Inspired by the proven model of deliberative assemblies worldwide, CCC and DemNext conceptualized, designed, and implemented the first of its kind tech-enhanced and student-focused version of a deliberative assembly in January of 2024. The assembly harnessed powerful technologies to create positive, constructive, human-led systems.
A Tech-Enhanced Student Assembly
Inspired by the proven model of deliberative assemblies worldwide, CCC and DemNext conceptualized, designed, and implemented the first of its kind tech-enhanced and student-focused version of a deliberative assembly and harness powerful technologies to create positive, constructive, human-led systems.
SETTING THE STAGE
Our pilot at MIT
Student Assembly participants pose for a group photo with the CCC and DemocracyNext organizing team at the end of the 3 days. Photo by Artemisia Luk.
CCC and DemocracyNext hosted the first-ever tech-enhanced Student Assembly at MIT during the Independent Activities Period (IAP) in January 2024. In this exploratory three-day event, a dozen students gathered to deliberate on ethical guidelines and regulations needed to govern the use of generative AI for faculty and students at MIT. Assembly members included both undergraduate and graduate students from a variety of academic disciplines:, from applied mathematics and computer science to urban studies and planning.
We started with the broad idea of how generative AI affects the lives of students and teachers, asking: How Might We Redefine Learning in the Age of AI? Relying on student feedback during the learning sessions, this framing became more tangible and specific.
The question
What overarching principles at the MIT-level should inform departmental guidelines for GenAI? How should departments enact those principles?
Participants were welcomed with materials and icebreakers to kickoff the Student Assembly.
What did students think before joining the assembly?
Advocates
“As long as it is used to help students think or achieve more, I am all for it!”
“AI will become an inevitable part of learning and it's most realistic to tailor pedagogy assuming students can access it. But, maybe fundamental mathematics and sciences will have to be taught in controlled environments, just like how the internet in general has been understood as being noisy and distracting for learning. What are the kinds of problem-solving skills that we can emphasize further if AI can figure out so many existing ones?”
“I recently used ChatGPT to come up with a climate change education career plan for primary school students... So I think that's pretty cool as well. Just make the lesson plan more creative.”
Skeptics
“We can’t ignore the risks of plagiarism, laziness, lack of personal thinking that comes with normalizing the use of AI in classrooms”
“There need to be guidelines for what is fair use of generative AI. There needs to be attention given to systemic concerns behind using generative AI (e.g. theft of training material, access of students to generative AI) and education concerns (e.g. students not being encouraged to doing themselves a disservice by taking the "easy way", and the quality of education if teachers are relying on AI).”
“Why do we need to make teachers more productive? Shouldn't there be a reasonable size of classroom? Is AI the tool to fix this problem of teachers feeling like they're not able to do their job?”
Before the assembly
Preparing the learning and evidence
Pre-assembly readings were distributed to help ensure shared knowledge among students from various fields of study - some of whom come in knowing very much about generative AI and others who are new to the subject. Reading material was intentionally selected with minimal bias: some sources were critical of the role of AI in education, while others were far more optimistic, leaving readers to draw their own conclusions.
Reading list materials
Click to open content from each topic.
What is generative AI?
What is a Deliberative Assembly?
What is a Citizens' Assembly?
What is the future of Gen AI in higher education?
What is the future of Gen AI in higher education?
What is the future of Gen AI in higher education?
What are hopeful futures for education, or ways it can go wrong?
What are hopeful futures for education, or ways it can go wrong?
What are hopeful futures for education, or ways it can go wrong?
What are hopeful futures for education, or ways it can go wrong?
Before the assembly
Engaging the community
We welcomed all MIT undergraduate and graduate students, as well as students in the Education graduate school at Harvard, to join our three day session. Students submitted their interest online, committed to all three days of collective deliberation, and then received an invitation to the assembly. They were offered an intellectual community and three days of meals.
promotional social media post
promotional poster
During the Assembly
Welcome and introduction
All students were greeted with personalized stickers generated by AI based on their values as a learner. We set a gentle, joyful tone and then introduced the students to three ideas:
Why a deliberative assembly here and now?
Why add tech to this process?
Why are we asking surveys and recording your conversations?
AI-generated stickers
Socializing and group-building
We set aside time for group-building activities to lighten the seriousness of the discussions as well as to ground participants in their bodies throughout the day. Sometimes this looked like shaking out each limb and sometimes it looked like imitating birds as we traveled around the MIT Media Lab!
During the Assembly
Identifying shared values
In order to frame this assembly as an opportunity to connect over values, not ideology, our first small-group breakout session consisted of a storytelling game. By identifying the values underlying participants’ views on the future of education, they were able to connect on a deeper level before incorporating fact-based arguments and ideologies to the deliberative process.
Analogia: translating personal stories into values
Analogia is a game for sharing honest and sometimes challenging stories using a combination of talking prompts and absurd images. The cards and images used in the game were specially generated by Cassie Lee to focus on questions of personal values and experiences in educational systems.
analogia image and prompt cards
During the Assembly
Learning
Students joined a panel discussion with experts, featuring Dr. Janet Rankin, director of teaching and learning at MIT, Dr. Jad Kabbara, research scientist at CCC, and Prerna Ravi, PhD student at the MIT Computer Science & Artificial Intelligence Laboratory (CSAIL).
Panelists (left to right) Prerna Ravi, PhD student at the MIT Computer Science & Artificial Intelligence Laboratory (CSAIL), Dr. Jad Kabbara, research scientist at CCC, and Dr. Janet Rankin, director of teaching and learning at MIT, answer student questions about the role of generative AI in higher education. Photo by Artemisia Luk.
Main takeaways
Our experts agreed on the inevitability that generative AI will have notable impacts on systems of higher education. They noted the need to train educators on how to update assignments such that students still learn and do not automatically generate essays or problem set answers from ChatGPT. They also agreed that there are numerous ethical issues involved in integrating AI into education; since AI cannot be simply banned, decision-makers must carefully regulate how AI tools are deployed.
Participant voices
During the Assembly
Deliberation
The deliberation stage of a deliberative assembly involves participants engaging in informed, respectful discussions to carefully consider various perspectives and evidence on an issue. The goal is to collaboratively reach a set of decisions or recommendations, each of which must be endorsed by at least 75% of the participants before it can be considered to “pass” into the final recommendations.
Final recommendations
Recommendations are written and formatted as the participants presented them. Edits for clarity are in brackets.
Equal opportunities
Scaffolding
Regulations
Meaningful learning experiences
Meaningful learning experiences
After the Assembly
Back tracking
To explore how technology could enhance the assembly process, Master’s student Michael Wong used AI to transcribe recorded small-group conversations from the assembly. He then employed a Large-Language Model (GPT-4) to identify key themes and tag each speaker's snippet with those themes. This allowed him to track how often important themes were discussed and use AI to trace how early discussions evolved into recommendations by mapping connections between themes across sessions. The width of each connection indicates how frequently linked themes were mentioned. The goal of such back tracing efforts is twofold: it allows those outside of the assembly process to peak inside the deliberation process without compromising participants’ privacy; it also allows participants to see the impact of their contributions in changing group opinions across the duration of the assembly, underscoring the value of each individual’s contributions and making them feel heard.
PhD student, Maggie Hughes, and CCC research scientist Brandon Roy, used data from this Student Assembly’s small group discussions to explore what a constructive conversation looks like. In their visuals, also known as “jellies,” Hughes and Roy make visible patterns within conversations in order to see dynamics between participants, the impact of facilitators, and -- hopefully -- even measure the quality of the conversation itself. The jellies are small network maps of each conversation, nodes being participants and comments passed between them. They hope that such visualizations will support conversation designers as they learn, reflect, and drive towards constructive communication.
Do small group conversations increase intra-community trust, psychological safety, and social tolerance?
PhD student Danny Kessler designed a set of surveys to be shared before and after the assembly as well as a few shorter surveys to be shared at the end of every small-group conversation. Results are promising, even from this small set of participants: they felt greater levels of trust, psychological safety, and tolerance for difference by the end of the assembly process. Future and longer assembly processes are expected to be even more impactful on participants’ worldviews.
Our participants came in as strangers and definitely left as thought partners, learning as much from each other as from the organizers, panelists, and readings.
.gif)
