Conducting user persona research with an AI
“Empathize” is the earliest stage of the Design Thinking process, which for those unfamiliar, is a non-linear, iterative process by which to create and continually improve upon end products that will likely result in successful adoption by their target audience. In other words, how do you make a product or service in such a way that people will actually appreciate it and not just how the designers or managers predict they will?
There have been many famous instances of corporations losing millions due to new products turning out to be complete flops. For example Google Glass, New Coke, Apple Newton, or Microsoft Vista. In most of these cases, these companies could have avoided financial disaster if they had spent more time in the earlier user research phases. This among other reasons is why professionals in the fields of User Experience and Design Thinking have become hot commodities for most well-informed companies.
However, conducting user research is often logistically complex, time-consuming, and a bit socially illusive due to a variety of unknown or unpredictable variables, mostly social/emotional. It’s conducted in a very similar fashion to most broad scientific studies involving people.
Typically, a precursor to the interactive portion of user research is “Desk Research” in which a UX researcher finds out through, you guessed it, sitting at a desk and researching a particular demographic. What I’ll be proposing in this article is appending an intermediate stage in which researchers use a new linguistic interface to the Internet in the form of an advanced AI chatbot, GPT-3, to conduct virtual interviews. Let me explain…
In previous articles, I’ve written about having text-based conversations with programmable AI chatbots that assemble very human-like responses fed from a massive sample of text data. In other cases, I’ve interviewed replicas of Carl Jung and Don Norman. Although there are some areas of logic or validity that are a bit questionable, the overall experience has been fascinating and insightful.
So, if we apply the same techniques but plug in a set of demographic information particular to our target audience and proceed to ask it some questions, could that lead to a more fleshed out idea of that audience prior to bringing in the human counterparts?
Below I attempt to answer that with a template (that you can feel free to use) and some follow up questions. Granted, you may not be able to copy/paste the template verbatim, but you can use the general premise to guide the research. I will likely follow this article with a more hands-on tutorial for how to conduct this virtual interview.
The initial persona programming:
#create a list of variable to plug in to the bracketed words below
#plug in the ones you know and have it figure out the rest.I am a [age]-year-old [gender] who lives with [familyMembers] in [cityName] in a [housingType].
My favorite hobbies are [hobby1], [hobby2], and [hobby3].
In school I get mostly between [gradeLow] and [gradeHigh], and my favorite subject is [subject].
My favorite kind of food is [food] and I love going to [restaurant] to eat.
I make about $[monthlyIncome] per month as a [job].
I would be willing to spend up to [spendHigh] on $[product] upfront and up to $[monthly] per month.
When I graduate, I’d love to [career] for a living.
One day I hope to be [relationshipStatus] to a [gender] and have [kids] kids.
You may not have all of the variables figured out, but since GPT-3 is essentially an auto-complete engine based on probability, you can feed it some of the completed sentences and then have it auto-complete the sentences with the variables you are unsure about.
The Semi-structured interview portion:
Once you’ve established some of the basics, you can then proceed with the questions that you are curious about.
In user research, the three most popular forms of interviews are Structured, Unstructured, and Semi-structured. The style I will be conducting is Semi-structured. This is similar to a talk show host interviewing a guest. They may have some initial questions on their queue cards or teleprompters, but they will also improvise follow up questions based on the guest’s responses.
Since I’m focused on the field of education, I’m most interested in building products or apps for parents wanting better educational opportunities for their kids, students who are looking for alternative ways of learning, or teachers who want to better reach their students.
Some of the initial sample questions I used came from some of the research tools provided by Panorama Education, a company focused on improving students’ sense of belonging, behavior, and academic performance. Once I got the ball rolling, I came up with my own followup questions.
The following contains excerpts from a conversation with an AI based on a 16-year-old American male Generation Z student living in California. My goal is to find out how to improve this student’s experience at school by finding out what their main challenges and goals are.
As a reminder, the responses this AI gives are based probabilistically on speech patterns, interests, and behaviors of someone fitting in this demographic found on social media, research data, and other writings.
When I asked the bot to give itself a name, it chose “Zach”.
Jonathan Earley: How sure are you that you can complete all the work that is assigned in your class?
Zach: I’d like to believe that I can get good grades in school.
JE: What’s something that most adults don’t get about you?
Zach: I think adults don’t understand how I feel about the world.
JE: And how do you feel about the world?
Zach: I believe that we should all be united.
JE: What can teachers do to help students become more united?
Zach: Teachers can help students understand about LGBT and other diverse people.
JE: Should this be a teacher’s role? Or should this be up to the students to decide?
Zach: I want teachers to help me better understand the issues associated with equality and diversity.
JE: What is the best way for a teacher to do that?
Zach: Teachers should make me aware of these issues, and they should be willing to have tough conversations with me.
JE: Do you think it’s ok for a teacher to bring their own opinions into the conversation? What if a teacher is not in favor of a certain group, religion, or lifestyle?
Zach: I think a teacher should make their students aware of different values and opinions by creating a safe environment for us to explore different points of views.
JE: What do you find most difficult about school?
Zach: I find that being with other people is the most difficult thing about school.
JE: What is so difficult about people?
Zach: Sometimes it’s intimidating to talk with other people. I feel bad if they’re offended by something I said, and I can’t talk to everyone.
JE: If a school offered the choice between these two things, which would you choose and why?
1. How to better get along with people or 2. How to score higher on the SAT
Zach: How to get along with people is what I would choose because I want to make friends and fit in.
JE: What is so important about fitting in?
Zach: I want to feel like I can hang out with a lot of people in my age group and not be uncomfortable.
JE: How do you think this might improve your life after you graduate?
Zach: If I can make friends in school, it makes it easier for me to find friends after I graduate.
JE: Do you think it would be important for you to be friends with whomever you worked with after school?
Zach: Yeah, at work it would be important for me to have good relationships with my coworkers.
JE: How do you feel about AI taking over most of the jobs that have traditionally been done by humans?
Zach: I feel that there may be a problem if all the jobs are taken by AI. I’m concerned that AI is becoming too efficient, and humans won’t have anything useful to do anymore.
JE: What should schools be doing to prepare you for if this happens?
Zach: I think schools should teach us how to be creative and how to think of new solutions to problems.
JE: What kinds of problems do you think you’ll be solving in the future?
Zach: I think I’ll be solving environmental problems and trying to find ways to make the world a better place.
Use this as a guide, not the end result
Obviously, this was not a real student, and even if it were, not every student would give the same answers. However, what I’ve found when checking in with my own students, especially in surveys is that they often don’t know how to express what they want or feel that what they want is socially acceptable to say out loud. When I present students with multiple choice or yes/no questions, they’re often much better able to express themselves.
This initial test can serve as a scaffolding that you could then use to ask more quantitative questions versus open-ended ones.
I hope you found this article intriguing and I’d love to read your thoughts in the comments.
Take care.
