Today it's hard to avoid AI, and at FlexMR we’re no different, especially as this summer we launched our second generative AI tool into the InsightHub in the form of PromptAI.
PromptAI is a new question type that allows researchers to follow up with intelligent probing of verbatim responses. Prompts are dynamically generated based on participant answers but guided by the constraints that researchers specify. This offers a balanced solution that helps insight teams gain more rich and expressive qualitative data from surveys, whilst offering participants a personalised experience.
Suffice it to say, I was keen to get my hands on this new tool, have a bit of a play and explore of what it could do and see how this new tool might affect the ways in which I approach the design and analysis of surveys. So, I thought I would do a little experiment, to get a better idea of how PromptAI works; how it constructs its follow-up questions and what sort of information we could then hope to get from using it. For the topic of experimentation, I'm asking PromptAI to explore the concept of Customer Salience.
| Tweet This | |
| AI is everywhere, and FlexMR is no exception. Nikki Foster uses FlexMR's PromptAI to test its capabilities and explore more about Customer Salience. |
One of the main things, that, as a researcher I have to get my head around when it comes to PromptAI is it’s not just about how you ask the participant the question. That is part of it, but you also have to write the background information, otherwise known as the question objectives. And it is this, along with the question text, that informs the AI’s follow-up question.
For this experiment, I have kept the question objectives identical for each reiteration: ‘we want to understand what market researchers understand by the term customer salience, what it means to them and how it affects the work they do.’
After this was established, I then asked my first question:
Mock participant answer: "Businesses keeping the customer in mind when making decisions."
AI: "That's a good definition. As a market researcher, how do you define customer salience and how does it impact the work you do?"
I have kept this first question as open as possible, to see how PromptAI responds when it has more information than the participant. Here the participant has answered the question I asked them succinctly, however, they haven’t addressed some of the question objectives, particularly around how Customer Salience affects the work they do. Therefore, the PromptAI’s follow-up question is constructed to prompt the participant to answer the question objectives, directing the participant to answer in their capacity as a market researcher, and requesting them to answer in a more detailed and specific way, whilst ensuring that the question’s objectives are still achieved.
I designed the second question to be more specific, providing the definition of Customer Salience given in the first question, whilst also ensuring that all of the question objectives were covered in the initial question to the participant.
Here, the PromptAI has recognised that the participant has answered the initial part of the question, however, although they have touched on ‘what it means to their work’ their answer is still somewhat vague. PromptAI’s focus on the ‘how’ is encouraging the participant to provide a more actionable and practical response. The result of this is that we learn how we can start to help embed Customer Salience into our work as market researchers; by using empathy. A skill which we can focus on building and making part of our day-to-day practice.
Using the PromptAI in this way creates answers which are not only on topic, but are more specific, more relatable and when it comes to using this for research purposes; more actionable.
For me, one of the challenges with an open text question, especially those where you want the participant to focus on a specific topic or task, is to ensure that you are providing enough information for the participant to understand what you want them to talk about, without, leading them. Therefore, this question is designed to elicit as practical and actionable a response as possible.
Here the participant has fully answered the question put to them, in a practical and actionable way, however, as the question objectives have remained the same since the first question, the PromptAI’s question is then modelled on these prompting a second response from the participant which presents the more holistic view of Customer Salience that Questions 1 and 2 were getting.
| Tweet This | |
| Understanding where AI tools like PromptAI excel and how to ensure they produce quality results is key to employing them correctly in research experiences. |
Given this research, I can say that working with PromptAI is great at ensuring that open-text questions are answered fully by participants. Where it excels is when the question text fully reflects the question objectives so that it can construct its follow-up question based on the participant's answer; probing deeper into the how what and why of the question, based on any gaps which the specific participant may have left.
And in terms of using it to explore the idea of Customer Salience, well:
Customer Salience is businesses keeping the customer in mind when making decisions. It means putting customers at the heart of our research and championing the value of insights with stakeholders across businesses so that they in turn become spokespeople for the value of insights.
It affects the way we plan, design and conduct research projects to ensure the customer is the central consideration; focusing on what customers want and using empathy to ensure we keep the customer in mind when conducting research, tailoring research questions in a way that will provide insights into understanding customers' wants and needs. Considering the customer at the design phase, ensuring that the research is designed in a way to be accessible, relevant and allows the customers responses to best reflect their opinions. This means the business gets the most accurate view of customer opinion and the relevant, actionable insights can be developed.