What humanness in AI actually means for conversation design
What humanness in AI really means for conversation design, from usability to social presence, and why human-feeling voice AI improves trust and outcomes.
Think back to the last time you interacted with a large language model (LLM), maybe ChatGPT or another application. Chances are, A.) you got something useful out of it, and B.) it felt nothing like chatting with a real person.
That’s all well and good for a text-based chat, but imagine that interaction over the phone, where timing, tone, and subtle cues matter even more. The gap between technically correct responses and genuinely human-feeling conversation is where conversation design makes all the difference.
At PolyAI, we break down humanness in AI into two dimensions: usability and familiarity/affinity. One ensures tasks get done efficiently. The other ensures users feel present, understood, and supported. Together, they define how human-AI interaction feels.
Usability: getting the task right
Usability is the hard stuff because it’s concrete and measurable. It’s about whether people can navigate a conversation intuitively and efficiently to complete their task.
- Can the system support natural turn-taking?
- Will it interrupt someone before they’re done speaking?
- Does it understand the user correctly?
- Does it follow the rules of human dialogue, like context-aware responses?
A big part of usability is spoken language understanding. If the system mishears or misinterprets speech, the conversation can break down quickly. For example, imagine calling your utility company to check on a power outage. A typical LLM might respond like this:
"In order to verify your account, I need your account number, then your date of birth, and then your service address. Once I have all that information, I can check your outage status."
It’s technically correct, but humans don’t communicate that way on the phone. A real agent would first ask for the account number, wait for your response, then ask for the next piece of information—breaking the process into clear, digestible steps. This kind of cooperative interaction keeps the conversation efficient and reduces cognitive load.
By reducing cognitive load, well-designed AI lets users focus on solving the problem, rather than figuring out how to navigate the conversation.
Familiarity and affinity: feeling present together
Usability alone isn’t enough. Conversations are social. They create rapport, trust, and a sense of presence. This is where familiarity and affinity—the soft stuff—come in. It’s about making the user feel like they’re working with the AI, not just receiving instructions.
- Natural intonation and timing: Pauses or unnatural rhythms can make conversations feel disjointed.
- Appropriate tone and register: Not too formal to feel cold, not too casual to feel socially unaware. This also varies across languages and cultures, which is why localization and expertise in target audiences matter.
- Cultural and linguistic adaptation: Each language and context requires its own nuance.
- Extemporaneous responses: Avoid scripted-sounding replies that break the illusion of real-time collaboration.
Speech synthesis introduces challenges like the uncanny valley: a voice that's almost human but slightly off can feel alienating. So, Managing these subtleties is key to social presence.
Find out how to design effective dialogue that delivers efficient and empathetic self-service
Dialog design: How to bring empathy to AI
When conversations actually feel human
When these elements are combined, users experience social presence, the feeling of being with another person. This isn’t just “making it sound nice.” Research consistently shows that even when users know they’re talking to AI, a strong sense of social presence:
- Increases trust and confidence in the interaction
- Encourages adherence to guidance
- Reduces repeat calls or follow-ups
- Improves overall satisfaction
This sense of social presence makes users feel like they are working together with the AI, not just receiving instructions from a machine.
Why AI still talks like a robot
LLMs have incredible capabilities, but they are trained mostly on text, not spoken interaction. This creates several challenges:
- Over-explaining: LLMs often ramble, giving more information than needed. In human dialogue, brevity and relevance matter. This is because they are trained on vast amounts of written text and do not inherently assume shared context or brevity as humans do.
- Ignoring shared context: Humans naturally assume what the other knows. AI needs explicit design to replicate this.
- Missing subtle engagement cues: Small language choices, like using “we” instead of “you,” signal shared involvement in a task. Active phrasing like “I’m checking availability” feels more present than “nothing is available.
The quirks that make LLMs so good at generating text—formality, neutrality, over-explaining—are the same things that make them sound off when dropped into a spoken dialogue. The job of a conversation designer isn’t just to rein in those habits, but to actively shape interactions that feel natural, collaborative, and human.
What to consider when designing AI that talks like a person
1. Keep responses concise and focused on the task
| Overly structured and formal | Relaxed and unscripted |
|---|---|
| “Our exclusive Birthday Bonus Rewards Program offers you the opportunity to enjoy a complimentary decadent dessert on your special day at any of our locations! Would you like me to enroll you so you can start celebrating in style?” | “I can also add your birthday to our rewards program so you get a free dessert next time you’re here. Would you like me to do that?” |
2. Signal engagement in the task
| Less effective | More effective |
|---|---|
| “Do you want to look for other tables around that time?” | “Do we want to look for other tables around that time?” |
3. Favor active, present-focused language
| Less effective | More effective |
|---|---|
| “Nothing is available.” | “I’m checking availability.” |
4. Calibrate tone and register
| Less effective | More effective |
|---|---|
| ”Could you please provide me with your account number” | ”Could you tell me your account number, please?” |
Why humanness matters
- Reduced cognitive load: Users can focus on solving their problem rather than figuring out how to interact with the system.
- Increased trust and confidence : Feeling understood encourages trust, even when users know they’re talking to AI.
- Higher adherence to guidance: Users are more likely to follow instructions or recommendations.
- Fewer repeat calls or follow-ups: Clear, cooperative, and engaging interactions reduce the need for repeated contact, and less screaming “I want to speak to a person!”
- Improved overall satisfaction: Conversations feel smoother, collaborative, and human.
Where usability and social presence meet
Humanness in AI is a combination of task efficiency and emotional engagement. Usability ensures users can complete their goals. Social presence ensures they feel supported while doing so. Thoughtful design bridges the gap between what AI can technically do and how users experience it.
Make every customer feel heard. Instantly. Speak to our team today.
