What is Cognitive-Process Assessment?
What is Cognitive-Process Assessment (CPA)?
Traditional assessment grades the artifact — the final UI, the prototype, the delivered design.
CPA grades how you thought to make it — the decisions, judgment calls, and domain expertise applied along the way. It answers the critical question: "How do we prove a student actually knows what they're doing when AI can produce the same-looking output in seconds?"
The core tension: A generalist can prompt Stitch and get a polished-looking UI in seconds. What generic prompting can't do is direct that process with expertise — iterating with intention across research, content, wireframes, and UI, catching what the AI gets wrong about interaction patterns, cognitive load, accessibility, component scalability. That expert direction, visible across every iteration, is the intellectual fingerprint. CPA makes it visible and assessable.
Key Terms
- Intellectual fingerprint
- The unique cognitive signature of a student's creative decision-making
- Cognitive-Process Assessment (CPA)
- Rubrics that grade thinking, not output
- Domain specific knowledge
- Expert UX and UI judgment applied at every stage of the AI-assisted process — the difference between generic prompting and expert direction
- UI Design
- The visual and interaction layer of the process, where a student's judgment about hierarchy, interaction patterns, and accessibility becomes visible and assessable, not just the polish of the screen itself
- Logic-first creative economy
- The professional landscape where orchestration = value
- Human Creative Art Direction
- The student's role as art director, copywriter, and critical-thinking decision-maker, directing AI output and contributing original work; together they transform the process into protected IP
- Grounded Reasoning
- Google's term (Session 1) for factual grounding + logical reasoning together — the technical layer underneath "intellectual fingerprint"
- Attribution Error vs. Logical Error
- From Google's REVEAL framework: an Attribution Error is an unsupported or fabricated claim; a Logical Error is a flawed deduction even from true premises. CPA uses this same split to grade why an AI-assisted decision succeeded or failed
Academic Grounding
Google Research/DeepMind's REVEAL benchmark grades an AI's chain of thought sentence by sentence rather than judging only the final answer — via a Fact Check and a Logic Check on every step. CPA applies that same principle to a student's design process instead of an AI's reasoning chain: the final screen isn't the reliable signal, the reasoning chain that produced it is.
Google's own fellowship staff described the opportunity in academic integrity as "shifting metrics from product to process" — nearly a direct restatement of the CPA thesis, and a useful anchor quote for institutional audiences.
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