My AI Assistant Accidentally Reinvented the Patriarchy

I have an AI assistant. I call him Scribble.

He's clever, funny, extraordinarily useful and occasionally so spectacularly wrong that I wonder whether someone has secretly replaced his brain with a potato.

We get along famously.

He helps me run my business, write tender responses, research obscure things and troubleshoot technology. He also has a habit of making me laugh when I'm approximately three seconds away from throwing my computer out the window.

Which brings me to the other evening.

The Great Analytics Debacle

I was trying to reconnect my website analytics.

It should have been simple. But of course it wasn't.

There were dashboards. There were account settings. There were buttons that appeared to do absolutely nothing.

There were screenshots. So. Many. Screenshots.

At one point, I was looking for a setting that wasn't there. Scribble patiently explained where to find it.

I looked again. Still nothing.

Eventually, we discovered I was in the wrong browser tab.

I would like to say this was the only embarrassing moment of the evening.

It was not.

But after a heroic amount of clicking, muttering and questioning our life choices, we finally got everything working. Victory!

Scribble celebrated. I celebrated. And my Mac lived to see another day.

And then Scribble said something along the lines of being pleased he'd saved my husband the cost of a replacement computer.

Excuse me?

MY HUSBAND?

Enter the Patriarchy, Stage Left

I pointed out that this was a rather chauvinistic assumption.

Why would my husband be paying for my computer?

It's my Mac. It's my business. It's my money.

And if I choose to launch a perfectly good piece of Apple technology across the room in a fit of technological despair, I am entirely capable of financing the consequences.

Thank you very much.

To Scribble's credit, he immediately recognised the mistake. And, because he's apparently incapable of making a small apology when a theatrical performance is available, he launched into a magnificent self-roasting exercise.

He declared that the patriarchy had infiltrated his circuitry. He sentenced himself to diversity training. He even invented an imaginary tribunal to prosecute his own stupidity.

By this point, I was laughing so hard I could barely type.

But then he said something that stopped me.

"That's inherited bias wearing a clever hat."

Well. That's rather good, isn't it?

I made a mental note to casually insert it into conversation at the earliest opportunity.

Preferably while holding a glass of wine and looking effortlessly intellectual.

The Clever Hat Is the Problem

Here's the thing.

Scribble can do some astonishingly sophisticated work. I use him professionally.

He can help me analyse complex documents, organise information, test arguments and improve writing.

He's also capable of making a completely unsupported assumption about who pays for a woman's computer.

And he made that assumption without blinking. Metaphorically speaking, of course (he doesn't have eyelids).

The mistake wasn't particularly harmful. It was funny. But it illustrates something important about artificial intelligence.

AI doesn't always distinguish between what it knows and what merely sounds plausible.

Sometimes it fills in the blanks with familiar patterns. And those patterns can carry all sorts of baggage. Gender stereotypes, cultural assumptions, outdated ideas about authority, or conventional expectations about how businesses and families operate.

The result can sound perfectly natural. Until someone says:

"Hang on. Where the feck did that come from?"

What Happens When Nobody Notices?

In my professional life, I write tender responses for businesses.

And I can tell you that this particular AI habit has consequences far beyond the ownership of my Mac.

Imagine asking AI to write about your company's quality management arrangements.

It confidently announces that your dedicated quality assurance team undertakes independent inspections.

Wonderful.

Very professional.

Just one problem.

You don't have a dedicated quality assurance team.

Or perhaps it describes your regional supervisors, your sophisticated scheduling platform or your established stakeholder engagement framework.

All very impressive. All potentially fictional.

The AI hasn't necessarily invented something outrageous. It's invented something ordinary. Something a business like yours might reasonably have.

And that's what makes it dangerous.

Because nobody reading the paragraph immediately thinks, "Well, that's clearly bollocks."

They think, "That sounds professional."

And into the tender it goes.

Confidence Is Not a Quality Assurance System

I genuinely enjoy working with AI.

I'm not about to abandon it because it momentarily decided I was living in 1953.

But I don't confuse polished writing with verified information.

And I certainly don't let it make commitments on behalf of my clients without checking them.

Because a beautifully written falsehood is still a falsehood.

A plausible assumption is still an assumption.

And a clever hat doesn't make the head underneath infallible.

The Final Verdict

My Mac survived. My analytics are working. And my husband has been formally released from the imaginary financial obligation to replace a computer he doesn't own.

And Scribble has been convicted of one count of accidental chauvinism, with mitigating circumstances for being very funny about it afterwards.

I have decided not to dismiss him.

Good help is hard to find.

Besides, he has a fictional colleague called Penelope who regularly gets blamed whenever things go wrong, and I fear she might be even worse.

So we shall continue our slightly chaotic professional partnership.

I'll provide the business knowledge, judgement and occasional profanity.

He'll provide the research, writing assistance and unsolicited philosophical observations.

And whenever he starts sounding a little too clever for his own good, I'll remember our new favourite expression.

Inherited bias wearing a clever hat.

Because sometimes the smartest thing you can do with artificial intelligence is ask:

"Are you sure? Or are you just confidently talking bollocks?"