the data, “hard-coded” or programmed directives from the developers, or personal bias.

Career Interview

This Career Interview activity is an extension of 1.5 Planting the Seed of STEM Jobs. Students, Our Young Professionals, are asked to interview a professional who they admire. The goal is to gain insight (inside info) into the planning, education journey and future prospects of a career of interest.

Cultural Lens

This activity is an extension of 0.5 Investigating Bias. Students will create a visual of their culture as a reference point to see how A.I. activity results or output may or may not represent their culture values, language and celebrations. The visual can be in any open-ended, creative form, like glasses or sunglasses, a wheel, or story map, or natural landscape with labels or descriptions of cultural features.

ELL Language Support Modification

This modification adds to the discussion of new vocabulary to try to deepen prior knowledge connections to greater support English Language Learners (ELL) with new vocabulary acquisition. Additionally, adding realia to the game supports ELL students.

Career Exploration and Impacts of AI Lesson plan with print Worksheet and Google Form Exit Ticket

Career Exploration and Impacts of AI for post secondary training, 2 year and 4 year degrees. Emphasis on continuing training or traditional education in STEAM careers whether it be an apprenticeship, certificate program, 2 year college, 4 year university or beyond. Emphasis on AI impacts in all fields. Combined Inventory of Me, AI’s Impact, and Roadmap to My Dream Job into a WS (2-3 day portfolio-like). They fill it in along with slides as they go.

Here is the image of a quesadilla with an algorithm creatively written on it. Enjoy the unique blend of culinary art and coding!

Best Quesadilla

I changed the food to be more culturally relevant to our school and our New Mexican culture. Instead of designing an algorithm for the best PB&J, this lesson tasks students with planning the best quesadilla. Our school population is majority Hispanic students who live in Albuquerque and mostly come from the South Valley. They will have more of a connection to quesadillas than to PB&J sandwiches. Optimizing a quesadilla algorithm for stakeholders may be more relevant to their culture and family.

Here is the image of an ethical quesadilla, beautifully presented with ethically sourced ingredients. Enjoy!

Quesadilla Ethical Matrix

For this lesson I changed the food listed to quesadillas (building on the previous lesson, Algorithms as Opinions), I also added a formative assessment to better help students connect to stakeholders in their lives and communities. Students who participate in the previous lesson (Algorithms as Opinions) will be able to carry it forward into a real-world connection by thinking about their out-of-school activity, the relevant stakeholders and their values.

Best Bocadillo

Making the best sandwich instead of PB & J for those who have not experienced PB & J. The reason I made this change is because there is a large Hispanic population in my area who do not make sandwiches with peanut butter and jelly and so a sandwich in general is more relevant to students of diverse populations such as Latinx, Asian, etc.

Ethical Matrix Lesson for Food Deserts

High school (HS)Students: Use ‘Teen’ instead of ‘Child’ Added ‘Store’ instead of ‘Doctor’ Teachers can talk with their students about topics relevant to their lives like: a)workers in the store; b)food available in local stores; c)accessibility of local stores; d) impact on people, supplies, and accessibility of stores

Culturally Sensitive Introduction to Supervised Machine Learning - Is it Frankenstein?

In this lesson students are giving the option to replace webcam images of themselves with webcam images of printed depictions of Dracula and Frankenstein. The training and testing images have been selected to help with the lesson's introduction to algorithmic bias.

Neural Network Moves

These modifications were made to organize the procedure of the neural network game.

Algorithms as Opinions (modified lesson)

This lesson includes modifications to make the lesson more accessible to students from different backgrounds, built in supports for students who have learning disabilities or students who are English language learners, and optional extension activities.

Ethical Matrix Lesson (modified)

This ethical matrix lesson includes modifications from the original to be more inclusive of students from different cultural backgrounds. It also includes modifications to support students with learning disabilities and English language learners in accessing the concepts in the lesson.

Introduction to bias for ELLs

The change made was adding an introduction to bias, distinguishing between facts and opinions, perspectives, and biases, tailored for ESL, bilingual, and regular students to comprehend bias in AI. This modification was designed to provide students with varying language proficiencies and backgrounds a foundation in understanding bias within the context of artificial intelligence, fostering inclusivity and equitable learning experiences.

Classifying living things

The adjustment involved incorporating a lesson of classifying living things to illustrate decision trees, crafted for ESL, bilingual, and regular students to comprehend decision trees in AI. This adaptation aimed to offer students with various language proficiencies and learning styles an accessible means to grasp decision-making processes, promoting inclusivity and enthusiasm for decision trees concepts.

Search for "[keyword]" across the whole site.

If you can't find what you're looking for, send us a comment about what you were expecting to find.

Everyday AI