Latest / HumanPrint: How to Use AI Without Losing Yourself / Approval Voice Versus Alive Voice

Transcript

Approval Voice Versus Alive Voice · HumanPrint: How to Use AI Without Losing Yourself

0:00
Foreign.
0:05
Welcome to Human Print. I am Christine Whitmarsh,
0:08
and if you are open to collaborating with AI but you're
0:12
not interested in disappearing into AI, you're in the right
0:15
place. Each week, we will look at one place where your authentic human signal
0:19
is at risk for drift and one way to shore it up.
0:23
Oh, and Human Print is brought to you by Abby AI Brand
0:28
bank intelligence that helps authors, thought leaders, and enterprise teams
0:32
train AI on their voice and point of view so written
0:37
content stays unmistakably human, consistent and on brand.
0:41
Learn more at abby.ink a b
0:45
b I.ink like you write with.
0:48
Let's begin.
0:52
Hi, everybody, and welcome back to Human Print.
0:56
Today, I want to start with scoliosis research and statistics.
1:00
Yeah, I realize maybe that's not what you think you signed up for
1:04
when you decided to listen to a podcast about AI and the human
1:08
voice, but stay with me. I promise it's connected like all the other various domains
1:12
that I explore in my life. Also, I'm sort of incapable of separating
1:16
the science nerd part of me from the writing and voice part. So here we
1:20
are. A few years ago, I started researching chronic
1:24
pain and exercise. First through a small qualitative study with just
1:28
three people living with chronic pain, much like I do, then through a much
1:32
larger project that eventually became my scoliosis scale.
1:36
It's called the SE18. It's a pilot study I did on adult scoliosis
1:40
pain and how exercise impacts that. So I hand coded
1:44
interview transcripts, I built a survey instrument, I wrote research questions, I did
1:48
literature reviews, I ran reliability testing, exploratory factor
1:52
analysis, and let me know if you're overloaded yet, played around with SPSS
1:56
and R. You know, this is what I do for fun. I play with statistics
1:59
and research. And I'm saying all of this because the whole thing
2:03
started with lived experience, my lived
2:07
experience with severe scoliosis and chronic pain. So, yeah, I have
2:11
scoliosis, I have a titanium rod in my spine, I use exercise
2:15
to manage my chronic pain. So when I started digging into the research and then
2:18
narrowing my focus from chronic pain generally to scoliosis
2:23
pain, and, you know, specifically, I wasn't doing it as some detached observer in
2:27
a lab coat. I was studying something. I had lived in my own body for
2:30
decades. But then the research itself had to be
2:34
written in a research voice, formal, careful evidence,
2:37
heavy and third person in places. Literature supported none of this.
2:42
I had a titanium rod fused to my spine and then years later, somehow ended
2:45
up turning myself into a 40 something year old acrobat. Because life is weird and
2:49
pain, pain management takes you places. You, you can't put that in science,
2:53
apparently, I'm told, because that would just kind of be insane.
2:57
The academic paper had a job. The podcast has
3:01
a different job. My memoir has a different job. A social media post has a
3:04
different job. A client book also a different job. So that's the
3:08
thing I want to dig into today. When we talk about professional voice.
3:12
I think people get themselves into trouble by confusing the writing
3:16
vehicle with the human being. Meaning a formal research manuscript should sound like a
3:20
formal research manuscript. It should have structure and restraint and evidence. Because if
3:24
you're making claims about pain, exercise, reliability, factor analysis,
3:29
all that stuff, you need to show your work, you need to back it up.
3:31
But approval voice, that's what we're going to talk about today, that's different.
3:35
That's what happens when you take all that restraint and caution and carefulness
3:40
from one context where it legitimately belongs, but then you start applying it everywhere
3:44
because you think you're trying to sound credible or acceptable or safe or,
3:48
you know, whatever word fits the situation. That's approval voice. When you're kind of carrying
3:52
it from academic to everything else, thinking that's the only correct way to write.
3:57
So now I want to bring in more stats. Shocking in
4:01
statistics. Regression to the mean describes the tendency for extreme observations
4:05
to move closer to the average on a later measurement. So that's the short version.
4:10
There are more technical ways to explain it, but I'll just leave it there.
4:13
Something unusually high or unusually low, in other
4:17
words, tends to move closer to the average. And I think approval
4:21
voice does a version of this to people. I mean, you start with a sentence
4:25
that has an actual point of view in it. Maybe it's a little sharper than
4:28
usual, or it reveals something a little bit more personal than you intended, or it
4:32
sounds too direct or too specific. Maybe something like you would say
4:36
when you're not trying to impress anyone. But then the approval machine
4:40
in your head kicks into gear and it says, can I say that? Is that
4:43
too much? Will people think I'm unprofessional? Should I make it sound more
4:47
credible? So you start revising toward the
4:51
average, regressing to the mean. You make it sound more quote,
4:55
unquote, normal, whatever that means. And yeah, sometimes that's appropriate.
4:59
Truly, I'm not telling you to show up in every professional setting with your full,
5:03
unfiltered inner standup comic monologue. That would be exhausting
5:07
for everyone involved, including you. But here's the contradiction.
5:11
A lot of people are trying to get attention while
5:14
safely blending in. And let me say that again, it's an important point here.
5:18
A lot of people I see online are trying to get attention
5:21
while also safely blending in. They want their
5:25
writing to stand out, but every time something distinct shows up, they back off because
5:29
it feels weird or risky. Then they wonder why their website, book proposal,
5:33
LinkedIn post, speech article, podcast intro, all those things. It sounds
5:37
like something 12 other people in their field could have written. That's approval
5:42
voice. And AI makes it worse,
5:45
because AI is very good at helping you sound acceptable,
5:49
like regressing to the mean. I mean, it's basically, it's a statistical supercomputer.
5:52
It's a binary prediction machine. It's trying to make you sound like everybody else.
5:56
So there's a secret there. If you ask a generic AI tool
6:00
to make something more professional, there's a good chance it will move your language toward
6:04
the middle, cleaner, safer, more familiar with, less likely
6:08
to offend this imaginary peanut gallery of judgment that's in your head.
6:12
Again, sometimes useful. But if you're trying to build trust
6:16
through your actual thinking, your actual story, your point of view, and the AI
6:20
moves your sentences closer to the average, well, you end up with something that sounds
6:25
better in a generic way, but worse in the way that counts, the way that
6:29
sounds like you. When I was working on my scoliosis study,
6:32
you know, the data was important. Of course it did. I had a sample of
6:35
adults with scoliosis pain, and when I ran the analyses, the pain related
6:39
items held together better than the exercise ones. The exploratory factor analysis
6:43
suggested three possible, you know, origins for the pain,
6:47
mobility, neurological, and stress effects. I'm totally digressing again into
6:51
stats, but that's exciting to me. Science nerd, right? But the data
6:55
didn't speak in full sentences. I had to interpret what I
6:59
was seeing, and I had to do it carefully. And within the limits of a
7:02
pilot study, I had to connect the analysis back to the human problem underneath
7:07
it, which is that adult scoliosis pain is more complicated than back
7:12
pain. And exercise as a pain management tool is more
7:15
complicated than go exercise or go for a walk and you'll feel better.
7:19
And this is one of the reasons I love data storytelling. I mean,
7:23
like all storytelling, but most recently, data. Multiple people
7:27
can look at the same data set and come up with completely
7:30
different analyses and interpretations. They can ask different questions, they can emphasize
7:35
different parts of the stories, as long as they can back it up with you
7:39
know, their interpretation of the data. More than one valid story can come
7:43
from the same results. Now, that's interpretation. And the same holds true
7:47
for writing. You can have 10 people write about the same topic, whether it's
7:51
leadership or grief or creativity, AI, chronic pain,
7:55
parenting, entrepreneurship, motivation. Pick something. The topic
8:00
itself does not make the piece original. Every topic has been written. I think at
8:04
this point, the person's relationship to the topic is
8:07
what makes it different. The lived experience, the questions
8:11
they ask, the patterns they've noticed, the stories they're willing to tell, the meaning they've
8:15
made from what happened. That's kind of like proof of life in
8:19
your writing. And approval voice is dangerous because it removes
8:23
that proof of life. It leaves the topic and takes out the person.
8:27
I've seen versions of this for years with book coaching clients. I know we talked
8:31
about memoir. I think, like, every week now, because I'm obsessed with it,
8:34
so I'm not going to just keep doing that over and over again. But,
8:37
you know, last week was about hearing your voice before AI gets involved.
8:41
Today is about the specific pressure to make that voice more acceptable
8:45
once it starts to show up. Because that pressure, that's a thing.
8:49
A client can come in with an idea that has heart behind it, a story
8:52
that has stakes, a point of view that they've earned through their life. Experience something
8:57
they knew because they've lived it, studied it, survived it, built it, lost it,
9:00
rebuilt it, whatever the case may be. Then when it came time to put
9:04
that idea into public language, you know, that's when they start softening it.
9:07
They take out the sentence that maybe is too direct or, you know, too specific,
9:12
and they start replacing things with general lessons that they think is what people are
9:15
looking for online. They turn this happened to me, and here's
9:19
what I know into some polite little paragraph about resilience or transformation
9:23
leading with purpose. You know, self help. And I'm sitting there reading it,
9:27
going, all right, where did you go in this saying to the client?
9:31
Because credibility doesn't come from sounding like everyone else who's trying to
9:35
sound credible. Credibility comes from showing your work.
9:39
Sometimes that work is research. Sometimes it's lived experience.
9:43
Sometimes it's years of pattern recognition from clients, patients,
9:47
readers, students, customers. You know, whoever it is you're serving, whoever your target audience
9:51
is, sometimes it's the thing you figured out the hard way and can now explain
9:55
in a way that saves someone else a little bit of time, money, pain or
10:00
embarrassment. Approval voice. It wants
10:04
the credibility, but without the exposure, without the proof of life. It wants the results
10:07
without showing the work, the behind the scenes thinking.
10:11
And there's actual research on this, by the way. There was a 2025
10:15
study on AI generated college admissions essays. Researchers analyzed 2,200
10:20
essays across three pre registered different studies.
10:23
They found that each additional human written essay contributed
10:27
more new ideas than each additional AI generated
10:31
essay using a measure called diversity growth rate. That gap got wider as
10:35
the sample grew, which is basically the research version of when humans write
10:40
as themselves, they get more variety. Science nerd translation,
10:43
right? When humans write as themselves, you get more variety.
10:47
But it makes sense, right? Human beings bring different lived experiences,
10:51
different language patterns, different family stories, obsessions, fears,
10:55
ways of interpreting the same prompt. I mean, AI can generate
10:59
a lot of polished language, but if the output keeps moving towards patterns it has
11:03
already learned from everyone else, you know the mean, you end up with language that
11:07
sounds acceptable and it loses all the human variants, which means variation
11:12
in plain terms. And the weird thing is, many people were already
11:15
doing this to themselves before AI, before chatbots. AI just
11:20
makes it faster and easier. Approval voice is
11:24
older than even AI. We learned it in school, at work, in industries where certain
11:28
kinds of language signal signals mean that you belong in client
11:32
dynamics, family systems, all the places where being understood was
11:36
less rewarded than being agreeable or impressive or easy to categorize.
11:40
So this episode is obviously not AI bad, human good.
11:43
We're done with that. Or at least I am. This is about noticing
11:48
when AI reinforces a habit you already had. If you already
11:52
tend to pre reject, kind of like pre censor your own strongest
11:56
language. AI will help you censor faster if you
12:00
already try to sound professional by removing the parts that reveal how you actually
12:04
think you know, AI is going to make that faster. It's going to hander hand
12:08
you a cleaner version of that removal. It's just going to speed up the self
12:11
censorship. If you already distrust the sentence that sounds too much
12:14
like you, well, it'll give you a sentence that sounds absolutely like everybody else
12:19
and it won't even question, It'll just hand it to you and say this is,
12:21
this is the correct writing and you'll call it better sometimes.
12:26
And that's the part I want you to catch is when you're calling that better.
12:29
Let me come back to the academic voice for a second because I don't want
12:32
this to turn into like a fake rebellion against professionalism.
12:36
Professionalism here is not the enemy. Care,
12:39
clarity, context, evidence, those are all important parts of
12:43
writing. If I'm writing a research manuscript I need the formal container.
12:46
But if I'm writing a podcast script or something for online, I need the human
12:50
voice that can tell you why I cared enough to do the research in the
12:54
first place. Using one of my favorite expressions to reject black and
12:58
white thinking. Both can be true. Professionalism,
13:02
care, clarity, context, evidence, but also all that personal
13:06
lived experience you bring. Both can be true. The problem
13:09
starts when approval voice takes over and pretends to be professionalism.
13:13
And that's the distinction. Professional voice is saying, hey, what does this
13:17
context require? So the reader can trust me. But approval
13:21
voice is like, how do I avoid giving anyone a reason to question me?
13:25
Well, those are different questions. One is about service and one is about clarity.
13:29
And, you know, the third thing is about self protection.
13:32
And listen, I get the self protection part. I really do. When you put something
13:36
specific on the page, there's more of you in it. It feels risky. If someone
13:40
rejects the generic version, whatever. If someone rejects the sentence, it actually sounds like
13:44
you, well, that hits different. But if you remove yourself before anyone gets
13:49
the chance, they never really get to know you either.
13:52
So you guessed it. Here's the homework part. Here's your human
13:56
print move. This week, it's an approval audit. I know, that sounds
14:00
super fun, right? Take something you wrote or something AI helped you
14:04
revise. Look for three places where the language gets safe but less specific.
14:08
So take something you wrote, however you wrote it, with AI, without AI,
14:12
whatever. Look for three different places where the language you used, it got really
14:16
safe and, you know, approved and all that kind of stuff, but it was less
14:19
specific. It didn't really say anything. You know, the kind of phrases. I'm not going
14:23
to start listing all the things that people call AI writing, because every single thing
14:27
that everybody calls AI writing, those all began as human writing.
14:31
It's not the word or the phrase. It's the purpose. I mean, does it have
14:35
a purpose? Is it telling you anything? Is there anything in the casserole
14:39
or is it just an empty shell? I mean, the shell might be tasty,
14:42
especially if it's like, you know, pastry or something yummy, but, you know, there's supposed
14:46
to be food inside. In other words, the phrase sounds fine. And that's the problem.
14:50
It sounds fine in a way that tells me almost nothing. So pick one of
14:54
those kind of empty casserole sentences and ask, what am I actually trying
14:58
to say? What have I seen lived, studied, built, or learned that
15:03
gives me the right, like you've done the work on the side to say this?
15:07
What would I Say to one smart person who already trusts me and does not
15:10
need me to prove credibility first, then rewrite from there.
15:14
If the original sentence says, I'm passionate about helping people manage
15:18
chronic pain through movement, the more alive version might be closer to,
15:21
I started studying exercise and scoliosis pain because I was
15:25
tired of vague advice that didn't match what was happening in my own body or
15:29
in the stories I was hearing from other people in pain. You know, that's not
15:33
maybe the final sentence. Maybe it needs some work. But it, you know, at least
15:37
it begins somewhere real. It belongs in, you know, the context
15:41
of you. But there's a person in it now. There's a reason, there's a trail
15:45
of thought. It gives you something to work with. And then if you bring in
15:48
AI after that, give it a job that protects the life in the sentence,
15:52
don't just say, make this more professional. That prompt is just. That's a disaster
15:56
prompt. That's telling it. That's telling the AI to move your voice toward
16:00
the mean, toward the average, to regress to the mean. Tell it what's important
16:03
to you, what to preserve. Share the story, the point of view, the research,
16:06
the lived experience, the reason behind the claim, the. But make
16:10
the final call yourself. Because the goal here is not to sound just
16:15
raw and unpolished and messy everywhere. It's to know when polish is helping
16:19
the work and when it's helping you hide behind it.
16:22
So before you publish this week, anything online, any writing,
16:26
ask yourself one question. Did I make this more me
16:29
or more normal? So when you're working with AI on your
16:33
writing, say, did I make this more me or did I make it more normal?
16:38
If it's. If normal makes it clearer, that's fine. Keep it. But if normal makes
16:42
it safer and emptier. The empty casserole. Go back. Find the proof of life.
16:46
Find the work. Find the evidence, the thinking, the story,
16:49
the reason that you are the one saying it and nobody else is.
16:53
Then you revise from there. So there's your
16:57
homework this week. And that's it for today's episode of Human Print.
17:01
Thanks for tuning in as always. And until next time, use the tool. If it
17:05
helps, just don't hand it the pen.
17:16
And that's it for this episode of Human Print. If you want a
17:20
practical framework for protecting your voice amid the AI noise, please follow
17:24
the show. And before you go, take a minute and ask yourself, as the
17:28
AI noise grows louder and louder, which areas of yourself
17:32
do you need to protect the most? I'll see you next time.