AI Chatbots and Your Attention Span: What the Data Says

The question of whether AI chatbots and your attention span are at war isn't hypothetical anymore, and a piece in MIT Technology Review put real numbers to it. In Are AI chatbots making us lose control of our brains?, the writer sat down at SXSW London with Gloria Mark, a psychologist at the University of California, Irvine, who has spent 30 years measuring how digital tools change the way we think.

I build with AI tools every day, so I read this less as a warning about chatbots and more as a question about which mental muscles I'm quietly letting go slack.

---

📉 The number that should stop you scrolling

Mark ran what she calls "living laboratory" studies, tracking people's attention, mood, and behaviour across two decades. The trend in how long we hold focus on a single screen is the headline finding:

Year	Average focused attention on a screen
2003	2.5 minutes
2012	75 seconds
2014–2020	47 seconds

That's a roughly 3x collapse in twenty years. She also paired this with heart-rate monitors and found a direct correlation: the more people switched attention, the higher their measured stress.

Key takeaway: The attention drop predates ChatGPT. Chatbots didn't start this fire; they're petrol poured on a habit we'd already built.

If you've ever opened your editor, hit a hard bug, and reflexively alt-tabbed to a chat window before you'd even finished reading the stack trace, you already feel this in your own hands.

---

🧠 "Depth of processing" is the skill you're trading away

Mark's sharper point is about what happens when you delegate the thinking itself. When you hand writing, summarising, or evaluation to a model, she argues, you give up depth of processing, the active engagement that makes learning actually stick.

Her analogy is blunt:

"If you're not constantly exercising your muscles, they can atrophy."

For a developer or a student, that lands differently than it does for a casual user. The things we're tempted to outsource are often the exact things that build competence:

• Reading a library's source vs. asking a bot what it does
• Writing the first draft of a function vs. tab-completing it whole
• Debugging by forming a hypothesis vs. pasting the error and accepting the first fix
• Summarising a paper in your own words vs. reading the AI summary

The first column is slow and a little painful. That friction is the learning. Skip it enough times and you end up able to ship code you couldn't write or explain on your own, which is a fragile place to build a career from.

This matters more here than in a richer market. If your edge as a Sri Lankan engineer competing for remote work is that you actually understand the systems you build, then quietly trading that understanding for speed is trading away the one thing clients can't get cheaper elsewhere.

---

🤝 The "synthetic companion" angle is quieter and stranger

The article also raises something I hadn't thought hard about: Mark's concern over synthetic companions, the chatbots people increasingly talk to for comfort. Her claim is that emotional intelligence, like attention, needs reps. Real friendship demands effort, friction, and reading another person who won't simply agree with you. A bot that's always available and always validating removes that effort, and the skill can atrophy the same way focus does.

I don't have data to extend that claim, and the article doesn't put a number on it, so I'll leave it as her hypothesis rather than dress it up as fact. But it's worth sitting with.

---

🛠️ How I'm trying to use AI without going soft

Mark's own prescription is low-tech: read books instead of summaries, favour in-person meetings, lean less on GPS. For those of us who write software, here's how I translate that into practice without pretending I'll quit the tools.

Habit	The lazy default	The version that keeps you sharp
Reading docs	Ask the bot to explain	Read it yourself first, then ask to check your understanding
New concept	Accept the explanation	Re-explain it back in your own words before moving on
Debugging	Paste error, take the fix	Guess the cause first, use the bot to confirm or correct
Long article	Read only the AI summary	Read the original; use a summary to decide whether it's worth your time

That last row is the honest tension in running a tools site. We host an AI text summarizer, and I use it. The trick is when: a summary is a good filter for deciding what deserves your full attention, and a bad substitute for the reading itself. Same with our reading time estimator. Use it to commit to the full read, not to talk yourself out of it.

Bottom line: Use AI to remove drudgery, not to remove thinking. The moment a tool starts doing the part that was teaching you something, that's the part to claw back.

---

💡 What this means for you

If you're a student in Colombo cramming for exams, a freelancer juggling client chats, or a one-person team shipping a side project, the takeaway is simple and a little uncomfortable: the numbers Gloria Mark measured are about us, not some distant generation of "phone-addicted kids."

• 47 seconds is the average focus span she recorded in her latest data. Notice how often you break before that.
• Attention-switching tracked with higher measured stress, so this isn't only about output, it's about how frazzled you feel at the end of the day.
• The skills most worth protecting are the boring, effortful ones: reading closely, writing your own first draft, and sitting with a problem before reaching for the answer.

None of this means stop using AI. I won't, and you probably shouldn't either when it genuinely saves you time. It means being deliberate about which mental reps you keep doing yourself, because those are the ones that compound into being good at this. The tools aren't the threat. The quiet decision to stop thinking is.