Deepfake Personas in UX

What are they? How do they measure up? Insights from a new study

Jul 6, 2023

Personas - fictional, yet grounded representations of the target user group - have been instrumental in shaping design decisions. Love them or hate them, personas are ubiquitous and one of the first things people think when they hear the term UX… Traditionally, personas have been mostly static. However, as technology advances, we're seeing the advent of a new form of personas, coined 'deepfake personas’, that could potentially alter how we understand and cater to the end-user.

Commonly, personas are represented as persona profiles, also known as templates or layouts (Nielsen et al., 2015). These profiles are often static, circulated via PowerPoint slides, PDF presentations, or printed banners on office walls. This static form of user representation has met with criticism from persona researchers, prompting calls for more dynamic and immersive methods of persona presentation. Despite advancements in technology, multimedia inclusion (including voice and video) in persona creation remains unexplored.

In a recent study Kaate et al. (2023) explored ways of incorporating deep fake technology into traditional personas as a way to increase their effectiveness.

“Deepfakes are synthetic content generated using advanced deep learning and AI technologies”. Gamage et al. (2022)

Kaate and colleagues used cutting-edge deep learning algorithms to create deep fake personas. They created the deepfake personas using commercially available software (Synthesia). They selected avatars among those available in the software that corresponded with the demographics of the identified personas, and for each persona they provided a text narrative in the first-person that the persona then articulated for study participants (i.e., telling their name, background, and other information). The created deepfake personas were exported as video files (.mp4). The final result was dynamic, hyper-realistic representations of the users, complete with facial expressions and body language.

The researchers argued that the use of deepfake technology holds the potential to elevate the sense of engagement with personas. In the context of personas, this technology can be used to bring a user persona to life, essentially creating a digital version of a user that can speak, emote, and even interact to some degree...

To test their hypothesis the researchers conducted a comprehensive comparison of deepfake personas, classic personas, and narrative personas, assessing them on several parameters, including empathy, credibility, completeness, clarity, and willingness to use. Classic personas refer to what most of us think of when we hear the term persona — static representations of the end-user, combining a representative image with detailed textual descriptions of the user's demographic details, goals, needs, and pain points. Narrative personas were textual descriptions of the persona profile, providing a more dynamic and engaging presentation of the same user details. The study participants were 100 tech workers with some design experience recruited online.

Types of personas used in the study (Kaate et al. (2023)

No significant differences were found between classic and narrative personas in any of the measures used; they appear to be perceived equally by the participants. The findings, however, indicated a lower user perception of deep fakes compared to classic and narrative personas, largely attributed to the uncanny valley effect. This phenomenon, first proposed by robotics professor Masahiro Mori in 1970, suggests that as a figure becomes more humanlike, our affinity for it increases, until it reaches a point where it's almost but not quite human. This near-human, yet off persona elicits feelings of discomfort or eeriness, which Mori described as falling into the 'uncanny valley'. This effect has been widely observed in fields like robotics and computer graphics. The discomfort and decline in user perception measures were unique to the deepfake personas, representing a substantial hurdle to their application in UX.

Another possible explanation for these findings has to do with what stakeholders expect from a persona. Traditionally, persona templates remain quite static, resembling more fixed profiles than living, breathing entities. According to research, the typical persona format that stakeholders tend to favour includes a static image accompanied by several sections of information. This format, most often presented in paper or PDF format lacks the dynamic and interactive elements that could enrich the persona experience.

Additionally, humans tend to gravitate towards what’s familiar as it’s seen as a comfort zone providing us with an illusion of control. Any deviations, no matter how innovative or potentially useful, are typically met with initial resistance. According to Kaate and colleagues deep fake personas could be falling into this category — users are simply not accustomed to them.

Interestingly, the study also revealed a niche group (17.8%) of participants who preferred deepfakes over the other persona types. This suggests that despite the general perception, there may be specific scenarios or user groups where deepfake personas can provide unique value.

At this point, it appears that deepfakes, while promising, are not yet ready for mainstream use in UX. However, given the continuous advancement of AI and multimedia technology, it is conceivable that future iterations of deepfake technology will overcome these challenges. Curious to see what the future of UX holds…

Have you ever used or come across deepfake personas? How do you feel about incorporating them into your UX toolbox?