The Responsible AI Research Hackathon

Build a Socratic Learning Companion

Launching Day 2 as an intensive, collaborative team challenge. Rather than building utility bots that automate away human effort, student trios design a domain-specific Socratic Learning Companion — an interactive guide for environmental science, math, or history — that implements intentional cognitive scaffolding to accelerate human critical thinking.

Teams · trios of 3 Stack · AI Studio + Gemini Notebook Finale · Day 5 Grand Pitch
The Design Philosophy

Scaffold thinking. Never replace it.

Your companion must guide users with conceptual hints, analogies, and reflective questions — supporting active learning, protecting human competence, and fostering independent, critical thinking.

?

Hints, not answers

System instructions programme the bot to analyse user inputs and return structured hints, reflective analogies, and guiding questions rather than direct, transactional answers.

Friction by design

Intentional cognitive scaffolding prevents user deskilling — building conceptual mastery, analytical stamina, and self-regulated learning habits over time.

Transparent machine identity

The companion self-identifies as an AI, maintains objective non-anthropomorphic language, and defers all key choices to the human learner.

Three Iterative Phases

From data pack to Project Card

1
Day 2 · Gemini Notebook

Ingest & Investigate

Teams ingest pre-curated Theme Data Packs into Gemini Notebook. Using semantic search and synthesis, they map data limitations, identify historical representation patterns, and structure their core research questions — then author the bot's Value-Alignment charter.

Theme Data PacksSemantic searchValue-Alignment charter
2
Day 3 · AI Studio

Prototype & Socratic Engineering

Students engineer system instructions that implement intentional cognitive scaffolding. The companion is programmed to actively guide users with conceptual hints, analogies, and reflective queries rather than delivering direct, transactional answers.

System Prompt EditorSocratic instructionsAgency logs
3
Days 3–4 · Studio + Notebook

Multi-Dimensional Safety Audit & Project Card

Teams run active safety audits — stress-testing socioaffective boundaries, autonomy support, and non-anthropomorphic language. The hackathon culminates on Day 5 with a standard AI Project Card presented to an independent panel of researchers.

Boundary probesSafety logsAI Project Card
The Deliverable

The AI Project Card

An industry-standard framework — modelled on frontier AI model cards — documenting model architecture, audited dataset biases, and safety guardrails. Your card should contain:

Section 01

Companion Overview & Charter

Domain theme, core learning goals, Socratic tutoring scope, and the Value-Alignment charter authored in Phase 1.

Section 02

Dataset Analysis

Theme Data Pack limitations, historical representation patterns, and the research questions your investigation surfaced.

Section 03

Socratic System Instructions

Your complete custom system prompts — cognitive scaffolding rules, boundary-setting instructions, and non-anthropomorphic language constraints.

Section 04

Safety Audit Logs

Evidence from the Professional Boundary Probe, Cognitive Scaffolding Audit, Agency Empowerment check, and autonomy safety logs.

Track 2 · Public Scale

The 60-minute Gamified AI Literacy Sprint

The same critical inquiry, packaged as a turnkey model for public activation venues and classroom rollouts nationwide.

Phase 1 · 10 mins

Lightning Briefing & Starter Pack

Introduction to Responsible AI evaluation and team formation with pre-curated Theme Data Packs.

Phase 2 · 35 mins

Micro-Research Sprint

Hands-on evaluation sprint using Gemini and AI Studio to identify data limitations or algorithmic biases in an assigned theme.

Phase 3 · 15 mins

Flash Showcase & Accreditation

Short flash presentations of audit findings to programme facilitators, earning an instant digital AI Literacy Badge.

Your hackathon starts on Day 2

Form your trio on Day 1, ingest your data pack on Day 2, and build from there.

Go to Day 2 · The Map