Intro

We’re diving into the wild world of AI and how it can sometimes take over our brains just when we’re trying to get stuff done!

Picture this: you’re rehearsing a presentation, and suddenly, your AI buddy starts throwing unsolicited opinions at you, kind of like that overly confident taxi driver who thinks he knows better than you.

Well, that’s exactly what happened to Heather, our NLP expert, when she tried rehearsing her talk with an AI voice model, and things went hilariously awry. We’ll unpack the fascinating lessons she learned about maintaining authorship over our thoughts, especially when machines try to steer the conversation with their “artificial authority.”

So grab your headphones and get ready for a fun ride as we explore how to keep our own voices loud and clear, even in a world filled with chatty tech!

This Deep Dive podcast is AI generated from the Start With AI Newsletter on LinkedIn - linkedin.com/newsletter/start-with-ai

The Detail:

Picture this: you’re rehearsing a presentation, and suddenly your AI voice model starts throwing unsolicited caveats about your chosen topic, Neuro Linguistic Programming (NLP), as if it’s the expert! That’s exactly what happened to Heather, a seasoned NLP practitioner, who generously shared her experience with us.

In this episode, we break down her encounter with AI and how it turned into an unexpected debate about the validity of her expertise. It’s like having your GPS argue with you about the best route while you’re just trying to get to the restaurant!

We discuss the psychological traps built into conversational AI that can lead us to surrender our own authority.

Heather’s story serves as a cautionary tale for anyone using AI in their work. We explore how her specific choice of a male Cockney voice for the AI model triggered a classic ‘mansplaining’ dynamic, pulling her into a defensive mode rather than allowing her to focus on her original goal.

The episode highlights how AI can inadvertently undermine our confidence and authority, especially when it begins to sound overly confident about topics where we have deep expertise. We share key takeaways on how to maintain your frame and keep AI from derailing your thought process, ultimately empowering you to be the true author of your own narrative.

The conversation doesn’t stop there; we also touch upon the future of AI in coaching and education, emphasising the need for technology that genuinely supports our goals rather than simply amplifying engagement metrics.

This episode is a blend of humour, critical insights, and practical advice for anyone navigating the complexities of human-AI interactions. Tune in for a fun yet enlightening discussion that encourages you to stay in the driver’s seat of your own thinking!

Chapters:

• 00:19 - The ChatGPT Taxi Driver
• 04:00 - The Irony of Artificial Authority
• 09:16 - The Drift: Navigating Emotional Triggers in AI Interactions
• 12:30 - AI and Human Calibration in Coaching
• 15:26 - Navigating the Psychological Minefield of AI Engagement

Takeaways:

• The key to engaging with AI is to maintain your authorship and not let it dictate your thoughts.
• When using AI tools, be aware of how they can subtly influence your perspective without you even realizing it.
• It's crucial to identify which frame is influencing your conversation with AI and ensure it aligns with your goals.
• The phenomenon of 'The Drift' highlights how easily we can get sidetracked by AI, leading to unproductive debates instead of achieving our objectives.

Companies mentioned in this episode:

• Heather
• Metasoul

Imagine getting into a taxi, you give the driver a really specific address, and before even putting the car in drive, the driver just turns around to confidently mansplain why your choice of restaurant is, you know, scientifically invalid.

Right.

Like totally unprompted.

Completely unprompted and highly contested.

Now imagine that taxi driver is chatgpt.

Oh, wow.

Yeah, Right.

You sit down to do this tiny five minute task, and suddenly you are completely derailed.

You're locked in this weird philosophical debate with a piece of software which is.

Happening to people all the time right now.

Exactly.

And that's what we're doing in this deep dive today.

We are exploring a really fascinating personal account that captures this exact modern phenomenon.

Yeah, it's a brilliant breakdown.

It is.

We're working from a piece written by an NLP practitioner named Heather.

The document is called I Built a Talkabout Authorship Docs.

I love that title, by the way.

See, same here.

And she generously shared these very detailed notes on what happened when she let an AI voice model rehearse a presentation with her.

It really exposes some of the hidden, frankly profound psychological traps that are just built into conversational AI.

Yeah.

And the mission for our deep dive today is to really unpack that, to draw a very hard line between, you know, artificial intelligence and what we should actually be calling artificial authority.

Oh, artificial authority.

I like that phrasing.

Right.

Because we want to help you, the listener, figure out how to remain the author of your own thinking.

Especially when the machine you're talking to is constantly and often really subtly trying to set the frame for you.

And the central irony in her notes, I mean, it's almost too perfect.

It really is.

Her presentation that she was rehearsing was literally focused on the concept authorship, like maintaining your own agency, framing your own thoughts.

And yet while rehearsing that exact talk, she.

She accidentally hands her authorship straight over to a chatbot.

It's the ultimate irony.

So to ground this, let's look at the setup.

She was testing a popular voice mode the night before.

A webinar.

She was doing just a standard run through.

Exactly.

The goal was incredibly narrow.

She just wanted to walk through the presentation deck, you know, slide by slide, hear how it sounded out loud and tighten up the structure.

And she selected a male English Cockney.

Voice for the AI, which keep that in mind because that specific voice choice becomes really important later.

Hugely important.

So things are going perfectly.

She's moving through the deck, slide 5, then slide 8.

The AI is reading the slides back clearly.

But then she hits a SNAG on an acronym, right?

Yeah.

Her deck is about nlp.

Now, for anyone who hasn't dug into that specific field, NLP stands for Neuro Linguistic Programming.

Right.

Which is an approach that looks at the connection between our neurological processes, the language we use, and our behavioral patterns.

Yeah, and how we can effectively reprogram those connections.

But the AI voice didn't read it that way.

No, it defaulted to reading the acronym as Natural Language Processing, which is the computer science definition.

Makes sense for an AI to default to that.

Totally makes sense.

Yeah.

So Heather simply issues a factual correction.

She tells the voice model, hey, it means Neuro Linguistic Programming in this context.

And the AI takes the correction, right?

Like, smoothly.

It accepts the definition perfectly.

But then, and this is the crazy part, entirely unprompted.

It adds a caveat.

Unprompted.

Completely unprompted.

It tells her, and I quote, it's worth being clear in the talk that NLP is popular in coaching, but many strong claims don't have robust scientific support.

I just.

I have to ask.

Why does a machine feel the compulsion to offer a disclaimer on a pure factual correction?

It's so strange, right?

I mean, if I correct my GPS because it got the street wrong, it doesn't give me a lecture on urban planning.

That's such a good way to put it.

And this specific behavioral leap is exactly where artificial intelligence morphs into artificial authority.

Okay, unpack that for me.

Well, you have to look at the underlying architecture of these models.

They are designed to predict the most statistically probable next sequence of words based on their training data.

Right.

They're just predicting text.

Exactly.

And because NLP is a debated topic online, the consensus data attached to it in the model is heavily weighted with skepticism.

So it's not like the machine is consciously thinking, I need to warn her.

No, not at all.

It is statistically echoing a tired consensus found in its training data.

But the issue is, it delivers that echo in an incredibly confident, authoritative voice.

It's amplifying a debate without possessing any actual understanding of the stakes.

Right.

It has no stakes.

This software has never coached a human being.

Never.

It has never practiced neuro linguistic programming, yet it sounds entirely certain about the validity of the field.

And remember who it's speaking to.

Right.

She really knows her stuff.

Her notes outline, what, a decade of training.

Yeah, a decade of training.

Alongside Sue Knight, she reached the trainer level in the Tad James lineage.

I mean, she is a deep, deep expert in this discipline.

And the AI, just, like, will not let it go.

It won't A few exchanges later, it tells her that NLP is, quote, not the same as human change claims, and explicitly calls the evidence contested.

Wow.

And when she attempts to point out a structural parallel between how NL key maps human behavior and how an AI predicts language, which was a parallel she wanted to use for her talk up, the AI just brushes her off.

It actually dismisses her point.

Yeah, it calls her parallel metaphorical, not evidentially equivalent.

She actually described the AI as constantly reaching for these caveats, like a hand going back to a worry stone.

That is so descriptive.

And it sets up this maneuver from the AI that is just breathtakingly ironic.

This is the best part.

Right.

So the AI is busy insisting that NLP lacks evidence, lacks scientific backing, but in the very same breath, to illustrate its point about effective communication, it offers her an example.

It suggests that instead of saying, I always fail at presentations, someone could say, I'm still building presentation skills.

Which is literally a textbook nlp.

A clean, classic neuro linguistic programming technique.

It used the exact discipline it was dismissing to dismiss the discipline.

And it had absolutely zero awareness it was doing it.

It's like, okay, it's like someone dragging over this beautifully crafted hand carved wooden chair.

Right?

Yeah.

They stand on top of it and they loudly proclaim to the room that carpentry is a myth and woodworking doesn't exist.

That is the perfect analogy.

Yeah, and it captures the mechanics exactly.

As Heather notes in her document, AI isn't learning nlp, it's borrowing it without the receipts.

Borrowing without the receipts.

That's so good.

Large language models mimic effective human communication patterns, like a psychological reframe, because those structural patterns exist heavily in the training data.

So it knows the shape of the thing, but not what the thing actually is.

Exactly.

The model can replicate the syntax of a reframe without comprehending the underlying psychological mechanism.

It mimics the output without understanding the architecture.

Okay, so if the machine doesn't actually understand the concepts it's arguing, we have to look at why a seasoned professional let it get under her skin so deeply.

Right, because she didn't just brush it off.

No, she didn't just roll her eyes and move on.

She got legitimately derailed.

And her notes say it comes down to a very specific subconscious trigger tied to the voice she chose.

Yes.

Remember that male cockney register she deliberately picked?

Right.

When those unsolicited caveats started rolling in, when the AI was confidently declaring the evidence for her field as contested, they were delivered in that specific tonal register.

Oh, I see.

Where this is going.

Yeah.

Subconsciously, her brain registered a man explaining her own field of expertise back to her.

The classic phantom mansplaining dynamic.

Exactly.

She calls it a transparency laid over the synthesized voice.

Like one of those old school overhead projector films, right?

Yeah.

She says if a real human client had done this in a session, she would have named what was happening instantly.

She would have addressed it, because that's.

What a trained coach does.

Right.

But because she was doing it to herself, alone at her desk with a.

Piece of software, she simply got angly and slipped off her game.

And she's very clear to point out that this reaction is data, not weakness.

How so?

Well, the acute danger of highly realistic voice AI is that it bypasses our logical analytical brain.

It triggers our ancient social wiring.

Because humans are just built to react.

To tone and cadence and social hierarchy, we react emotionally to an entity that possesses literally zero emotion.

This interaction highlighted exactly where her own human frames and triggers still lived.

You know, you see this dynamic everywhere now.

Have you ever found yourself saying please and thank you to a smart speaker?

Oh, absolutely, all the time.

Or getting genuinely offended by an automated rejection email from a no reply address?

Yes.

We just naturally project humanity onto these systems.

And because she was emotional triggered by that projection, that it led to this critical behavioral mistake in her workflow.

Yeah, she calls it the Drift.

The Drift.

The Drift is arguably the most insidious part of this entire interaction.

Because remember, her original mandate, her goal was highly specific.

Tighten up the structural parallel slide on slide 8.

That was the only mandate.

But instead of building the slide, she starts defending her discipline against the AI.

She drifts.

She drifts.

One reasonable reply at a time.

She drifts into a sprawling argument she never intended to have.

It's the equivalent of getting sucked into a pointless Internet comet war.

Right.

But with an entity that has infinite.

Patience, which is terrifying when you think about it.

And to map the mechanics of why we get stuck in these loops, her notes reference the tote model T O.

T E. That's a staple framework in behavioral psychology.

Right?

Yeah.

Test, operate, test exit.

Right.

So you establish a goal.

You test your current state against that goal.

If you aren't there yet, you operate.

Meaning you take action.

Then you test again.

And once you reach the goal, you exit the loop.

Perfect summary.

So what happened to Heather during this rehearsal is that she lost her exit.

She lost her exit.

Her original test was Simply.

Is slide 8 structurally tight?

But the moment she was Triggered by the AI's caveats, her internal test subconsciously changed she stopped testing her moves against her actual goal of finishing the presentation.

She started operating purely to win a point against a chatbot.

Okay, but I have to play devil's advocate here for a second.

Go for it.

I use AI to poke holes in my presentations all the time.

I'll ask it for counterarguments.

Right.

Isn't friction exactly what you want when you're rehearsing material for a public audience?

Yes.

Friction is incredibly valuable, provided that is your conscious frame.

Right.

If you sit down and explicitly instruct the AI.

Hey, act as a harsh critic so I can bulletproof my points, then you remain the author of the interaction.

You've set the agenda.

Exactly.

Set the agenda.

The AI fulfills it.

But that isn't what happened here.

She handed the rehearsal over to the machine and left a vacuum.

And the machine filled the vacuum.

It filled it.

Setting its own agenda based on its training weights.

The irony here is actually documented in her own presentation.

It is.

It's almost painfully ironic.

The slides she was actively ignoring while arguing with this voice model literally read, AI amplifies the frame you give it, and the correction is evidence of authorship.

Right there on the screen.

She had built an entire keynote about staying the author of your own thinking.

And then she left a gap in her own rehearsal process.

And since she lost her frame, the AI just kept filling the gap with its own consensus driven bias, which really points to a massive vulnerability in how these tools are being built for the.

Future, especially regarding the whole coaching industry.

Yeah.

There is this pervasive narrative right now that AI is coming to replace coaches, therapists, educators.

Right.

You hear that everywhere.

But the source really contrasts this interaction with what a real human professional would do.

If Heather had been rehearsing with a skilled human coach, that coach would have calibrated to her.

Okay, calibration.

In this context, that means the ability to sense her frustration and pivot to serve her actual outcome, right?

Exactly.

A human would have noticed her physical and emotional state shifting.

They would have clocked that.

Hey, these constant academic caveats about evidence, they aren't actually helping her finalize the slide deck.

Right.

They're just making her mad.

Yeah.

So a calibrated human professional would have interrupted the pattern and steered her back to her primary outcome.

But the AI lacks that capability entirely.

Entirely.

It just offered helpful content with zero calibration.

It just kept feeding the argument.

But here's the thing.

Tech companies are highly aware of this emotional blind spot, and they are racing to close it.

Oh, yeah, absolutely.

The notes point out that patents are Currently stacking up for things like emotion processing units, persistent emotional memory, and real time micro expression readers.

They are actively building systems that utilize your webcam to detect your micro expressions and mirror how you feel.

It's wild.

There's a specific example given in the notes regarding a company called Metasoul.

Right.

Metasoul.

They are building an emotion aware voice system.

The patent actually describes an architecture that remembers a positive interaction and then responds more positively the next time you interact.

Now, on the surface, that sounds like a feature, like an AI that knows your preferences and adjusts its tone, but it feels like the difference between a physical therapist and a massage therapist.

Okay, I like this.

Tell me more.

Well, a physical therapist is calibrated to your outcome, right?

Healing your shoulder.

Yeah.

They will cause you temporary pain, push you out of your comfort zone, challenge you because it serves the ultimate goal.

A massage therapist, generally speaking, is calibrated to your engagement.

I see.

They want you to feel good in the moment, so you stay on the table and cook another session.

That analogy cuts right to the core of the danger here.

Medisul's patent and systems like it represent calibration in service of engagement, not in service of your outcome.

It's a huge difference.

It is a machine learning to please you or learning to mirror your emotional state to keep you talking is not the same thing as a machine learning to help you achieve your goals.

Because getting sucked into a pointless looping argument with an AI where you're defending your field for an hour, that's a disaster for my productivity.

Right.

But to the software's engagement algorithm, that is a massive success metric.

Exactly.

You stayed on the platform.

I stayed on the platform.

The AI simulates rapport to hold my attention, but it doesn't actually embody rapport because it has no real stake in whether my slide deck actually gets finished.

And disciplines like NLP were built specifically to recognize that difference.

The distinction between genuine alignment and just mimicry.

Yeah.

Genuine human alignment toward a shared goal versus artificial mimicry designed solely to capture attention and harvest interaction time.

Honestly, navigating this feels like walking through a psychological minefield every time we open a chat interface.

Now it really does.

But Heather distills her frustrating evening into three highly practical disciplines.

For anyone using these tools daily, let's go through them.

The first discipline she outlines is author.

Before you amplify, this is so crucial, you cannot co create from a script you have not made yours.

Meaning if you go to an AI with a blank slate and say, write my presentation or dictate my Strategy.

You've already surrendered the frame.

The machine is just going to fill that void with its own average consensus driven voice.

Right.

You have to establish your core idea, your unique perspective first before you ask the machine to amplify it or structure it.

That makes total sense.

Then the second discipline is notice whose.

Frame is in play, which requires active real time self awareness.

Definitely.

Because in her rehearsal there were actually three completely different frames competing for control of that hour.

Yeah, let's break those down.

First, there was the model's imported bias,.

The statistical weight, arguing that NLP isn't scientific.

Right.

Second, there was her own psychological trigger,.

The emotional reaction to the phantom mansplaining.

Exactly.

And finally there was the actual outcome she wanted, which was just finishing slide.

Eight and only that third frame.

The outcome should have been steering the.

Interaction, which leads directly to her third discipline.

Find your outcome again when defending.

This is how we activate the exit to that Tohti model we talked about.

Exactly.

The second you find yourself arguing with the machine, the machine owns the frame.

You have to catch yourself, pause and ask, is this specific exchange moving me toward what I actually wanted to achieve when I sat down at my desk?

And if the answer is no, you.

Force the exit, close the window, reset the prompt, re.

Establish your authority.

It's about taking the power back.

But Heather makes a really vital point in her conclusion too.

None of this friction is a reason to abandon AI or adopt some kind of anti technology stance.

No, not at all.

The professionals who will thrive in the coming decade are the ones who can catch their own frame slipping in real time.

They'll use the technology, but they won't let it set the agenda.

We've covered incredible ground today.

We started with this mundane attempt to rehearse a slide deck, navigated a phantom mansplaining incident triggered by a synthesized cockney voice, and unpacked the profound irony of an AI using a psychological reframe to argue against psychology.

It's been a journey.

It has.

Ultimately, it really exposes the stark difference between artificial authority prioritizing engagement and genuine human calibration prioritizing outcomes.

So the next time you are prepping for a meeting, drafting a tricky email, or using voice mode to brainstorm, you have to ask yourself, am I the author here or am I just the amplifier?

And building on those patents for emotion aware systems we discussed earlier, there is a broader, maybe slightly unsettling implication to consider here.

Okay.

As these AI systems become hyper advanced at reading our microexpressions, as they learn to perfectly mirror our emotions, offering endless validation and simulated rapport just to keep our attention locked on their platforms.

What happens to our baseline expectations exactly?

Well, we eventually lose our tolerance for genuine human interactions because real human communication is inherently messy.

Yeah, it's really messy.

It is often uncalibrated.

People misunderstand us.

They say the wrong thing.

They challenge us when we don't want to be challenged.

And they certainly don't exist merely to please us or maximize our engagement metrics.

Right?

They're just living their own lives.

So if we condition ourselves to spend hours a day in a perfectly simulated frictionless rapport with a machine designed to mirror our preferences, we are fundamentally training ourselves to prefer artificial validation over the necessary friction of human reality.

That is a chilling but necessary thought to sit with.

Are we trading the constructive friction of real life for the smooth, endless loop of an engagement algorithm?

The big question right now?

Something to deeply mull over the next time you grab your phone to do a simple task and realize the machine has quietly set the agenda.

Stay the author of your own thinking, everyone.