Session 1 — Grounded Reasoning: Assessment in the Age of AI
Session 1 — Grounded Reasoning: Assessment in the Age of AI ✅
July 15, 2026 · COMPLETED
🎯 My Challenge Statement (Final)
Translate your challenge into a concrete solution. What does CPA actually look like in practice? Google's own session theme turned out to be the same problem CPA solves — see the recap below.
What was prepared
- Complete Digital Workbook from Kick-off
- Post both Community Hub items (Challenges thread + Prep thread)
- Build out the CPA framework with rubrics and evidence types
- Prepare 90-second pitch
Session Recap: Grounded Reasoning & REVEAL
Grounded Reasoning = factual grounding + logical reasoning together. Google framed the "Assessment Crisis" as the direct parallel to AI hallucinations: outputs that are plausible but built on fabricated knowledge, from students cognitive-offloading — essentially Google's own words for the CPA problem statement.
REVEAL (Google Research/DeepMind) is a public benchmark dataset that grades an AI's chain of thought sentence by sentence rather than judging only the final answer, via a Fact Check (Fully Supported / Partially Supported / Contradictory-Hallucination / Unsupported) and a Logic Check (does the conclusion validly follow, even assuming prior steps are true). It's a research dataset, not a plug-and-play classroom tool, but its labeling schema is directly citable prior art for CPA.
Reframe worth adopting: moving from "AI Policing" to "Thinking Design" — make thinking visible, build AI-resistant assessment, teach AI with intention not deference, balance efficiency with cognitive resilience.
Two breakout questions Google itself posed, which CPA directly answers: "How do students document their intellectual history?" and "Is asking to see prompt history a creative infringement?"
Citation: Jacovi, A., Bitton, Y., Bohnet, B., Herzig, J., Honovich, O., Tseng, M., Collins, M., Aharoni, R., & Geva, M. (2024). A Chain-of-Thought Is as Strong as Its Weakest Link: A Benchmark for Verifiers of Reasoning Chains. Proceedings of ACL 2024. arXiv:2402.00559. doi.org/10.48550/arXiv.2402.00559 · Dataset: huggingface.co/datasets/google/reveal
CPA Framework: What Gets Assessed
| Dimension | What it measures | Evidence | Tools |
|---|---|---|---|
| Prompt Architecture | Strategic intent, vocabulary depth, and contextual framing behind AI instructions. | Raw prompt logs, systematic variable testing, and multi-turn iteration history. | Google Stitch / Gemini |
| Orchestration Logic | How the student chains multiple AI tools together into an end-to-end design workflow. | Workflow maps, tool-handoff documentation, and architectural decision rationales. | Gemini Notebook |
| Strategic & Content Development | The student's critical thinking in making and justifying strategic content decisions — feature scope, content priorities, and value proposition — grounded in research findings rather than default or generic choices. | Feature-scope prioritization frameworks, content/IA decision logs, and value-proposition rationale documents. | Gemini Notebook / Gemini |
| Critical Curation | The student's ability to evaluate, stress-test, and refine raw AI outputs against constraints — naming each catch as an Attribution Error (an unsupported or fabricated claim) or a Logical Error (a flawed deduction, even from an otherwise-true premise). | Revision trails, error-tracking sheets, source-verification/fact-check logs, and written critiques of AI-generated content. | Gemini |
| Creative Art Direction | Taste, aesthetic consistency, and unique conceptual choices that override default AI biases. | Annotated style matrices, mood boards, and mood-to-asset translation logs. | Google Stitch |
| Domain Specific Knowledge Application | Deep visual communication craft — typography, hierarchy, color, and compositional judgment — applied skillfully enough to elevate AI-assisted output into original, intentional design work that reads as more than generic Gen AI. | Typography and hierarchy exploration notes, color/composition rationale, design system documentation, and before/after visual craft comparisons showing the shift from generic AI output to an original visual direction. | Google Stitch, Figma |
| Reflective Reasoning | The student's meta-cognitive ability to explain why they made specific curatorial choices — framed as a legible reasoning chain, where each step should be independently defensible, not just the final conclusion. | Timestamped process notes and file annotations with rationale, and a brand- and persona-driven reflective approach. | Gemini Notebook / Gemini Gems |
| Direct Creative Contribution | Human authorship: what the student sketches, develops into original artifacts or moodboards, writes, or fundamentally transforms. | Original source files and before/after authoring comparisons showing hand-finished or transformed work. | Gemini Notebook, Google Stitch, Figma |