To do this we use aptitude and behavioural user research techniques. Depending on what we want to clarify, we will use quantitative or qualitative techniques.
Quantitative research techniques resolve uncertainty in the ‘What’ and need to have sufficient critical mass for statistical validity.
Qualitative techniques provide certainty about the ‘why’ and the ‘how’ and delve deeper into the personal experience of users.
With generative artificial intelligence, so-called ‘synthetic users’ are being introduced. Synthetic users are artificially created entities thanks to AI that imitates the behaviour, preferences and usage patterns of real users.
The use of synthetic users aims to contribute to:
- Cost reduction
- Time reduction
- Avoiding data regulation and privacy issues as they are not real people
- Breaking down barriers between quantitative and qualitative invitation.
Disadvantages of using synthetic users in research:
- Lack of authenticity: Basically, we are talking about models that have not experienced real life or emotions.
- Flattening of people: AI, at least until now, has had a problem of homogenising experiences and identities and this means that nuances that are key when we talk about user experience are lost.
- bias: As a field of social research, it is extremely sensitive to biases which, in the case of human biases, we are very familiar with, but in AI they are still being studied.
And I would add an important drawback when it comes to qualitative research: the loss of direct interaction between researcher and user. We are not just talking about verbal communication, but also non-verbal expressions, spontaneous emotions or unexpected insights that did not fit into the script but which the user has shown you by serendipity, arising in real interactions.
Having said that, I think it’s a world to explore and my recommendation if you are going to use synthetic users is:
- Take care of the data sources and the information that will feed into it
- Work very well on the hypotheses so as not to feed back into biases
- Only do it with quantitative techniques
- Review the results carefully and think critically. After all, these models are built to meet our expectations as researchers, so they can be too complacent with the hypotheses formulated.
- Triangulate the study with qualitative techniques using real users to validate the results.
- Share your experience with the entire UX community
