Conducted in-depth research on three projects and practitioners working in cognitive assessment: ACE-X (Adaptive Cognitive Evaluation), the NIH Toolbox, and José Azuara's work bridging assessment design and user engagement. Also reorganized the repo structure with professor guidance to support M2+ workflow requirements.
AI was used for two things: providing an initial research lead, and reformatting handwritten notes into clean markdown. All analysis and conclusions were mine.
Initial Research: Asked AI to surface a starting point for research direction. From that lead, gathered all information independently — reading documentation, taking notes by hand, writing observations in the blog.
Used AI to restructure handwritten notes into clean markdown — adding headers, bold emphasis, visual hierarchy. AI did not reorganize or rewrite the content; it only enhanced presentation. This is the right use: AI as a formatting tool, not a thinking tool. The analysis stayed mine.
AI offered to synthesize and summarize the three sources after I shared my notes. I declined — I wanted the raw research voice preserved, not polished into AI's summary style. When AI summarizes research, it tends to smooth over the contradictions and tensions that are actually the most interesting parts. I kept those rough edges.
The three sources taught me a core tension in assessment design: medical credibility vs. user engagement.
ACE-X prioritized engagement and accessibility — adaptive, fun, trackable over time — but risked feeling too "game-like" for clinical adoption. NIH Toolbox prioritized credibility and efficiency — professional, structured, quick — but lacked motivational hooks that keep users engaged. José Azuara's work showed it's possible to bridge both, but with hard constraints: FDA approval locks your methodology, medical validation means building within a rigid framework.
Key insight: I can't optimize for everything. My product needs deliberate tradeoffs. The research clarified what those tradeoffs actually are and what they cost.
The research showed what's already possible when you combine assessment design with technology — and what's still missing. The gap: between "valid assessment" and "something people actually want to use." Baseline sits in that gap. The "NOT diagnostic" boundary also became clear here — Azuara's work showed that claiming clinical validity without the full FDA/validation infrastructure is the wrong move for a student project with a quarter timeline.
Identify the specific problems the prototypes can solve that existing tools don't address well. Then develop the Position Statement for M2, which will lock in the stance on medical validity vs. engagement, and define what will and won't be compromised.