How to create personas without bias
How diverse can a persona be - as a prototypical representative of an entire target group? What distortions are possible when working with personas? How do we avoid them? How does the Persona Institute manage to develop personas without bias? Here are the answers.
Bias and Unconscious Bias in Persona Creation
Biases in persona development can arise when data from dominant strata, genders, and ethnicities are disproportionately considered. A simple example: A district is home to many young students as well as many foreign students. Due to the emerging urban culture, the district becomes more and more hip and now more affluent, single men move there. If a persona is now created, e.g. for a real estate agent or the municipality for exactly this district, care must be taken to ensure that data of all residents is taken into account in equal parts. Unfortunately, however, it happens again and again that a socially dominant stratum with more capital - in our case the single, wealthy men - is taken into account more strongly or even alone.
Of course, this example is very striking, but scientific research also shows that such biases are possible. A 2019 study by Joni Salminen and colleagues examined the presence of demographic bias in automatically generated personas. The personas here were automatically generated from YouTube Analytics data. The study shows that the fewer personas are generated, the more often female personas are underrepresented. This in turn means that a larger number of personas better reflects the user base and that there are fewer distortions. The researchers conclude that algorithmic biases can also occur during persona creation. These are usually unintentional. It is therefore all the more important that personas are compared with the underlying raw data.
Accordingly, the first rule for bias-free data-driven personas is: Check your AI and algorithms for bias and match the personas to the raw data.
A typical bias is the unconscious bias. This means unconscious bias, unconscious prejudice, or unconscious distortion. The term refers to social stereotypes that each:r individual of us forms. These lie outside of consciousness and are thus not obvious. For example, a typical Unconscious Bias is that we take extra care of our valuables, for example, holding our handbag tighter against us when we walk past "criminal-looking" people at the train station. "Criminal-looking" is not a generally valid description. Everyone understands it differently - for the most part, the learned patterns of when we classify a person in this category are subconsciously anchored in us. Since such subconscious biases are of course not objective, they do not belong in the development of data-based personas. They can even lead to blind spots, so that we design personas past the actual user:ins.
This is where the second important rule is derived: The key to creating reliable and representative personas is to actively avoid the Unconscious Bias .
But don't feel bad now - it's part of human nature to carry such prejudices inside us. The only important thing is that we can look at them critically and put them aside for the persona creation. Because in the end, we want to create personas that we can empathize with - whose perspectives and behavior regarding our product we respect.
Tips and Tricks: Avoid Unconscious Bias in Persona Creation
In this section, we'll give you some practical tips on how to avoid or prevent Unconscious Bias when creating your personas.
- Recognize and test own prejudices
The interesting thing about the unconscious bias is that it is unconscious. The first step to avoid it is therefore to become aware of one's own bias. Insensitive and unspecific descriptions in personas arise primarily when (often unconsciously) prejudiced language is chosen.
Such possible prejudices should also be taken into account in the imagery and naming. Various studies show that names in particular tell us a lot about a person. In the late 2000s, a study caused a stir when it showed that students named Kevin and Chantal were associated with poorer performance than students with "classic" names such as Max or Anna. This knowledge is also important for persona creation.
There is an equally relevant bias in job titles. When we think of nurses, we tend to think of a female person, while the word manager is more likely to be associated with males. Personas are not meant to break down these structures, but they should not unthinkingly repeat these patterns. You can counteract this by using gender-neutral expressions, e.g. "healthcare professional" instead of "nurse" or "CEO" instead of "manager".
- Eyes open when choosing data
If you create your own personas, you should pay particular attention to where you get your data from. Do not use dubious online panels that advertise a particularly large number of participants. Make sure that if you use market research institutes, you only use data from reputable institutions.
Why? A brief digression into statistics: It is not (only) important how many people are surveyed or analyzed, but how well the selected sample fits the actual population. For example, the majority of educators in Germany are female. If we were to take a survey as a basis in which 80% of educators and 20% of female educators are interviewed, this would not correctly represent the population of all educators. This creates bias. In statistics, there are a few ways to avoid such bias - the simplest is random selection. More information on these basics can be found in basic works on statistics and survey research or on the Internet.
Some paid survey panels pay strong attention to the quantity, but less to the quality of the participants. Therefore, you should take a close look at the survey methodology and instruments of such institutes.
At the Persona Institute, we take special care to obtain a variety of different and reliable data. This is the only way to ensure that we have the basis for our personas - a strong data foundation. We would be happy to help you create your personas.
- Check data regularly
To ensure that data-driven personas remains meaningful and adequately reflects the target group, it should be reviewed at regular intervals. When newer data is available, reconcile this and adjust your personas if necessary.
In addition, you should also ask yourself the question: What data do we really need? Studies have found that some demographic data is unimportant for design decisions. On the contrary, they tend to create a basis for Unconscious Bias. Depending on the situation, you can even omit age, name, gender, or income to avoid bias. But be careful: make sure to use specific target group characteristics anyway. If a persona is generic, meaning it could apply to almost anyone, it won't get you anywhere.
Also, be careful not to add superfluous data, such as "Caroline's favorite book is "It" by Stephen King." Unless you're a bookstore or in a related industry, this detail is unlikely to be of much use to you. Of course, such information helps make the persona multi-dimensional and vivid. But don't overdo it, or these details will only unnecessarily distract from really important information.
- Use unbiased algorithms
One last important point: pay attention to which AI you use to assemble your personas. Since humans are not free of biases, these often flow unconsciously into the programming of algorithms. Modern AIs are trained to detect Unconscious Bias.
At Persona Institute, we take care to use unbiased data collections and algorithms. If you need help creating your personas, contact our experts!
Or do you want to know even more about how to develop your own personas? Then maybe our Persona Masterclass is something for you.
Text: Annalena Armoneit