January 25, 2026 4 minutes minutes read Admin

Offloading our thinking to AI risks weakening the very resilience that writing quietly builds

We’re outsourcing more and more writing to AI – and with it, some of the hard mental and emotional work that actually builds resilience. If writing is “adaptation in progress”, what happens when we stop doing it ourselves?

Offloading our thinking to AI risks weakening the very resilience that writing quietly builds

We talk a lot about “AI for productivity,” but much less about what we’re quietly losing as we type less and prompt more.

I just read a piece arguing that ordinary writing – journaling, messy drafts, even rage-texts you never send – literally rewires the brain and builds resilience. Writing forces you to retrieve memories, organise them, name emotions, and turn chaos into some kind of structure. It’s not just communication, it’s cognition plus emotional regulation on paper.

Now contrast that with our current AI era.

Instead of wrestling with a blank page, we paste a vague bullet list into a chatbot and get a fluent, polished paragraph back. Nice dopamine hit. But the brain work that actually builds resilience – the slow, uncomfortable process of finding your own words – is quietly outsourced.

A few things worry me here:

  • When AI drafts your emails, papers, reports, you get the result of thinking without the experience of thinking. The emotional “distance” and clarity that come from writing your way through a hard situation never fully happens.

  • If we stop naming our own emotions in words (“this reviewer email hurts”, “I’m afraid of this grant rejection”), we also stop training that neural circuit from amygdala panic → prefrontal clarity. The machine carries your feelings in its buffer; your brain never learns to shelve that metaphorical book itself.

  • In research, we already see a temptation to let AI “smooth” the story: generate abstracts, introductions, even whole sections. It feels efficient, but it also removes the friction where you reconcile contradictory data, notice gaps, and confront your own confusion. That friction is where scientific insight – and personal resilience – is forged.

  • For younger scientists, this could be especially damaging. If every tough email, rebuttal, or application is “co-written with AI”, you may grow a portfolio but not the internal voice that can face failure and reframe it.

Don’t get me wrong: I use AI tools every day in my own work – from molecular modelling to writing. But I want them as exoskeletons, not wheelchairs.

Some practical ideas for our AI-heavy research world:

  • Keep a handwritten or typed lab/journal that is 100% you. No autocomplete, no LLM. This is where you write the unfiltered version of what’s happening – scientifically and emotionally.

  • Before asking AI for help with a difficult email or rebuttal, draft a raw version yourself. Let the model help with tone and structure afterwards, not instead of your own thinking.

  • Treat writing as a cognitive experiment: the moment you feel overwhelmed, annoyed, or stuck, write for 5–10 minutes before you react. Notice whether the urge to “just ask AI to fix it” is stronger precisely when your brain would benefit most from doing the work.

In short: if writing is “adaptation in progress”, then fully delegating it to AI is, at best, adaptation outsourced – and at worst, adaptation cancelled.

Personal note: I’ve noticed that the pieces of writing that changed my own career and life the most were the ones I struggled through without shortcuts – grant failures, harsh reviews, big decisions. They were unpleasant to write and invaluable to have written.

How are you balancing AI assistance with the need to keep your own writing (and thinking) muscles strong in your scientific work?