Make ChatGPT Your Research Assistant | ChatGPT for Students - Software Development PDF Download

Introduction

Qualitative user research goes through multiple phases and ChatGPT can assist you with each step.

ChatGPT is not a replacement for human judgement, expertise, or skills. A person conducting user research still needs to understand how to perform their job, but ChatGPT can help streamline the process.

Planning

Clarify research objectives.

PROMPT: Pretend you are a user researcher assistant. I'm going to give you a research objective and I'd like you to help me clarify it.

The software is for large spas. This research is focused on helping optimize revenue by ensuring maximal bookings for each room and service provider. It is discovery research and we'd like to understand how this happens today, where the pain points are, and what goes well today.

OUTPUT

I wasn't looking for research questions yet, but ChatGPT provided 5 questions as part of the answer. Even though it's not quite what I asked for, it's useful!

If you don't get what you want, don't be afraid to ask again.

PROMPT: i'm asking you to rewrite the research goal
At the planning stage, ChatGPT can create a draft of your research plan, design a survey, generate research questions, check it for typos, create research questions, and more.

Recruitment

While ChatGPT can't email people or make phone calls, it can create screeners or help make sure that your screener is on point.

Screeners are critical to ensure that you're getting the right people in your research. A bad screener for 1 : 1 methods can waste your time and can cost money if there is a gift or payment to the participant.

Even if you're great at writing screeners, getting ChatGPT to take a look at what you already have and asking for feedback, can help you get it dialed in more tightly.

PROMPT: Write a screener for the person who takes the spa booking

Having previously worked with spa software and revenue optimization, this is a great start to a screener.

Data collection

Collecting research data isn't a core strength of AI/ML yet.

There are some technologies coming soon for real time translation that could facilitate researchers fluently communicating with participants when there is no common fluent language. Microsoft Teams has already launched this technology but having tried it, it's not ready for critical business use. (It can create some eyebrow raising translations!)

Data analysis

Data analysis of 1 : 1 interviews or testing sessions requires reading your notes, other's notes, reading through transcripts, re-watching recordings, identifying themes and patterns in the data, looking for emotional indicators, collecting quotes, and more.

It is the most time-consuming part of interview user research methods and will typically take 3x longer than the interview itself.

It's also the area where AI/ML will have the biggest impact, speeding up the time needed for generating actionable insights. It can summarize transcripts, find recurrent themes, and blur the boundaries of deductive vs inductive tagging, since it can assist in identifying the appropriate tags.

Summarization

This is one of the most exciting aspects of emergent technologies and user research. Analysis of large amounts of data can be challenging. Tagging and coding data is one of the most challenging parts of user research and it's time consuming. AI/ML, including but not limited to ChatGPT, excels at finding patterns in data and can create a label for that cluster of data that is meaningful to humans.

You can feed an AI/ML UI a transcript of data and ask it to pull out and categorize insights. ChatGPT 3.5, (the free version as of April 2023), can only process around 800-1000 words at a time. A 60 minute user research transcript will vastly exceed that capacity and adding in multiple transcripts is out of the question.

ChatGPT4 reportedly can process much more, though there is conflicting information online about where the boundaries are - 3,000 words to 25,000 words. The upper end may be suitable for processing multiple transcripts.

Workaround to input limitation

PROMPT: I'm going to give you multiple sections of a transcript from a user research session. Summarize the insights for me for each section. I will then give you the next section until we are done.

Coding data

Finding patterns in the data is one of the most time-consuming tasks of user research.

There are some research repository products that help make it easier, including tagging while looking at a transcript.

But if you pull everything into rows in a spreadsheet, you can copy the cells to ChatGPT and get it to help tag data and summarize themes.

Tagging

Capture what seems important to you in a spreadsheet. Feed the spreadsheet (copy-paste works) into ChatGPT and ask for assistance coding and tagging.

PROMPT: I'm going to paste a spreadsheet of data to you. I'd like you to identify patterns in the data and tag each row with a relevant label.

Finding themes

PROMPT: I'm going to paste a spreadsheet of data to you. I'd like you to identify patterns in the data and surface the themes that are most important.

Reporting

ChatGPT can help you with grammar, extend something you've written, summarize what you've written, and also help with identifying actionable steps.

There are other tools that integrate with Microsoft Office, (Microsoft Copilot), and Google Suite that can help do similar work in the context of your final document.