Three types of synthetic data: fully synthetic (AI-generated from scratch), partially synthetic (AI-generated but stripped of identifiers), and augmented synthetic research (AI expands real interviews to scale them up). Augmented synthetic preserves the original research's accuracy while multiplying its size without sacrificing privacy.
So there's fully synthetic data in this case, which is, again, as I mentioned, completely made up by the AI, basically following specific guidelines, but it's completely AI-generated data. Then there's partially, partial synthetic data, which is basically AI-generated data that closely mirrors your real data, but basically excludes any sensitive information like names or specific the information that you want to exclude. Then there's the most important, I think, most undervalued one, which is augmented synthetic research, which is conduct your research, your human research first, or at least a basis for it, and then expand it using, using AI. So you basically get the same kind of accuracy, the same directional power and effectiveness, but at scale, because let's say that I can interview 5 people. With synthetic AI data, I can basically turn those 5 interviews into as if I did 20 interviews because AI is able to spot patterns, expand them, replicate them, multiply them, and so you get a lot more data that can be actually useful and accurate at the same time.
