NNGroup Using AI for UX Research + Designer Competencies in the AI Era (12 Articles · Accelerating Research / Planning Research / Generating Mock Tables / Red Lines for Synthetic Users / Evaluating AI-Simulated Behavior / Methodological Pitfalls / How GenAI Works / Redefining Design Skills / The Return of Generalists / AI Literacy)
12 curated articles from Nielsen Norman Group’s “Artificial Intelligence” special topic, organized around two main threads. ① Using AI to assist UX research: Across the four research stages (Planning / Conducting / Analyzing / Reporting), AI is strongest at text processing (strongly recommended for analysis; recommended for planning and reporting) and weakest at behavioral observation (use with caution / not usable for conducting, as AI cannot truly observe behavior); a 5-step method for writing research plans with generative AI (do not ask for a complete plan in one sentence; break it into background / questions / methods / screening criteria / materials); 7 key strategies for using AI to generate realistic mock tables and charts for prototypes (context / data volume / sorting / attribute formatting / generating tables from charts / distribution control / Python validation), plus 7 strategies for charts, and differences between Claude 3.5 Sonnet and ChatGPT 4. ② Red lines for Synthetic Users: AI-generated virtual users can only be used for desk research or as hypotheses, and must never replace real user research (they exhibit sycophancy, idealize experiences, lack prioritization of needs, and produce unreliable imagined experiences); NN/g evaluated three academic studies (survey-based digital twins achieved 78% accuracy on missing data / 67% on new questions, r=0.98 vs 0.68; digital twins based on two-hour interviews achieved GSS 0.85 / Big Five personality 0.80 / economic game 0.66 across 1052 people; Neeraj Arora’s 605 synthetic users for refrigerated hot dogs only showed trends and lacked diversity), interview-based models have 36%–62% lower bias than demographic models, and more accurate predictions for white / high-socioeconomic groups raise fairness concerns; methodological pitfalls hidden in research tools (setting only one success URL misjudges successful users, analytics tools cannot see behavior, conflating evaluative and exploratory research in discussion guides, AI-generated tasks leading users). ③ How GenAI works (text fill-in probability prediction / neural networks / Word Embeddings / Transformer self-attention; do not treat it as a real person or a memory database). ④ Designers in the AI era: 5 principles for redefining design skills (keep strategic thinking in-house and outsource tactics / balance trust and scrutiny / design for both humans and AI Agents / embrace team augmentation / focus on inequality), the return of generalists (AI raises the bar for “knowing what to do,” making the horizontal bar of the T-shape more valuable), AI literacy = Prompt Fluency × Output Literacy (naive experts are the most dangerous). Audience: product managers; terms explained in industry language; decision-oriented (when to use / how to choose / counterexamples).