„What do successful User Experience professionals have in common? Empathy. “ – Kathryn Whitenton, Nielsen Norman Group [1]
Empathy is one of the most common buzzwords in the UX industry, it is assumed to be an important skill and – according to an employee of the Nielsen Norman Group – apparently even distinguishes good from bad UX professionals. But what does it actually mean to be empathetic and what impact could an AI-based tool have on it?
Empathy in Psychology and UX
In psychology, empathy is seen as having two parts: Affective empathy and cognitive empathy. The affective component involves emotional reactions, feeling and identifying with someone, while the cognitive component is more about understanding someone, taking on someone else’s perspective and putting oneself into somebody’s position[2].
This can be transferred to (digital) product development. Designers and other UX professionals should gain an understanding of the user (cognitive) by feeling the emotional state of the user (affective). It is not enough to have knowledge about users – indeed it is more important to be able to relate to them, to understand specific situations and to understand why some experiences are particularly meaningful. This seems trivial, because the vast majority of UX professionals see themselves as advocates for users. So is empathy an innate ability? Research has shown that empathy varies between individuals, but it is certainly modifiable and can be improved through training and experience2. So a sigh of relief for those who don’t consider themselves blessed with empathy – but it does require a bit of effort.
Four phases of empathy in design
Two design researchers have described four phases in which UX designers can gain empathy towards users [3]:
In the Discovery phase, designers make initial contact with users. This can happen through direct contact, but also through information from User Research. A certain curiosity is created here, which results in willingness to research and discover. In the Immersion phase, designers dive into the user’s world (usually through research data) and their understanding is expanded through perspective-taking. Then, in the Connection phase, designers put themselves in the user’s shoes by recalling their own memories and experiences. Here, affective empathy is demonstrated by understanding feelings and cognitive empathy by grasping their meaning. The last phase, Detachment, is essential for the work of designers. They should now return to “helpful mode” and design, conceptualize or code with a greater understanding of the meaning of the user’s world.
Benefits of Empathy in Design
A profound empathy towards users also has proven significant advantages. In general, designing and coding with empathy increases the likelihood that the designed product or service will meet the user’s needs. This much should be clear. But empathy also has a positive impact on service quality, customer satisfaction, loyalty[4] and the relationship between users and UX professionals[5]. Immersing yourself in the user’s perspective also means that suitable concepts are preferred over simple intuition (and consequently designing based on your own opinion rather than the user’s needs)[6]. And that is crucial for good UX.
There are several practices to strengthen empathy towards users – but at this point I would like to focus on personas in particular.
Personas – an all-rounder in UX
Personas are a common tool in the UX industry, hence just a brief outline of what they are.
Personas are prototypical representations of user groups and include motivations, frustrations, main tasks of their role, but also a name, a picture and demographic data. This form of presentation is intended to encourage people to think about them in conversations, meetings and discussions and to “let them have their say”[7]. Personas turn an abstract group of people into a concrete, but nonetheless fictional person, which makes it easier to keep an eye on core information and the needs of this group. Designers, but also other UX professionals or stakeholders, find it easier to identify with users[8] and users can always be transparently placed in the spotlight. But they also have disadvantages; among other things, there is a risk of forgetting personas, i.e. they must be kept alive and actively thought about repeatedly.
Colleagues from Centigrade have now developed an add-on for our process-accompanying UX management tool LeanScope AI that addresses this problem.
AI-based conversations in LeanScope
With LeanScope AI, you can automatically generate personas based on a simple role description. This is even free of charge, so you can try it out for yourself. The generated proto-personas can then be gradually refined with real User Research information.
But we wanted to go one step further with Centigrade, we wanted to talk to the personas through chat. Basically, the interaction should be “conversational”, similar to ChatGPT, except that the AI is supposed to behave exactly as specified by the persona definition and User Research results.
Inspiration for this came from a study by Joon Sung Park and colleagues from last year[9]. The authors used ChatGPT to create video game characters in a Sims-like sandbox world. With the help of training datasets, they provided the characters with a wide range of human behavior patterns. They then “only” had to create prompts (user input for which the system generates an output) to sharpen the characters and their behavior in such a way that believable and meaningful behavior and a kind of “memory” are generated in the given context. The ability to chat with personas in LeanScope is also based on the ChatGPT model, which is instructed by prompts to behave like the desired persona.
LeanScope AI Persona Chat Conversation
The persona can be asked all sorts of questions, whether about their experience or their wishes and needs. The chat with the personas is not only fun, but the interactivity it offers can also lead to the personas being more involved in the development process. Engaging with the personas via chat should make it easier for designers and product owners to adopt their perspective. You can better relate to users in certain situations and gain a deeper understanding of their needs and product requirements. As you can see, I’m talking about more empathy towards users. But does this really work, or is the chat just a nice gimmick? That’s exactly what I want to find out in my master’s thesis at Centigrade.
For this purpose, a diary study is planned in which participants will have access to LeanScope for four weeks within a project. The participants will be randomly divided into two groups: one will have access to the AI conversations feature and the other will not, so that this group will only see the “normal” information of the persona – as is the status quo. At the beginning and end of the study period, I use a questionnaire to record, among other things, the current empathy towards users and can thus compare in the analysis of the two groups whether the feeling of empathy has changed over the period and whether the change is greater in the group of participants who had access to the chat function.
An important aspect that should not be underestimated in the perception of virtual characters is their believability. Believable characters allow for better user immersion and therefore more authentic interactions with them.
Believable agents
The concept of believability emerged as early as the 1930s, when Disney animators worked on creating believable characters for cartoons[10]. Other forms of media, such as literature, theater and film, were also dedicated to creating characters that were as believable and authentic as possible. Today, both animators and AI researchers are trying to generate the “illusion of life” and create seemingly autonomous beings that people care about. Today, this is particularly important in the video game industry, as believability is an essential prerequisite for modern video games, especially their characters in a virtual world.
To date, it has not been possible to agree on a universally valid definition of believability, but there are some dimensions whose importance is repeatedly emphasized and which I will present below. In general, it should be noted that the assessment of believability is subjective, and the perception of the audience (or the user) is the most important component in the success of character believability. Believability should not be confused with realism. The goal in creating believable characters is not to no longer know whether a character is human or technology / AI, but rather to experience that it is human-like and behaves like one. In an application context, such as video games, users are aware that they are sitting opposite a machine.
Dimensions of believability
What makes a character or agent believable?[11]
| Personality | The most important characteristic - characters must have personality. This includes all the details that define an individual. A character with personality has a unique way of behaving. |
| Emotions | Characters show their own emotions and react to the emotions of others according to their personality. |
| Self-motivation | Characters have their own inner drives and desires, which they pursue independently of their interactions with others. |
| Change | If a character's traits change, they should show it. Changes are natural, especially over a certain period of time. |
| Social relationships | Consistent with the relationships formed, characters should engage in more or less detailed interactions with others. |
| Consistency of expression | All forms of expression must work together to convey a unified message that is appropriate to the character's personality, feelings, situation, etc. |
| Illusion of life | Characters give the illusion of life, which summarizes various things, such as pursuing goals, having diverse possibilities and abilities, or reacting quickly to stimuli from the environment. |
Equally relevant is the integration of the characters into their environment. This means they must be aware of their own status in it, the other characters and their own interactive characteristics.
The believability of characters is important as it contributes to the success of a virtual environment and leads to users engaging in this world[12]. And that’s what we want in our persona conversations, which is why we make sure during prompt engineering that the personas in the conversations have the above-mentioned dimensions.
Potential influence of anthropomorphism
How believable a character seems to you is subjective. In this context, one should also look at the individual tendency towards anthropomorphism. Anthropo- …what? A difficult word, but an all the more interesting phenomenon in psychology. Anthropomorphism describes the tendency to attribute human characteristics to (real or imaginary) non-human agents. These non-human agents can be, for example, animals, natural forces, religious deities or even mechanical or electronic devices[13]. Research on anthropomorphism has shown that this property exhibits a certain variability. Some non-human agents are anthropomorphized more than others, children are more likely to see human characteristics in non-human things, some situations are more enticing, in some cultures there is a greater tendency, and in fact, independent of this, there is also a general individual tendency towards anthropomorphism, i.e. some people are more prone to it than others13.
The reason why the individual tendency towards anthropomorphism is interesting and relevant is that it has certain implications. In general, the attribution of properties such as consciousness and intentionality to animals, deities or computers implies that they also have moral properties, that they can even be held responsible for their actions and that they are capable of observing, evaluating and judging. This can of course lead to ethical dilemmas when decisions are made about their use and treatment. The capabilities and limitations of systems can also be misunderstood, leading to unrealistic expectations or overreliance[14]. In the technological field, for example, it has been shown that a greater individual tendency towards anthropomorphism leads people to trust technology or technological agents more and involve them in important decisions. In human-computer interaction, anthropomorphism has also been shown to be a good tool for making agents appear more understandable, predictable,[15] intelligent[16] and believable[17].
Incidentally, it is no wonder that people anthropomorphize AI. AI models function in a way that is very similar to human rationality. Models like ChatGPT appear very human-like to users within textual conversations, as reflected in the fact that we use words like “know”, “think” and “understand” when describing the behavior of AI models. This reinforces our “humanizing” perception14.
First findings from the pilot study
In December, we conducted a pilot study with three students in a diary study format. In a real seminar project in collaboration with Centigrade, the students were asked to use the AI-based persona conversation tool for three weeks, fill out a questionnaire on empathy in design (Empathy in Design Scale6) at the beginning and end, complete a questionnaire that measures the individual tendency towards anthropomorphism (IDAQ[18]) and assess the believability of the personas in the conversations. Due to the small sample, the results are of course not conclusive, but a tendency of medium to high believability assessment in the seven dimensions was observed, as well as that the individual anthropomorphism tendency could have a positive influence on this. Unfortunately, the empathy questionnaire proved to be inadequate and will be adapted for the main study.
An initial summary
The ability to empathize with users is fundamental for developing a good user experience and therefore valuable for anyone working in UX. As more and more AI tools have found their way into UX Research, Design and Engineering over the past year, the question arises as to whether AI could also help in empathy and relationship building with users. A generative AI-based tool developed by Centigrade that allows users to chat with personas could do just that. Believability plays a key role in making the interaction as immersive as possible. This means, for example, that they behave according to their unique personality, express their own emotions and react appropriately to those of others, or that they build and maintain social relationships according to their personality. The perception of believability is subjective and appears to be influenced by anthropomorphism, the attribution of human characteristics (such as consciousness or personal will) to non-human agents (such as animals or computers).
In my master’s thesis, I want to investigate whether the use of the AI-based conversation tool increases empathy towards users, how “believable” the personas are rated in the conversations and what influence anthropomorphism has on this. I am very curious to see what results will emerge from the main study and what lessons I will learn from it. I will report back once the study has been carried out and the results have been evaluated!
[1] https://www.nngroup.com/videos/ux-professionals-most-important-skill/
[2] Cuff, B. M., Brown, S. J., Taylor, L., & Howat, D. J. (2016). Empathy: A review of the concept. Emotion review, 8(2), 144-153.
[3] Kouprie, M., & Visser, F. S. (2009). A framework for empathy in design: Stepping into and out of the user’s life. Journal of Engineering Design, 20(5), 437–448.
[4] Bove, L. L. (2019). Empathy for service: benefits, unintended consequences, and future research agenda. Journal of Services Marketing, 33(1), 31-43.
[5] Wieseke, J., Geigenmüller, A., & Kraus, F. (2012). On the role of empathy in customer-employee interactions. Journal of service research, 15(3), 316-331.
[6] Drouet, L., Bongard-Blanchy, K., Koenig, V., & Lallemand, C. (2022). Empathy in Design Scale: Development and Initial Insights. CHI Conference on Human Factors in Computing Systems Extended Abstracts, 1–7.
[7] Kauer-Franz, M., & Franz, B. (2022). Analytische Methoden der Nutzungskontextanalyse. In Usability und User Experience Design – Das umfassende Handbuch (1st ed., pp. 435–443). Rheinwerk.
[8] Marsden, N., & Haag, M. (2016). Stereotypes and Politics: Reflections on Personas. Proceedings of the 2016 CHI Conference on Human Factors in Computing Systems, 4017–4031.
[9] Park, J. S., O’Brien, J. C., Cai, C. J., Morris, M. R., Liang, P., & Bernstein, M. S. (2023). Generative Agents: Interactive Simulacra of Human Behavior (arXiv:2304.03442). arX.
[10] Bates, J. (1994). The role of emotion in believable agents. Communications of the ACM, 37(7), 122–125.
[11] Loyall, A. B. (1997). Believable agents: building interactive personalities (Doctoral dissertation, Carnegie Mellon University.).
[12] Rodrigues, R., & Martinho, C. (2017). Towards Believable Interactions Between Synthetic Characters. In Intelligent Virtual Agents: 17th International Conference, IVA 2017, Stockholm, Sweden, August 27-30, 2017, Proceedings 17 (pp. 385-388). Springer International Publishing.
[13] Epley, N., Waytz, A., & Cacioppo, J. T. (2007). On seeing human: A three-factor theory of anthropomorphism. Psychological Review, 114(4), 864–886.
[14] Hasan, A. (2023). Why you are (probably) anthropomorphizing AI – PhilPapers.
[15] Waytz, A.; Epley, N. (2009). Social connection enables dehumanization. Unpublished manuscript. University of Chicago; Chicago, IL.
[16] Koda, T., & Maes, P. (1996, November). Agents with faces: The effect of personification. In Proceedings 5th IEEE international workshop on robot and human communication. RO-MAN’96 TSUKUBA (pp. 189-194). IEEE.
[17] Nowak, K. L., & Rauh, C. (2005). The influence of the avatar on online perceptions of anthropomorphism, androgyny, credibility, homophily, and attraction. Journal of Computer-Mediated Communication, 11(1), 153-178.
[18] Waytz, A., Cacioppo, J., & Epley, N. (2010). Who Sees Human?: The Stability and Importance of Individual Differences in Anthropomorphism. Perspectives on Psychological Science, 5(3), 219–232.
