How AI Personalities Can Transform the Data Experience

Why does emotional engagement with data matter?

Data is at the heart of decision-making, driving everything from product development to customer service. Every presentation of data is ultimately telling a story, from a dry spreadsheet to Florence Nightingale’s famous casualty diagram. But how effective, impactful and, perhaps most importantly, intentional is that story? Emotional engagement is vital because it not only makes data more relatable, but also invites curiosity and deeper reflection. By intentionally crafting the personality of conversational AI systems, we can create experiences that enhance functionality through emotional connection.

Personality-driven AI interfaces help transform the process of making sense of data into an experience that can spark empathy, understanding, and even enjoyment. This approach offers a way to help AI systems humanize data, making the interaction with that data feel less transactional and more like a conversation.

The role of personality in AI systems

Integrating personality into data presentation systems is more than a party trick to add life to your pie charts. It’s about designing tools that can communicate with users in ways that reflect their mood, tone, and even sense of humor. Large Language Models (LLMs) make this not just possible, but accessible to everyone who needs to present data. By generating natural-sounding dialogue and infusing it with personality, LLMs allow for conversations that resonate emotionally—and at times playfully—rather than just delivering dry data. We believe all data tells a story, and LLMs can help you tell a better one.

What can that look like in practice?

During a recent workshop we hosted at HCI International 2024, we asked participants to design AI-enabled systems using an evolved version of the "Talking to Docs" framework. Our goal was to strip back the LLM interface as much as possible, until it contained just the elements we wanted to explore: a base prompt and a searchable vectorized document set. This simplification let our participants experience how setting intentions around personality and emotional framing through prompt engineering can recontextualize their unstructured data (held in the vectorized document set). Each participant took a unique approach to how their unstructured data could be presented in a more intentionally humanized way, blending functionality with personality. Here are some of our favorites.

1. Exploring Newton’s Principia Mathematica with an enthusiastic fitness coach

One participant created a conversational system to explain Newton’s Principia Mathematica in the style of an overly-enthusiastic fitness instructor. This personality is rooted in humor and contrast. Principia contains some of the most revolutionary ideas in physics, not to mention that it’s written in Latin–not an inviting combination. By choosing the framing of a fitness coach, the system immediately added a whole different set of cultural signifiers. But, the benefits went beyond style. The system actually framed its explanations of the ideas in the text as fitness metaphors, making them more relatable to a common human experience—like when it explained the concept of gravity.
‍
"Just like your muscles build momentum during a run, gravity pulls that ball down the hill! Keep up the energy,—you’ve got this!"‍

This example shows how AI can make even the most challenging material feel dynamic and engaging and, perhaps most importantly, welcoming to audiences who might not be seen as the core one for a particular subject​.

What could this mean in real-world applications? Imagine large multinational corporations creating engaging experiences for AI adoption and upskilling their workers; hospitals could explain complex medical treatments to patients in digestible ways, and schools and education companies could make tough subjects more fun and accessible to learners.

2. Demystifying Picasso’s Guernica for the curious

Another participant used a friendly, approachable art teacher personality to explain Picasso’s Guernica to younger audiences. By using relatable language and a conversational tone, the system helped break down the complex themes and techniques of the painting, making art more accessible as a patient teacher might:
‍
"Picasso broke things apart and put them back together in strange ways to show the confusion and chaos of war. It’s like a giant puzzle, and you’re the detective figuring it out!"‍

LLMs can be prompted to use existing pedagogy techniques (for example, Socratic dialogue or reciprocal teaching) to deepen understanding of topics a user might normally just skim over. This approach could be used by art and fashion players like museums and brands to explain the backstories of their artifacts, or by travel companies to add greater context to a visitor’s experience of a new place. Or, even more provocatively, this technique could be used to create understanding, consent and trust when agreeing to data usage, especially in highly-sensitive but complex areas like genetic data and other personal information.

3. Adapting the approach to user behavior

A third participant introduced a humorous twist to the bureaucratic process of filling out tax forms by creating an interface that changes its entire persona based on the attitude of its user. If the system detects the user’s tone as positive  and polite, it responds with the personality of a bubbly, cheerful government worker right at the start of their career. But if the system sees that the user is grumpy or rude, it replies with a sarcastic and grudging attitude, and adopts the persona of a government worker on their very last day on the job.
‍
"Ugh, all right. You do not put your ITIN here. Grab a coffee and take a deep dive into section 4 of Pub. 15 if you really want to know more."‍

What happens when you no longer have to design your interface with an assumed user intent, but instead can dynamically adjust it based not just on explicitly gathered user preferences, but on their tone and attitude?

There are many applications where creating some intentional friction when users don’t behave as intended is desirable, from financial planning to medical treatment plan adherence. Instead of building these frictions on a one-size-fits all design, we can use LLMs to estimate user attitudes and create the right behavioral nudges at the right time. And maybe they can even manage to be funny while doing it.

The future of emotionally-engaged AI design

Designing AI-enabled systems isn’t just about efficiency and ease. It’s about creating meaningful interactions that leverage how humans think and feel to make data more accessible, relatable, and even enjoyable. With their ability to understand and transform human language, AI systems naturally evoke emotional responses that differ from those triggered by traditional software or hardware. Even the absence of a defined personality becomes a personality in itself. As designers of these systems, it’s our responsibility to think deeply about how they’ll relate to users, not just how technically accurate their responses will be. The applications of personality-driven, emotionally-infused AI interfaces can span sectors from technology to healthcare, culture, media, education, government, consumer products, and beyond, potentially making data more vital, relatable, and actionable in all of our everyday lives. It all depends on design.

As AI technology advances, so does the potential to reimagine how we experience data—where even the most abstract information can evoke curiosity, empathy, and engagement.

Try it for yourself

We created this process to make engaging with the fundamentals of AI tools more accessible to designers–and that means you too! If you’d like to try your hand at our personality crafting exercise you can find instructions here. We’d love to hear about what you make! And if you’d like to partner with us to help solve an AI problem of your own, get in touch.