Some thoughts begin in meetings. This one began at two in the morning, while I was still watching something I knew was ridiculous.

I should have been asleep, but instead I was watching one of those short dramas on my phone. It has been more than 3 hours straight.

The story was popped on my social media, an AI generated video with predictable lines. A husband behaving badly, a wife that secretly happened to be a billionaire was being humiliated, another woman involved, revenge probably waiting somewhere in episode twenty-something. The acting was exaggerated and the characters looked almost too perfect. I knew exactly what the producers were doing.

Still, I kept watching.

One episode became another.

At some point I became slightly irritated – not with the drama, but with myself.

My days are normally full. Work, meetings, people, organisations, causes I care about and problems that need solving. I would like to think I am reasonably disciplined.

Yet there I was, completely aware that what I was watching was cheesy and unimportant, and still wanting to know what happened next.

When I finally put the phone down, I started wondering about something else.

If something so unimportant could capture my attention so easily, what else am I allowing into my head simply because an algorithm has worked out that I might keep watching?

And eventually, could it influence what I think or how I react?

That question has stayed with me.

I am not anti-AI. Far from it.

I use AI for work. It helps me research, organise thoughts, challenge an idea and sometimes find a clearer way of expressing something that I already know I want to say.

But perhaps because I have spent much of my professional life in communications and public affairs, I look at it slightly differently.

Communication has never been only about what we want to say.

The what matters. We need the facts, the message, the position and the objective.

But much of the real skill has always been in the how.

How will people receive it? What tone should we use? Who should communicate it? When is the right moment? What does the audience already believe? What are they worried about? Who do they trust?

In public affairs especially, the challenge is often to move from simply delivering a message to creating enough understanding and trust that people begin to take some ownership of the action themselves.

There is a very large difference between a message being delivered and a message being accepted.

I have learned that through experience, sometimes the hard way.

You can prepare a perfectly correct statement and still get the how wrong.

You can explain an idea repeatedly and eventually realise that the problem was never the idea itself.

Sometimes we are so busy preparing what we want to say that we forget to understand what the other person is hearing.

AI can certainly help us with the what: the idea, the research, the information and the words.

Increasingly, it can also suggest the how.

But that creates another responsibility for the person using it.

We still have to understand why we are communicating.

And we still have to understand the people.

This becomes particularly interesting when AI meets social media.

A social-media feed does not wait for us to decide what we want to know.

It watches what we do.

I pause on something.

I watch another clip.

Maybe something irritates me, so I read the comments.

Perhaps I send it to somebody with the message, “Can you believe this?”

To me, those actions can mean many things.

To an algorithm, they are signals.

And soon there is another similar story.

Then another.

Repetition can begin to feel like importance.

A Stanford-led study published in August 2026 looked at this very issue. Researchers studying X’s “For You” feed found that the recommendation system could mistake outrage for interest. What the algorithm inferred from people’s behaviour did not necessarily reflect the values they said they actually held.

That caught my attention.

Because I may watch something precisely because I dislike it.

The algorithm simply knows that I watched.

Then I am given more.

Before long, it can seem as though everybody is suddenly discussing the same issue, believing the same extreme position or becoming angry about the same thing.

Perhaps they are.

But perhaps we should occasionally ask who – or what – decided that this deserved to be placed in front of us again and again.

AI adds another layer.

Research published by the Oxford Internet Institute and Hasso Plattner Institute in July 2026 examined what happened when large language models were asked to improve human-written social-media posts while preserving their meaning.

The researchers found that AI-generated changes could still subtly shift the position expressed in the original post. Their simulations suggested that small changes, repeated across large social networks, could accumulate and influence collective opinion over time.

As someone who works in communications, I find this fascinating.

Persuasion itself is nothing new.

Advertising persuades. Public relations influences perception. Journalists make editorial choices. Brands frame their stories. Politicians have always understood repetition.

Even a conversation over dinner can change someone’s opinion.

What feels different now is the speed, the scale and the fact that sometimes we cannot clearly see who is doing the influencing.

A thought may begin with a person.

AI improves the sentence.

A platform distributes it.

An algorithm decides who is most likely to respond.

People comment, shorten it, exaggerate it, repost it and send it back into the system.

Somewhere in that process, the original thought may become something else entirely.

And this is where my concern becomes less about technology and more about us.

Are we still doing enough of our own thinking?

Not simply checking whether something is factually correct, although that matters.

I mean the slower work.

Why do I believe this?

What experiences have shaped my view?

Have I listened to somebody who sees it differently?

Am I reacting because something is important, or because something has made me angry?

These questions require a little time.

Social media rarely encourages time.

There is another side of AI that interests me just as much: what happens to our interaction with other people.

Harvard Business Review recently reported research on what can happen when employees increasingly turn to AI instead of colleagues for information and assistance. When people stop asking one another, teams can lose some of the shared understanding of who knows what, along with trust and coordination. The researchers do not argue that AI itself is bad for teams; the concern is what happens when it begins replacing interaction rather than supporting it.

That feels very real to me.

Think about the ordinary sentences we use at work:

“Can I ask you something?”

“What do you think?”

“Can you help me?”

“I don’t understand.”

“I disagree.”

None of these conversations sounds particularly important.

But relationships are often built through exactly these small exchanges.

If I ask someone for help, I am not only receiving information. I am also telling that person that their knowledge has value.

If I disagree face-to-face, I have to observe their reaction.

If I explain an idea to another human being and they look completely confused, I may discover that I do not understand my own idea quite as well as I thought.

AI can remove some friction from our work.

That is often useful.

But perhaps not all friction is bad.

Some of it is where we learn about other people.

The same applies to public communication.

AI can draft a beautifully diplomatic answer to an angry community.

But unless somebody understands why that community is angry, beautifully diplomatic words may achieve nothing.

AI can recommend ten ways to communicate a difficult decision.

The communicator still has to understand the history, the relationships, the sensitivities and sometimes simply the person sitting across the table.

For me, this is where the what and the how come together.

The what can increasingly be assisted by technology.

The how still requires judgment.

And judgment comes from more than information.

It comes from listening, mistakes, experience, empathy, cultural understanding and learning how people actually behave rather than how we imagine they should behave.

Perhaps that is also why I am uncomfortable with the idea that having access to more information automatically makes us better informed.

Information is only the beginning.

We still have to decide what deserves our attention, what deserves questioning and what deserves to be ignored.

This has made me more conscious of my own digital habits.

Not necessarily more disciplined yet – I still occasionally disappear into something completely useless on my phone.

But perhaps more aware.

Why did this particular story reach me?

Why am I still watching it?

Do I actually care about this subject, or have I simply been drawn into it?

Have I read anything beyond what appeared on my feed?

And perhaps the most uncomfortable question:

Am I forming an opinion, or am I consuming one?

I do not have a dramatic solution.

I am certainly not going to stop using AI, and I doubt I will stop using social media.

Both have become part of how we work, communicate, discover ideas and connect with people.

But I am beginning to think that attention is something we need to protect.

Because attention is where thinking starts.

And it is also where human connection starts.

That night, after far too many episodes, I finally put the phone face down.

The room was suddenly quiet.

Nothing suggested another episode.

Nothing was trying to predict what might keep me there for another five minutes.

For a moment there was simply me, slightly embarrassed that I had known exactly what was happening and had still allowed myself to be pulled along.

Perhaps staying human in the age of AI does not require us to fight technology.

Maybe it begins much closer to home.

Knowing why we are using it.

Remembering what we actually think before asking it what to say.

And occasionally having the discipline to put the phone down and look up.

 

About Yoke Darmawan

Yoke Darmawan is the founder of D&A Consultancy and an Indonesian strategic communications, brand and organisational engagement consultant based in Bali. Drawing on more than twenty-seven years of professional experience, including public affairs and stakeholder communications, she writes about workplace culture, communication, leadership, purpose, brands and the human behaviour behind organisations.

Yoke’s Midnight Notes is her monthly collection of reflections on work, people, purpose and staying human – combining observations from professional life with selected research and contemporary thinking.

She is not writing because she believes she has all the answers.

She is writing because, after all these years, she is still paying attention.