As you write your interview guide, you have the opportunity to outline “probing instructions” to your AI moderator. These are instructions on how to follow up on your initial question and make sure you get the most out of the conversation with your participant.
Well-crafted prompts ensure meaningful follow-ups and deeper insights, but they do require some trial and error. Here are some best practices our team has developed.
1. Treat the AI like your intern
Use the probing instructions to write directions you would want to give any moderator. The goal of these probing instructions is to direct the AI moderator just like you would with a human moderator.
While you can write probing instructions as follow-up questions, we find that they’re most effective when written as explicit directions to the AI.
✅ "Probe deeper into the reasons behind their answer”
❌ “Can you tell me more about why you feel this way?”
2. Update the "amount of probing" according to your needs
The “Amount of probing” option has options for “None”, “Standard,” or “Deep exploration.”
For best results, make sure to match the depth of probing to the type of probing you need. If you have more in-depth probing instructions and also want the AI moderator to potentially ask other unscripted follow-ups on this question, you’ll want to select “Deep exploration.”
3. Keep it concise and specific
Avoid long paragraphs. Instead, use one or two short follow-up questions or instructions. Clarity in wording matters, as this is direction to the AI moderator - keep the instructions clear, concise, and specific.
✅ "Find out why the respondent thinks this is important."
❌ "Can you ask them to explain their reasoning in detail about why they made this choice and how it relates to their past experiences, and whether this has changed over time?"
4. Keep the topic focused with your probing
Make sure to keep the question and the probing instructions focused on the same topics/ themes. Having unrelated follow-ups within the same question can make your analysis more complex after interviews are completed, as participants will be responding to different themes within the same question.
For example, let’s say you’re testing a concept -
Guide question: What are your thoughts on the branding of this product? [Image of product displayed]
Probing instructions:
✅ “Follow up to find out their thoughts on the design and colors used, and whether they think it matches their image of the brand.”
❌ “Follow up to find out their thoughts on the design and colors used, and what is the maximum price they would be willing to pay, and how frequently they would buy it”
In the example above, the good probing instructions have two simple follow-ups, and the follow-ups are related to the main question.
In the bad probing instructions, there are too many follow-ups listed, and the topic of willingness to pay/ frequency of purchase are not related to the guide question of branding. These pricing and frequency of purchase questions would be better as their own guide instructions rather than probing follow-ups.
By following these best practices, you can ensure your AI moderator effectively probes responses for richer insights, leading to more valuable data.