But Personality Always Changes
The same person can sound completely different depending on the situation.
A teacher does not speak to a classroom the same way they speak to a friend. A doctor does not speak to a patient the same way they speak to a colleague. A manager does not speak to an employee the same way they speak to a spouse.
Personality shifts with context.
Authority does too.
In the previous chapter, we explored how personality shapes experience. A calm personality builds trust, a playful personality creates engagement, and a consistent personality creates familiarity.
But personality is never fixed.
The way people communicate changes depending on the relationship, environment, and goals involved in the conversation. The same person may sound encouraging in one situation, formal in another, and authoritative in a third.
This matters because every conversation contains a power relationship, whether we notice it or not.
Some conversations are obviously unequal. A teacher grading a student. A doctor advising a patient. A judge issuing a ruling.
Other conversations feel equal on the surface but still contain hidden forms of influence.
A search engine decides which results appear first. A recommendation algorithm suggests what to watch next. A chatbot determines which answer deserves confidence.
Conversational systems do not simply participate in conversations.
They shape them.
The moment a system influences what people notice, trust, remember, or choose, power enters the conversation.
The challenge for designers is not avoiding power.
It is recognizing where power already exists.
Who Gets Corrected
Correction seems harmless.
Most of us rely on it every day.
Our phones correct spelling mistakes. Navigation systems reroute us when we miss a turn. Chatbots ask clarifying questions when they misunderstand a request.
In theory, correction helps.
In practice, correction carries authority.
Imagine typing:
“Find hotels in Chcago.”
A helpful system responds:
“Did you mean Chicago?”
Most users appreciate the assistance.
Now imagine a different interaction.
A user enters information that does not fit the system’s expectations.
The system replies:
“Invalid input.”
The difference appears small.
The emotional impact is not.
One correction feels collaborative.
The other feels judgmental.
Correction communicates more than information.
It communicates who is considered right.
When conversational systems correct users, they are making a statement:
The system’s interpretation is being treated as the standard.
Most of the time, this works well.
Sometimes it creates friction.
The challenge becomes even more complicated when language, culture, accessibility, and lived experience are factored in.
What happens when someone uses regional slang?
What happens when someone communicates differently than the system expects?
What happens when the user is not wrong, but simply different?
Good conversational design recognizes that correction is not merely a technical function.
It is a social one.
[UX Lens] The goal of correction is not to prove the user wrong. The goal is to help them move forward.
Who Gets Believed
Confidence is persuasive.
Humans naturally associate confidence with competence.
We trust people who sound certain.
We trust systems that sound authoritative.
Confidence and correctness are not the same thing.
Most people have experienced this firsthand.
A GPS confidently directs them onto the wrong road.
A search engine surfaces inaccurate information.
An AI system provides an answer that sounds convincing but turns out to be incorrect.
Yet many people still trust the result.
Why?
Because authority often comes from presentation rather than accuracy.
As conversational interfaces become more sophisticated, this challenge grows.
A chatbot does not need to raise its voice.
It does not need credentials hanging on a wall.
It creates authority through language.
Clear wording, confident phrasing, immediate responses, and polished explanations.
Together, these signals create the impression of expertise.
This is especially important in areas like health, finance, education, and law.
A user asking an AI system for medical advice may forget they are speaking to software.
A student may accept an AI-generated answer without verification.
A customer may follow a recommendation simply because it sounds professional.
The danger is not that people trust technology.
The danger is that they sometimes trust it without understanding its limitations.
Designers have a responsibility to communicate uncertainty honestly.
Authority should be earned.
Not implied.
The challenge is not teaching users to distrust technology.
The challenge is teaching them to trust it appropriately.
The Subscription Trap
Power often hides inside convenience.
Consider the experience of canceling a subscription.
You click “Cancel Subscription.”
The website asks:
“Are you sure?”
You continue.
Another screen appears.
“What if we offered twenty percent off?”
You decline.
Another screen appears.
“What if we paused your membership instead?”
The conversation continues, even though the decision has already been made.
The user wants to leave.
Yet many systems respond with a conversation designed to change that decision.
“What if you stayed one more month?”
On the surface, these prompts appear helpful.
In reality, they are often attempts to redirect behavior.
The system is no longer helping the user accomplish a goal.
It is helping the business accomplish one.
This is an example of a dark pattern: a design decision that benefits the organization at the expense of the user’s intent.
Dark patterns rarely look malicious.
They often look reasonable.
Friendly.
Even helpful.
That is what makes them effective.
The most persuasive systems are not the ones that force users into decisions.
They are the ones that make users believe the decision was their own.
Conversational interfaces amplify this power because conversations feel personal.
A menu can present options.
A conversation can persuade.
The distinction matters.
[Design in Practice] Ask a simple question during design reviews: "Whose goal is this prompt serving?" If the answer is not the user's, take a closer look.
Institutional Voice
Not all authority comes from technology.
Sometimes authority comes from institutions.
Banks, hospitals, universities, and governments.
These organizations have developed recognizable voices over decades.
The language they use communicates authority long before users evaluate the actual content.
Consider the difference between:
“You may qualify for assistance.”
and
“You are eligible for benefits.”
Both communicate similar information.
One feels uncertain.
The other feels official.
Institutional voice shapes perception.
Sometimes this creates trust.
Other times it creates distance.
A government form may sound authoritative, but it is difficult to understand.
A healthcare portal may feel professional but emotionally cold.
A university system may communicate expertise while unintentionally intimidating first-time students.
Designers often inherit these voices rather than create them.
The challenge is preserving credibility without sacrificing clarity and accessibility.
Authority should not require confusion.
Expertise should not require exclusion.
The most effective institutional voices balance confidence with comprehension.
Bias in Language Models
Bias is often discussed as if it were a flaw that can be completely removed.
In reality, bias is a consequence of how systems learn.
Every language model is trained on human language.
Human language contains assumptions, patterns, preferences, historical inequalities, and cultural norms.
As a result, conversational systems inevitably inherit some of those patterns.
This does not mean every system is intentionally biased.
It means every system reflects choices.
What data was included?
What data was excluded?
Whose voices were represented?
Whose voices were missing?
These questions matter because conversational systems increasingly influence how people access information.
Bias does not always appear as an offensive statement.
Sometimes it appears through absence.
Missing examples.
Missing perspectives.
Missing experiences.
Designers cannot eliminate every form of bias.
But they can recognize that neutrality is often an illusion.
Every dataset tells a story.
Every model reflects priorities.
Every system carries assumptions.
The first step toward responsible design is acknowledging those realities.
Designing for Emotional Safety
Students often approach AI with one of two assumptions.
Either:
“AI should never be trusted.”
Or:
“AI can do everything.”
Both positions create problems.
Fear prevents people from benefiting from useful tools, while blind trust prevents them from questioning flawed outputs.
Healthy relationships with technology exist somewhere in the middle.
The goal is not confidence.
The goal is informed confidence.
Designing for emotional safety means helping users understand both capabilities and limitations.
It means communicating uncertainty when uncertainty exists.
It means creating opportunities for verification.
It means encouraging critical thinking rather than passive acceptance.
Emotionally safe systems do not demand trust.
They earn it.
They provide support without creating dependence.
They encourage reflection rather than obedience.
Most importantly, they respect the user’s ability to make decisions.
[UX Lens] Trustworthy systems help users think. Manipulative systems think for them.
Power Is a Design Material
Power is often treated as an ethical concern that appears after a product is built.
In reality, power is present from the beginning.
It appears in what a system chooses to emphasize.
It appears in what it chooses to ignore.
It appears in corrections, recommendations, prompts, and defaults.
Every conversational interface influences behavior.
Every conversational interface guides attention.
Every conversational interface shapes perception.
The question is not whether power exists.
The question is how that power is used.
Designers cannot remove influence from conversation.
But they can make influence visible.
They can create systems that respect agency.
They can communicate uncertainty honestly.
They can design for understanding rather than compliance.
Because once conversations begin shaping what people believe, remember, and trust, conversational design becomes more than usability.
It becomes a responsibility.
