In the rapidly evolving landscape of insights and strategy, the allure of “Synthetic Data” as a cure-all for meeting the demands of agility and cost-effectiveness is undeniable. The promise of lightning-fast data processing, pattern recognition, and uncovering insights has positioned AI as the superhero of the data-driven era.
Qualitative insight has traditionally relied on recruiting flesh and blood respondents to participate in methods such as group discussions, participant observations and interviews. Across 2023, there has been an emergence of synthetic respondents, which are essentially consumer personas created by computer models derived by understanding their attitudinal and behaviour patterns.
Synthetic data might sacrifice accuracy at the altar of cheapness or efficiency, but there’s no doubt that there are several use cases across our industry. This is especially urgent in the attention economy where the need for agility is critical for marketers to adapt to ever-shifting consumer preferences and behaviours.
However, amidst the excitement surrounding AI’s capabilities, it’s essential to surface some criticisms and watchouts associated with relying solely on generative AI for insights.
Firstly, human responses to external stimuli, be it a simple advert or complex phenomena such as Joy are deeply influenced by cultural factors. To flesh and blood respondents these cultural influences remain embedded in their being and tend to be impossible to codify. Thus, with synthetic respondents, there is a risk of overlooking or worse, misinterpreting cultural nuances, especially in multi-market projects where insight often emerges by comparing one culture to another.
Secondly, not enough scholarship exists, yet, that points to the veracity of insights produced through synthetic respondents. This is perhaps a case of when, and not if there will be further research into the authenticity of insight. The potential risk here is data collected might feel overly flattened, linear, or rational simply because data models are not great at capturing human–level nuances. Insight is most powerful when it’s rooted in emotion and feeling, and overly neat synthetic data quality might lead clients to the wrong path.
Lastly, qualitative insight is not simply used to test business or brand hypotheses but to fundamentally arrive at brave, original, new ways to generate theories of penetration, growth etc. Of course, a human researcher is using their skill, knowledge, and experience to arrive at these strategies, but interesting, unusual, or edgy respondents speed this process along. Synthetic respondents are programmed to be predictable, but real-life humans are not.
Unlike Mark Ritson’s assertion, we are not cigar-chomping cynics. But there must be pause and circumspection when adopting something so fundamentally core to the insights industry. From where we stand, magic is created when good quality, open-ended, unstructured data is interpreted by researchers and strategists with the ability to triangulate cultural context, brand problem and human data. In this future, perhaps a convergent approach of having both AI-enabled data and researcher interpretation can generate insights that are brave and point to new futures for businesses.
Together, AI and human interpreters make a dream team. AI does the heavy lifting, while we add the soul. It is a partnership that not only promises speed – but does it better, with compassion, creativity, and a whole lot of heart.
