How psychology can make AI-powered customer personas more useful Clio

How psychology can make AI-powered customer personas more useful

 Clio

The marketing characters are fictional, but still useful. Today, many tools claim to create virtual characters that turn out to be little more than elaborate LLM instructions.

Combining decades of psychological research with real-world behavioral data could make virtual characters more effective at predicting customer behavior.

Why traditional customer personas are still important

In the 1959 song “Maria” from “The Sound of Music,” lyricist Oscar Hammerstein asks how can we “solve a problem like Maria?” It is an attempt to encapsulate the complete character of the aspiring nun in a few lines.

We get behavioral traits (she goes to mass and hides curlers under her wimple) and attitudinal traits (she is described as “a fickle gibbet” and “volatile”). The nuns themselves can’t agree on whether they are good or bad (e.g. whether she is a demon or a lamb), but we get a clear picture of the story’s protagonist.

For decades, psychologists, designers, and marketers have tried to distill a person’s entire personality, behaviors, and traits. We’ve all worked with characters: encapsulated descriptions of an individual (real or fictional) who represents a larger group.

Personas help with UI/UX design, understanding product usage and functionality, and bringing abstract segmentations to life through an individual who represents the ideal customer or user.

By anthropomorphizing you, characters help us understand how a real person, who may not be anything like us, purchases or uses our products and services. Despite their limitations, they remain an effective tool in our marketing arsenal. Now we have their shiny new cousins: virtual characters.

10x your SEO with Semrush for business.

The most powerful SEO platform in the world, created specifically for businesses.

Request a demonstration

How virtual characters work

In their simplest form, virtual characters involve entering character descriptions into an LLM, along with instructions and guardrails, so you can chat with them. We can ask them to describe how they would react to new situations or decisions they would make.

This can bring individuals or segments to life and provide deeper insight into their motivations and likely reactions. To experiment, try pasting a description of Maria into a chat session and asking her to speak in her voice. You can use a simple prompt like:

  • “Please speak to me in Mary’s voice, as described below. Make your answers short and conversational. Speak as if you were her, not as someone describing her: “

This fun exercise is reminiscent of the way virtual characters are often generated today, albeit with much more specific instructions and guardrails.

Virtual personas like this, populated with in-depth behavioral and attitudinal data from real people, can be vital tools as you explore the depths of your customers and market.

Like simple characters, they have limitations. We rely on LLMs to fill gaps in what we know about individuals of origin, which can lead to generalizations and stereotypes. This risk increases as we move away from embedded data.

If you have data on how a specific Gen Z individual views car ownership and brands and then ask their virtual persona about pet ownership, LLM will complete with what it thinks a Gen Z member would respond.

How psychological research can improve virtual characters

Psychological research in recent decades offers a way forward. Rather than a set of behaviors and attitudes, we have data within a specific framework that may allow us to infer behaviors beyond what we have observed or measured—that, at least, is the hope.

Researchers debate which structures best predict behavior. Consider a commonly cited example: the Big Five personality model. This framework has its roots in 1930s work on the lexical hypothesis, which postulated that language reflects personality differences (an interesting precursor to today’s language models). This model evaluates individuals on five dimensions: openness, conscientiousness, extraversion, agreeableness, and neuroticism. We could give Maria a score like:

  • Opening – Very high: Daydream, sing at the wrong times and improvise.
  • Conscientiousness – Low: He is late, forgets the rules and returns with the missing items. His heart is in the right place, but his agenda is not.
  • Extraversion – High: She’s hot and loud, draws energy from people and wins people over with music.
  • Pleasantness – High, with an edge: She is kind and trusting, but confronts people when she thinks they are wrong. It has a high level of heat, a low level of compliance.
  • Neuroticism: low to moderate. She is mostly resilient and optimistic, but runs away when her feelings overwhelm her.

A model like this could apply to domains beyond the primary dataset. If we wanted to know whether Maria would be more likely to sign up for a hiking trip or an all-inclusive resort, the psychological picture would give us much more to work with than a description of her behavior in the convent.

Where will virtual characters go next

The process of building personas today is moving towards hybrid descriptions of people that combine real behavior, attitudinal statements and psychological profiles such as the Big Five, as well as more decision-focused models (such as Prospect Theory). There is also a move beyond chatter towards models designed to predict actual behavior rather than simply describe it.

I often remind people that the characters have always been invented, but that hasn’t limited their usefulness. By using virtual characters and beyond, we have the opportunity to integrate a long history of research and build characters true to the humans behind them.

Leave a Reply

Your email address will not be published. Required fields are marked *