Better #41

Kia ora,

Today’s edition of Better is bought to you by caffeine and slog 💪

Honestly, who else is feeling like the year is streaming past so quickly the edges are starting to blur?

Soon it will be September (WTAF?), which I frankly find a bit rude. After that, new mayors and councillors will be elected in New Zealand, social media will be banned for under-16s in Australia, and then we’ll wrapping up both work and presents in time for Christmas.

What ever happened to time passing a bit more reasonably, or even generously?

Maybe I’m just getting old, or my approaching holiday is putting the squeeze on everything.

Either way, no mucking around today, just straight into it.

Cheers 👋
Seamus Boyer

I haven’t had much to say about Artificial Intelligence.

Honestly, there’s already been plenty of commentary (so much!) and it’s mostly way outside my area of expertise.

But over time, I’ve started to work out where I do have something to add.

Not on the terrifying rise of “AI psychosis,” or the deeply troubling dynamics emerging between chatbots and vulnerable people. Those are critical issues, but far beyond the scope of my little newsletter about comms.

What I am increasingly concerned about is the impact of AI on government communications and content. How AI-generated content can subtly push us to (further) optimise for safe, bland, and generic comms. And how it can create yet another layer of distance between us and the other humans we’re trying to reach.

I’m no Luddite - I think we should embrace good, smart tech tools. And lots of AI tools do seem useful and positive and will makes things be faster and easier and better.

But we also can’t afford to just hand over the keys to these systems that literally do not care about me, you, or what kind of results we get in our work.

And on that note, here’s my take…

👇

‘Show some damn respect’

I actually want to start with an example - one of my favourite posts ever, by any council, anywhere in the world.

It appeared on the Southland District Council Facebook page back in January, and I reckon is pretty much perfect in every way. Accompanied by two images of a trashed toilet, it begins:  “A message to whomever is responsible: I hope next time you are in desperate need of a public toilet there isn't a scrap of toilet paper to wipe your butt. 😡”

It then mentions that it was the second time in two weeks, and that the poor cleaner had to deal with it. 

“To see the community's newest facilities treated like this by some brainless numpty (or numpties if plural is appropriate) is incredibly disappointing. Next time, show some damn respect.”

As fine a post as you’ll read, in my books.

How did it go down? Well, most Southlanders seemed to love it, it got shared well beyond the local area, eventually making its way to me - 1,000km away.

There were a couple of naysayers, notably Gene, who grumped about the tone, which sparked Kate (the OP) to reply that she’d already tried being polite and it clearly hadn't worked. Fair enough, I say.

Now, if Kate had fed her post draft through ChatGPT, imagine what advice she might have received? Well, we don’t need to imagine, I plugged Kate’s caption in and got this:

“Your caption already has a strong voice and will definitely get attention — it’s conversational, emotive, and clearly conveys frustration. But on Facebook, especially for a council page, you’ve got two extra balancing acts:

• Keep community support onside — so people are angry at the vandals, not at the council’s tone.

• Encourage positive behaviour — not just vent about the bad one.”

Some strengths are then offered (“The toilet-paper curse is memorable and relatable”), as well as weaknesses (“‘brainless numpty’ will be loved by some, but can backfire if a commenter thinks the council is stooping to name-calling”).

It then offers a new hook: “To whoever is responsible: May your next toilet visit be tragically short of paper. 😡“

Now, TBF to ChatGPT, it at least keeps the hook (and the curse), but it absolutely strangles the life out of it - by ‘perfecting’ the language and grammar.

And what happens if Kate had then got a bit nervous because of all the pro-ing and con-ing and asked for a second opinion? 

ChatGPT: “Yeah — I think you’re right to hesitate. A council page has to walk the line between being human and staying credible, especially since your audience includes ratepayers who expect professionalism even when you’re annoyed.”

Hmmm, I think I know how this ends up…

One more barrier between us and our audiences

LLMs like ChatGPT all operate in a similar way, anticipating the next ‘token’ based on probability (which is why they get described as “most-likely-next-word machines”). This means they can be great at producing content that’s - unsurprisingly - unsurprising.

Another word for this is ‘safe’, and a big problem with that is that in much government comms, safe is already the default.

In my experience, government comms systems generally aren’t designed to maximise helpfulness or boost audience engagement, but instead to minimise the chances of saying something problematic.

That’s why we see sign-off processes that involve 2, 3, or even 4 layers of approval, and where the process is largely about pulling back or toning down rather than being bolder, funnier, or taking more risks.

When you’re already set up to sand off the edges, adding an AI layer to that sanding process means everything is likely ending up super smooth indeed.

Right now, I believe we need boldness and innovation in our government communications more than ever - particularly on social media, where we’re pitted against an increasingly unhelpful algorithm. So in order to be effective, we need to swap some of our caution for courage, not pile on more caution.

So, why don’t we just change our prompts to produce more ‘human’ content? Fewer em dashes and ‘its not about x, it’s about y’ set-ups? Maybe even purposefully introduce some harmless typos to give it that authentic, harried-human vibe…?

Because, ultimately, AI is one more barrier between us and our audiences. And moreover, it’s a barrier that doesn’t care about us, our audiences, our communities, or any of the work we’re doing.

Let’s remember why we’re here

In scarier moments, I picture the technocrats circling. They’ve been waiting patiently for years, silently shaking their fists in rage at those creative, inefficient comms maniacs. AI is their golden opportunity to optimise and automate communications once and for all, and with every ChatGPT upgrade, they are growing bolder…

Nek minnit, comms and engagement are reporting into the Chief Digital Officer. The result?

• Iron-clad templates for everything

• Every draft gets run through the machine to ensure it’s 100% on-brand

• Optimal risk scores are produced for social posts so we can always be just on the right side of edgy

In comes efficiency and order, out go typos, but also any mention of numpties (both singular and plural). In comes ‘perfection’, out goes the fun and the weird, and that incredible feeling of not being sure if the post will land ok [it will BTW].

In this world, there will be absolutely no place whatsoever for Kate and her perfectly imperfect, perfectly human Facebook posts…

😔

Can AI tools help with elements of our work? Of course they can, and they will - increasingly. Our job is, after all, to be effective, and if AI can help with that, in whatever way, then we should at the very least explore the possibilities.

But let’s just make sure that these tools complement our judgment and instinct - not substitute it - and that we still maintain space for the magical messiness of truly memorable human > human communication.

We don’t need AI tools to help us speak to each other. Never have, never will. And that should probably go double for us comms and social pros. We know our craft, we know our audiences, we know (even if we can’t always articulate it) what works.

And sometimes we’ll get it wrong, but as a mate of mine recently put it, ‘I’d prefer human slop over AI slop any day of the week’.

That’s it for today.

If you’re still reading, I’d love to know where you sit on the use of AI in comms?

Love it and use it everyday? Used to but have grown wary of it? Ain’t into it at all?

Hit reply to this email and let me know. And don’t worry - I will never share your feedback with a soul (or any AI tool without a soul for that matter 😄).

See you soon! 👋