We have found that leveraging AI can improve live research and identify nuanced patterns and connections that live below the surface. Perhaps most valuable, AI serves as a dynamic activation catalyst for businesses, transforming deliverables into a platform that offers several ways to understand and interact with data.

Our last post discussed how AI can enable better analysis and in-survey outputs.  In this post, we cover how you can take AI even further, by using structured and unstructured data to unlock new insights by querying AI-generated versions of users or segments. Specifically, we will focus on AI personas and synthetic data.

Have you ever finished a study and wished you could have conducted a series of qualitative interviews with a subset of survey participants? Or have budgetary, time, or feasibility limitations that have forced you to skip or curtail research?

AI personas can address these issues, enabling researchers to interact with virtual representations of research participants, allowing for virtual IDIs with digital personas that mirror the style, language, attitudes, and stated behavior of the real-world participants they’re based upon. These algorithmically-created “digital twins” give researchers the ability to conduct deep exploration without limits on time, the amount of covered content, or number of participants.  Conversations can be conducted on a respondent level or among specific segments within the target audience.

Although it’s important to note results are directional in nature, there are a wide range of benefits in working with AI personas:

• It allows for deeper audience exploration among hard-to-reach, sensitive, or expensive populations (e.g., patients, B2B professionals).
• There are no time constraints – whether the length of a conversation or when the interview is conducted
• It is less expensive in the long run – no incentives, scheduling, and management required, and sample sizes can be far larger than in traditional qualitative approaches.
• They can work as an early-stage test, surfacing early issues, barriers, or “non-starters” before conducting more thorough research

It’s not uncommon to present data and have someone ask, “Did you ask this question?”, where no is the answer.  This is one of many situations that can be addressed through synthetic data, a valuable way to conduct or augment a quantitative study with questions answered via an algorithm trained on one or multiple data sources.

This includes:

• Asking questions that are deemed secondary objectives and initially cut due to survey length concerns
• Directionally understanding potential responses to new questions which arise based on study results
• Address sensitive topics that may be difficult to handle with human survey participants.
• Rapidly iterate/experiment before conducting more expensive real-world research.
• Acting as an early-stage idea screen – vetting ideas to eliminate those which likely do not show any potential for adoption or usage by the target audience

It’s important to note that the development of synthetic data is still at an early stage, and while it shows real potential and is rapidly improving, researchers must be cognizant of its limitations when communicating the data it produces.

Digital personas and synthetic data can become even more powerful when incorporating multiple data sources – for example, customer survey data paired with sales/behavioral information, passive digital behavior, reviews, or qualitative inputs.

Russell Research has been actively investing in, developing, and testing AI applications for custom research, and are currently using/employing AI in some form throughout the entirety of the research process.  We’ve found that incorporating artificial intelligence across the process represents a remarkable opportunity to improve quality and more deeply understand target populations.

Be on the lookout for the next installment of this series where we’ll provide more details on how to effectively work with AI and to understand its current limitations.