AI in Research
The introduction of Generative AI has given rise to questions about how to incorporate it into research. Currently, I use AI in my workflow as a "research assistant" to speed up and complete tedious tasks more quickly and as a second set of eye while being in compliant with organizational AI policies. Below are examples of AI use cases.
AI Use Cases
1. Develop a Custom GPT Research Repository
-
Created a custom GPT to function as a research library for use by UX design and product stakeholders.
​
-
Objective: Make research reports more accessible to stakeholders and enhance storytelling with data. This GPT was replicated by researchers in other product spaces.
​
-
Included all discovery and evaluative research reports for the product space, research-created store associate personas, and business OKRs and KPIs.
​
-
Used to quickly and accurately answer stakeholders' research questions, summarize past research, identify themes and pain points across initiatives, make recommendations, tell data-grounded stories through associates' eyes and link all research to business OKRs and KPIs.
​​
-
Guardrails implemented:
-
New files cannot be uploaded by users​.
-
GPT can only create output related to it's intended purpose and denies other requests.
-
All direct quotes pulled from reports are cited and reported in full to trace accuracy and monitor for hallucinations.
-
All output includes in-text citations that include report file name and slide numbers, as well as a references section at the end of each output listing all reports.
-
Add caveat that outputs include historical data and may include now-outdated information current-state experiences.
-
Users are provided with the link to the main research repository to download reports and are instructed to direct questions to the UX team and researcher.
-
2. Analyze Transcripts to Identify High-Level Themes
-
Use ChatGPT to review transcripts from user interviews and contextual inquiry sessions to identify high-level finings based on research questions and objectives. This does not replace traditional analysis methods but it guides and speeds up the process. Additionally, it can be used to share interim research updates with stakeholders.
​​
-
Prompt Example: Transcripts are from contextual inquiry sessions with store associates on the fulfillment team. Fulfillment team associates are responsible for picking and fulfilling online and in-store pickup and delivery orders and use the fulfillment app to perform job functions. Identify pain points associates have with the app, what strategies associates use to pick and stage orders efficiently and include app improvement recommendations.
1. Review Usability/Concept Testing Scripts
-
After completing an initial draft of a usability testing script, I upload it into ChatGPT to review. I use it to identify leading questions, suggest re-worded questions, identify gaps in questions being asked and for feedback on overall script flow.
​​
-
Prompt Example: This is a usability testing script. The goal of testing is to identify pain points with [a new feature]. Review overall script flow, provide question suggestions, identify leading questions and re-wording suggestions for leading questions.