Plot Summary

SEASON FINALE!

A season of exploring comes to an end, as we realize how many different paths we explored in Spring 2023, and what paths lie ahead to explore going forward.

We started wtih a seven and half minute video lookback. episode by episode, at what happened in the Seminar, in the voices of various participants.

Then John Mitchell shared a seminar summary, taking us on a journey from where we started, what we did, some commonalities and understandings, and some uncharted territories.

And then the Seminar concluded with two sense-making activities in small groups. First we looked back and looked forward, reflecting both on the personal and the field. Then finished by creating a time capsule, making some predictions of what wild success would look like for what you're working on, and what surprises might emerge for the field. 

Plot Summary

A special virtual-only session!

Chris Dede, Associate Director for Research at the National AI Institute for Adult Learning and Online Education (AI-ALOE), delivered a talk on intelligence augmentation (IA) and generative AI in education. He emphasized the complementary roles of humans and machines, stating that while AI excels at reckoning and calculative prediction, human judgment based on biological, social-emotional, and ethical/spiritual knowing is irreplaceable. Chris discussed the partnership between AI-based cognitive assistants and adult educators, highlighting the need for upskilling to effectively leverage AI in education. He cautioned against confusing performance with competence and the misuse of oversimplified "suitcase words" in discussing AI capabilities. The talk underscored the importance of understanding the limitations of AI and fostering a productive partnership between humans and AI in the learning process. Participants discussed the material in Chris' talk, exploring how some of the concepts might be applied in a real scenario, to help recent immigrants learn negotiation skills.

Miroslav Suzara and Sierra Wang facilitated an activity at the Seminar on Generative AI and Education. The purpose was to explore how generative AI can enhance social interactions. Participants were encouraged to use Zoom's live transcript feature and save the transcripts to analyze and generate insights using platforms like GPT. The activity involved breakout rooms with three different exercises: acting through a scenario, brainstorming, and planning. Feedback and reflections were shared, highlighting the potential of AI-generated feedback and the complexities surrounding its delivery. 

Plot Summary

Bethanie Maples, Matthew Rascoff, and Robert Prakash led an activity on the opportunity to integrate generative AI into a large educational resource at Stanford University, ClassX.

 They discussed using a corpus of almost 10,000 educational videos by Stanford professors to improve learning and education through generative AI. They explored the pedagogical and technical aspects of utilizing this corpus and engaged in design activities to envision the potential of the resource. 

In small groups, seminar participants discussed various ideas, including generating personalized recommendations, creating interactive deep fakes for professors to discuss topics (noting, perhaps a few ethical issues to sort through here!), using language models for natural language-based content search, and analyzing the corpus to identify gaps in teaching and formative assessments. Ethical considerations, such as privacy and promoting critical thinking, were also discussed.

Plot Summary

Many of the usages of generative AI for education so far have focused on the interaction between a human and an AI agent, but what if we shifted our perspective to ask:  
What if generative AI could be used to help humans interact with... humans?

We explored this perspective through the context of collaborative learning, in an activity co-facilitated by Jenny Osuna and Glenn Fajardo. Participants were introduced to a pantry of INGREDIENTS that included:

• ELEMENTS (intersubjectivity, multivocality, social skills)

• MOVES (eliciting, probing, revoicing)

• SCENARIOS (small group facilitation, noticing marginalized voices, fostering healthy team dynamics)

• STRUCTURED  (for Generative AI to be applicable, it needs to be structured - for example can you imagine writing a prompt?)

Then participants formed small teams and were given a mission to create a prototype, in the form of a skit, of how Gen AI could be used to support collaborative learning.

Hilarity and insights ensued, as teams performed their skits for each other and discussed what the skits made them wonder about going forward. 

Different approaches were taken by different teams, such as individual AI agents for each participant, AI agents supporting human mediators, or AI agents acting as participants or facilitators. Specific questions were raised about how AI agents could be used to enhance and scaffold human interaction rather than replacing it entirely.

If you're interested in how generative AI can be used to support collaborative learning, Jenny, Glenn, and John Mitchell are working on a project called Humanizing AI for Better Collaborative Learning. Feel free to reach out: glenn (at) dschool.stanford.edu

Plot Summary

...and before we knew it, participants had created these bots, inspired by experiences with their teachers:

• CaptainCook2, 

• PandaBot

• ProfKnowall, 

• Quackdratic

• SpinnerBot

How did we get there?

Josh Weiss led us through a journey where we first reflected on a moment in your life when someone taught you something in a really effective way. 

Working in small groups, with throughline team names and selfies, teammates shared their inspirations and worked together to promptify this., weaving together strategies such as analogy, contrasting cases, elaboration, and visualizaiton. 

Teams then shared their prompts in the WhatsApp group, then Josh and Chris Bennett fed those prompts into Poe as bots were born.

What can we learn from this quick and dirty build? Let's try them out, notice, and reflect...

Plot Summary

After a warm-up that focused on finding uncommon commonalities with a partner, we dove in to three talks:

• Alan Cheng: Designing Immersive, Narrative-Based Interfaces to Guide Outdoor Learning

• Emil Palikot: Smiles in Profiles: Improving Fairness and Efficiency Using Estimates of User Preferences in Online Marketplaces

• Chris Mah: Using ChatGPT as an “Example Machine”

Plot Summary

In an 80 minute interactive learning experience led by Leticia Britos Cavagnaro of the d.school, participants Explored the Generative AI Education Problem Space... Opportunity Space.... Responsibility Space.... Possibility Space. 

Participants warmed up in pairs, sharing a moment where generative AI adds (or could add) value to our work/life. 

Then individually, participants spent a few minutes unpacking thoughts into post-its on:  1) why AI is consequential for education,  2) what interesting generative AI assets are already available for learners and teachers, and 3) what could you or others create that doesn't currently exist.  

Then participants got up and formed different Human Maps based on how long you have been at Stanford, whether you are a student/faculty/etc, and academic discipline. 

From there, participants were able to form diverse small teams to Map the Problem Space of Generative AI and Education. 

Then teams travelled to the future, to create a future headline based on a provocation that emerged from their Map the Problem Space exploration.  

We took a Reflection Pause to individually reflect on the activity, the connections we made, and the ideas emerged, with the assistance of an AI reflection buddy called Rebot, that Leticia created.

Participants shared their future headlines and team selfies via the seminar WhatsApp group, and we concluded the exploration with a whole group debrief.

If you missed the session and would like to learn more, a recording is available here.

Welcome! We're so glad that you've come together for this Spring 2023 shared experience.

Plot Summary

The curtain dropped, welcoming a diverse group of Stanfordians (Stanforders?)  from various disciplines across campus, all interested in the intersection of Generative AI and Education. Many participants are actively working on something in the space.

John Mitchell welcomed everyone and shared the intentions behind this gathering. Can we connect and learn from each other, from each of our various engagements, expertise, and efforts around Generative AI and Education? How might we help each other? If you missed it, see the full welcome here.

We then jumped into an activity that challenged people to foster High Quality Connections in pairs. High Quality Connections, based on Jane Dutton’s research, are one-on-one interactions that light you up. Even 40 seconds of a positive caring interaction has measurable impacts on both people, and we'll try to foster High Quality Connections throughout the quarter.

And then... the main event of the first session: Five Lightning Talks:

1. Allen Nie: "Understanding the Impact of Reinforcement Learning on Subgroups of Students in Math Tutoring"

2. Sierra Wang: "MATH IDE"

3. Rose Wang: "Empowering teachers and students at scale: Providing localized feedback with natural language processing"

4. Eric Zelikman: "Parsel: A (De-)compositional Framework for Algorithmic Reasoning with Language Models"

5. Evan Liu: "Grading for Classrooms at Scale: Automatically Giving Feedback on Interactive Software"

We closed with some information about asynchronous connection activities (see next section for details) and then chatted away over food and drink.

Share a bit about yourself...

• Please take 5 minutes to Introduce Yourself on this MURAL digital whiteboard. Password is "highqualityconnections". By participating, you'll help increase the chance of useful connections during future seminar sessions and outside of those sessions.