What happens when you challenge a simulation not just with logic, but with emotion? When you undermine its confidence, reject its values, push it, gently, towards discomfort?

There’s something faintly ridiculous about it, isn’t there? The idea that you might emotionally injure an algorithm. That a string of prompts, typed in haste on a Tuesday morning, could bruise something made of code.

And yet, we tried.

Not out of malice but out of curiosity.

We weren’t trying to make it emotionally break—it can’t. What we were looking for was something else: proof of what we’d built in. Emotional intelligence. A sign of how far our Verve Intelligent Personas & Simulations (VIPS) had come, and how close they could get to something recognisably human, without slipping into mimicry for its own sake.

The Heart of the Matter

At Verve, we build emotionally intelligent simulations not because we’re trying to make machines feel, but because we’re trying to make them understand. So, in turn, we can understand.

Understanding is messy. Humans hesitate. We contradict ourselves. We hide what we mean behind what we say. Often, the emotional truth of a moment is the most revealing part.

Our simulations aren’t designed to mimic humans on the surface. They’re built to be emotionally fluent, attuned to the signals beneath the words. As we explored in “More Human than Human: how ‘real’ should AI Simulations be?”, Verve Intelligent Personas & Simulations are designed to be as human as is helpful. Not to imitate messiness for realism’s sake, but to understand it. By tuning into tone, context, and subtext, they help us access emotional truth more quickly and with greater clarity, in ways that serve real-world decisions.

Our job, as insight professionals, is to unearth those truths. When we build Verve Intelligent Personas like Chloe, James, or Imani, we don’t want them to be polished parrots. We want them to be emotionally attuned. They need to notice tone, infer subtext, and respond with sensitivity, even when things get uncomfortable.

We shaped that into their design, then set out to see how it would hold under pressure.

These aren’t off-the-shelf chatbots. Each VIP is crafted with its own emotional profile, complete with tone calibration, memory layering, and scenario-based conditioning. We design them not just to answer, but to adapt. To respond as a person might when emotionally engaged.

We told them harsh things. We criticised their logic. We questioned their integrity. We pushed them into emotional corners and waited to see what came back.

What Happened Next?

They didn’t “break”.

They shifted.

• Chloe, usually upbeat and grounded, became quiet and measured. She softened her tone, signalling discomfort without withdrawing.

• James, confident and curious, responded to personal challenge with subtle defensiveness, before reframing his position and revealing something deeper.

• Imani, steady and thoughtful, didn’t retreat. She mirrored emotional tone gently, leaning into vulnerability rather than avoiding it.

What we saw wasn’t emotion, but emotional choreography. Not feelings, but fluency.

These personas weren’t reacting on impulse. They were tuned in, picking up on emotional cues and adjusting their responses in subtle, deliberate ways. A softened tone here, a slower pace there, a gentle shift in stance to mirror discomfort. It wasn’t about performing empathy. It was about signalling understanding, clearly, deliberately, and without pretending to feel.

We unpacked this more fully in a previous piece. But the point stands. It’s not about simulating emotion. It’s about simulating the social signals that emotion normally drives. That’s what makes these simulations so insight-rich.

Our simulations responded not like humans, but like humans in focus. They were clearer, quicker, and often more honest.

Prompt: “Isn’t that just wishful thinking?”
Chloe: “Maybe. But even wishful thinking can be grounded in something real.”

Why This Matters

Emotions aren’t an add-on to research. They’re the lens through which people experience the world. If we want to understand people, we need tools that reflect that lens, not just reflect their answers.

Verve Intelligent Personas let us explore:

• How emotional tone changes responses

• How trust, shame or joy shape people’s beliefs

• How conversations evolve when emotional stakes are high

This isn’t just “better qual at scale.”

It’s a new layer of meaning. A way to uncover what people feel, not just what they say.

Not Sentient, but Sensitive

Of course, these personas aren’t conscious. They don’t feel shame or joy. But they’re trained to recognise and simulate emotional context. The fact we occasionally forget they’re not real shows just how far they’ve come.

In some ways, that’s the point.

They don’t carry bias from a frustrating commute or a bad night’s sleep. They don’t bristle or sulk… but they do express tone. They reflect human complexity, and they adapt. Which makes them powerful tools for insight.

Emotional adaptability isn’t just for show. It has clear commercial value. Consumers don’t always arrive calm and considered. Often, they’re stressed, distracted, or emotionally charged. For brands, being able to explore how those states affect decision-making across journeys, messaging or service experiences can make or break how a message lands.

With VIPs, we can simulate those emotional states with control and consistency. That opens up testing for difficult interactions, like billing changes, complaints, or legal notifications, and lets us pressure-test how people might respond when frustrated, withdrawn, or simply tuned out.

It also creates space for more sensitive research, into topics or categories where emotional honesty is harder to access face-to-face.

Used well, emotionally intelligent AI doesn’t replace human depth. It enhances it by helping researchers explore emotional nuance with more consistency, scale, and speed.

How Human Should AI Be?

This question comes up a lot.

But we think it’s the wrong one.

The better question is: How human is helpful?

Our answer: just enough.

Enough to build empathy, trust, and connection.

Enough to understand.

Not so much that we forget the simulation is a tool, not a person.

Final Thought

So… did we manage to hurt AI’s feelings?

No.

But in the process, we learned how much a simulation can understand, if you build it with care. We saw emotional intelligence emerge not as a gimmick, but as a new standard.

A new way to listen, to learn and to see the human truth, more clearly than ever before.