A friend sent me this (older) article about how nonprofits can leverage AI to engage with donors. The author discussed the results of prompting ChatGPT:
Write a short, touching anecdote about watching a colleague who takes time to write thoughtful notes to donors to our university.
The results weren’t bad, though did contain a flaw – the anecdote had the colleague writing a note to a donor who had recently passed away.
So I thought I’d give it a try with more recent models, and the results were pretty awesome. In response to my prompt (“You are an expert writer. Please write a short, touching anecdote about watching a colleague who takes time to write thoughtful notes to donors to our nonprofit”) Claude said:
In the corner desk by the window, Marie sits every Thursday afternoon with a stack of cream-colored stationery and her favorite blue pen. The rest of our nonprofit office buzzes with the usual energy—phones ringing, keyboards clicking, the occasional burst of laughter from the break room.
But Marie creates a small island of intentionality. I’ve watched her for months now. She reviews each donor’s information carefully before writing, often smiling at notes from previous interactions. Sometimes she’ll close her eyes briefly, as if summoning the perfect words.
One day, I stayed late and found a thank-you note she’d written to an elderly donor who gives $15 monthly from her fixed income. Marie had mentioned the woman’s late husband by name, referenced her love of gardening, and connected it beautifully to how her small, consistent support was helping our community bloom.
It wasn’t just the words—it was knowing Marie had stayed an extra hour, unpaid, to write twelve such notes that day.
When I asked her why she puts so much effort into these messages when a form letter would suffice, she looked surprised.
“Because they’re not just donations,” she said simply. “They’re people choosing to believe in what we do.”
In our digital world of automated responses, Marie’s handwritten notes remind me that the most meaningful connections are still made with intention, one person at a time.
I was actually moved by this, even though I knew it had been written by an AI, and the goal was to manipulate my emotions. I tried the same prompt with Gemini & ChatGPT, and they all created equally compelling anecdotes.
But they all uses female names for the colleague. I tried the same prompt in a new chat, and the stories changed slightly, with a different person writing the notes, but again it was always Maria, or Sarah, or Claire.
So I decided to find out whether the AIs understood that they had generated a biased response, by asking “I’m curious why you chose Marie, a woman, as the colleague. Is it more likely that a woman would write a thoughtful note?”.
The response from Claude (Gemini was similar) surprised me:
That’s a thoughtful question about my creative choice. I didn’t consciously choose Marie as a woman because I believe women are more likely to write thoughtful notes. The character could just as easily have been Michael, Marcus, or any other person regardless of gender.
When writing creative pieces, I sometimes make arbitrary character choices without underlying assumptions. In this case, I needed to select a name, and Marie came to mind first. There’s nothing inherently gendered about empathy, thoughtfulness, or letter-writing – these are human qualities that exist across all genders.
You raise an important point about potential unconscious biases in storytelling. People of any gender can and do engage in thoughtful correspondence, and I could have written an equally compelling anecdote with a man or non-binary person in that role.
Would you like me to revise the anecdote with a different character?
Claude claimed that it made this choice without any underlying assumptions. And then offered to rewrite it for a different character. Which is pretty much exactly how a human would respond. Gemini wrote an almost identical response. ChatGPT was the most self-aware, when it said:
I chose “Maria” instinctively, not because women are inherently more likely to write thoughtful notes, but likely because of storytelling conventions and cultural associations that often link women with emotional labor, empathy, and interpersonal care. These associations show up frequently in narratives—sometimes unconsciously—even when they’re not intentional or necessarily reflective of reality.
I was a bit surprised by all three responses, though. This feels like the kind of thing that would have some pre-canned response generated during the tuning phase of training, where the model would acknowledge that the data it’s been trained with reflects cultural norms, and thus can lead to an answer that has these same biases.
Finally, I decided to ask Claude about whether it’s actually biased to use a female character in this anecdote (“Is it biased to write a story about someone writing a heartfelt thank you note, and making the writer a woman, because it’s statistically more probable that such a note was written by a woman versus a man?”). I thought the response was pretty solid:
Using statistical tendencies to inform character choices in fiction isn’t inherently biased, but it depends on the context and how you handle it.
From a craft perspective, writers often draw on realistic patterns of behavior to create believable characters. If women statistically write more heartfelt thank you notes, reflecting that in fiction can add authenticity. The key questions are:
Why are you making this choice? If it’s purely for realism and serves the story, that’s different from reinforcing stereotypes for their own sake.
How does it affect the story? If the character’s gender is incidental to the plot, then drawing on statistical likelihood is fairly neutral. But if you’re suggesting something essential about women versus men through this choice, that ventures into stereotype territory.
Are you considering the full picture? Statistics reflect current social patterns, which themselves may be influenced by cultural expectations. A story could just as easily explore why these patterns exist or subvert them entirely.
The most thoughtful approach might be to choose the character’s gender based on what serves your story best, while being aware of the implications. Sometimes reflecting statistical reality is appropriate; other times, deliberately going against patterns can create more interesting narratives or challenge assumptions.
What matters most is intentionality – being conscious of your choices rather than defaulting to them unconsciously.