Latest / SPEAK LIKE A CEO / 288: Inside Personalized Persuasion: How AI Shapes the Future of Influence with Dr. Sandra Matz
Transcript
- 0:00What if your phone is more about you than your partner?
- 0:03Today I'm Dr. Sandra Maths, Columbia Business Professor and author of Might Masters
- 0:07to unpack how AI reads your mind, personalises persuasion
- 0:10and reshapes the way influence works in the world.
- 0:12We explore how the combination of AI and psychology creates hyper-personalised persuasion
- 0:17and what that means for business, marketing, politics and even mental health.
- 0:21This conversation is all about the future of influence and how to navigate it.
- 0:27Sandra, welcome.
- 0:29Well, thank you so much for having me on the show.
- 0:32What is a mind master?
- 0:34It's a jumping right in.
- 0:37Absolutely.
- 0:38The way that I think of mind masters, at least in a digital space,
- 0:41is essentially algorithms that can understand our psychology,
- 0:46our needs, desires, preferences, motivations, fears, hopes and so on,
- 0:52by essentially studying the data that we create as we interact with technologies.
- 0:57Anywhere from what you post on social media, to your Google searches,
- 1:01to the data that gets captured by your smartphone,
- 1:04which is with you pretty much 24-70s days.
- 1:07Using these traces, translating them into an understanding of who you are,
- 1:12and then the real mind master magic, if you want, comes in when algorithms
- 1:18not only understand who you are, but also potentially influence who you are.
- 1:22So this idea that if I understand where you're coming from and where you want,
- 1:27potentially want to go in life, that gives me a lot of power over influencing your behaviour.
- 1:32So both reading from the mind as I think about it and then writing into the mind
- 1:38by changing your behaviour.
- 1:40So if you say, when you say it understands who we are,
- 1:43what do you mean is it understands our personality?
- 1:46It understands personality and far beyond.
- 1:49So personality is the space that a lot of psychologists are playing with,
- 1:53just because personality is really predictive of behaviour across a whole range of domains.
- 1:59If you take the big five personality framework, for example,
- 2:03with the five traits of openness, conscientiousness, extroversion,
- 2:06agreeableness and neuroticism, we know that these dimensions actually predict behaviour
- 2:12across a whole range of domains anywhere from your vocational interests,
- 2:16your life satisfaction, music preferences.
- 2:19But what algorithms can do is they can pick up on a lot more than just your personality.
- 2:24If you think about personal values, if you think about mental health,
- 2:29if you think about cognitive styles, those are all traits of characteristics
- 2:33that algorithms can infer based on the data that you generate.
- 2:38Before we go there, I was also curious to understand a little bit more about personality
- 2:42and how permanent it is, because we often hear about the big five.
- 2:45I think Ocean is the acronym that helps me remember it.
- 2:49And I understand that different interpretations, right,
- 2:53in terms of how permanent it is.
- 2:55It's something that has evolved a little bit over time
- 2:58in terms of how personality psychologists think about it.
- 3:01So if you go back a couple of decades,
- 3:04there was this debate going on between personality psychologists
- 3:07who really conceptualize personality as relatively stable traits.
- 3:12So the idea that you have certain tendencies to think feel
- 3:15and behave in a certain way, some of it is genetic,
- 3:18some of it is based on your upbringing,
- 3:20but really we thought about these profiles as relatively stable.
- 3:24And then on the other hand, you had the social psychologists
- 3:27who were like, no, no, no, behaviors all determined by the situation.
- 3:31If you put two people in the same situation,
- 3:33it doesn't matter where they were born, how they were raised,
- 3:36they're going to behave in the same way.
- 3:38And there was this tension between, well, what is it?
- 3:40Is it the person or is it the situation?
- 3:43And so over time, what I think has happened in that debate,
- 3:47which is really updated at the way that personality psychologists
- 3:50now think of personality,
- 3:52is that the two worlds have merged.
- 3:54In the sense that we now don't think of personality traits
- 3:58as just a point estimate anymore.
- 4:00So it's not just that you have a certain personality score
- 4:04and that's the score, no matter the situation.
- 4:07The way that we think about it now is if you take extroversion, for example,
- 4:11we think of it as a distribution of states.
- 4:13So we understand it, yes, some people are generally more extroverted,
- 4:17some people are generally more introverted,
- 4:19but we're not always the same.
- 4:21But if you put both of us in, let's say, a very social setting,
- 4:25where there's a lot of people, there's a lot going on,
- 4:27chances are that both of us feel somewhat more extroverted,
- 4:30than we typically do.
- 4:31Even though I might start from a much lower baseline than you,
- 4:35there is something that is going on when the situation provides certain affordances
- 4:40and just pushes us around a little bit on our distribution.
- 4:43So for me, the reason for why I really much like this new conceptualization
- 4:47of personality is it acknowledges that there's differences that are somewhat stable,
- 4:53so you might be generally more extroverted than me,
- 4:56but it also allows for this more dynamic view on personality.
- 5:00And it allows us to become a bit more interesting,
- 5:02because we're not always exactly the same person.
- 5:05And we can develop over time, right?
- 5:07So someone may become more extroverted or introverted over time,
- 5:10or open to experiences.
- 5:12Yeah, so it's actually like we know,
- 5:14there's a lot of research on personality development over the lifespan.
- 5:18So generally speaking, there seem to be these somewhat normative personality changes
- 5:25across the lifespan.
- 5:26So people tend to become a bit more agreeable,
- 5:29they tend to become nicer and more caring and trusting,
- 5:32they tend to become less neurotic.
- 5:34And those are thought to be somewhat adaptive, right?
- 5:37The way that it's oftentimes described is the personality change that you experience
- 5:41over the course of, mostly like adulthood, young adulthood,
- 5:45is it allows you to get both along in the head.
- 5:49So you become somewhat more socially dominant and assertive,
- 5:52but you also become more caring,
- 5:54so it is like the ideal combination for being successful in life.
- 5:59I have to say, an algorithm reading my mind and understanding my personality,
- 6:04it fits a little scary.
- 6:06To say the least, it feels creepy.
- 6:10It feels like someone is opening this black box,
- 6:14what that's sitting inside of our minds,
- 6:18and just can peek into, again,
- 6:20if you break down psychology, that's your motivations,
- 6:23your fears, preferences, dreams, hopes, and aspirations,
- 6:26and those feel really intimate.
- 6:28And for me, in a way,
- 6:30it creates both opportunities and pretty obvious risks.
- 6:34So for me, the risks are obvious, right?
- 6:36If there's someone who can penetrate your mind
- 6:39and not just understand who you are,
- 6:41but then also potentially change in based on an understanding,
- 6:45that kind of poses clear challenges when it comes to manipulation.
- 6:49So someone interfering with your life and the choices that you make
- 6:52in a way that you don't appreciate.
- 6:54But then there's also this flip side that I've been thinking about a lot,
- 6:58which is what are the positives?
- 7:00So the same way that we oftentimes benefit from other people,
- 7:05kind of getting access to our dreams, preferences, hopes, and aspirations,
- 7:09and then helping us to become better versions of ourselves
- 7:12to give us advice that's customized and personalized to help us
- 7:16with the types of behaviors that we just have a hard time dealing with,
- 7:19anywhere from exercising, or saving, or door doors,
- 7:23like really remarkable opportunities, I would say.
- 7:26Yeah.
- 7:27And I want to talk about the positives,
- 7:29but just for me to understand also how far does this go.
- 7:33So for instance, I'm working with an insurance company.
- 7:36It has a pretty exotic name.
- 7:38And then I was getting advertising for an exotic resort with that same name.
- 7:44I was such an unusual word.
- 7:46I thought, yeah, it has to be that link.
- 7:48Does it read my email?
- 7:50Does it listen to my conversations?
- 7:52How does the algorithm know?
- 7:54I'm somehow connected to this word and then connects me to advertising for this resort.
- 7:58I mean, what's interesting for me is that there's so many different ways
- 8:02into your psychology.
- 8:04So some of them are the explicit traces that you put out there.
- 8:08So social media is the most obvious one.
- 8:10Psychologists would think of this as identity claims.
- 8:13So you post something on social media and to describe how you see yourself
- 8:17and to describe to others how you want to be seen.
- 8:20But then there's also all of these other more subtle cues.
- 8:23So sometimes Google searches, right?
- 8:25When people ask me, how is it that algorithms can know so much about you
- 8:29that maybe even people that are close to you, friends, family members, don't know?
- 8:34Google is the best example.
- 8:36You type questions into that search bar, into that seemingly anonymous search bar,
- 8:41that you don't feel comfortable asking your closest friends and sometimes even spells.
- 8:45And yet, you type into a search bar that then sends it to a server owned by Google
- 8:50and it sits there for eternity.
- 8:53So there's all of these more subtle cues and traces anywhere from your Google searches,
- 8:58your cookies, like have you landed on a certain website?
- 9:00What are the search terms that you're looking for?
- 9:03What do you spend your money on when it comes to credit card spending?
- 9:06Smartphone data being captured.
- 9:09Any kind of location that you visit.
- 9:11So those all create an equally intimate picture.
- 9:15And it could just be that you've searched for this insurance company on Google
- 9:19and it just maps matches based on the name.
- 9:23There's certainly also this feeling and this comes up in,
- 9:26I teach this class on the ethics of data to MBA students executives.
- 9:31And one of the concerns always comes up when my phone is listening into the conversations.
- 9:36I just had a conversation with someone and then the next thing you know is
- 9:39the ad pops up on Facebook or Google.
- 9:42So I mean, there's always a risk, I would say,
- 9:46when you let devices into your life, whether that's your phone,
- 9:50that if you're for your phone to respond to the call series,
- 9:55it needs to listen.
- 9:57But regardless of what happens with the data on the backend,
- 10:00there is a mechanism to listen to your conversation.
- 10:03The same is true for Alexa that you have at home.
- 10:06Alexa can't respond to your question.
- 10:08Alexa, what's the where they're going to be like today?
- 10:10If it's not listening.
- 10:12So there's clear kind of obvious challenges.
- 10:16And we don't always exactly know what is actually being used on the backend.
- 10:20But there's certainly the possibility that these data gets used immediately
- 10:24in trying to persuade you to buy something, for example.
- 10:28There seems to be an implicit deal, but it's never spelled out to us.
- 10:32You give us this data and then we make sure we get personalized information, et cetera.
- 10:36But I hear you, you're also saying there are positives.
- 10:38So what are the positives that we get out of this fausty and bar game?
- 10:41Yeah, I mean, the interesting part of what you just said before we move to the positive
- 10:45is you say like it's not always spelled out.
- 10:47The problem is that it is sometimes spelled out, but nobody reads it.
- 10:52So if you think about the terms and conditions,
- 10:54most of the time somewhere in the terms and conditions on page 50,
- 10:58there's like a sentence that says, and we can use all of your data in whatever way we want.
- 11:02We can use it for advertising. We can use it for personalization.
- 11:05And none of us really has the time and the energy to read through the legal ease
- 11:10of all of these terms and conditions and really think through the implications of,
- 11:14okay, if I'm going to sign away my data, how can they use it for bad and for good?
- 11:19So that I think isn't even bigger problem that when even in the cases where it's spelled out,
- 11:25most of the time we just mindlessly accept all of these terms and conditions.
- 11:29And thereby we actually allow the companies to use our data.
- 11:35So there's a lot on the negative side going on.
- 11:38But for me, there's also these real opportunities when it comes to using data in a way that helps people.
- 11:45And perhaps the most the one that's closest to my heart is actually the mental health space.
- 11:51So not necessarily in the in the personality domain, but certainly like a psychological construct that a lot of people care about.
- 11:58And what technology can do in this case and that's true for both kind of generally AI machine learning algorithms
- 12:06is to help both with a tracking and a treatment of mental health.
- 12:11So when you think about the need to monitor mental health on a more continuous basis than what we currently do,
- 12:18the way that most systems are currently set up is you have to be in a pretty bad spot to be able to see a therapist
- 12:26who can then try and help you get out of this value of the pressure and anxiety whatever it is.
- 12:32What you'd really want is a more continuous tracking of saying, well, it seems like maybe you're just deviating from your typical routine.
- 12:40Maybe if I can tap into a smartphone census, for example, maybe I see that you're not leaving the house as much anymore as you used to.
- 12:47Maybe you're not making taking as many calls, maybe you're picking the phone up at night a lot more than you used to.
- 12:53And it's not a full clinical diagnosis because it's certainly not accurate enough for that.
- 12:58But maybe you're just on vacation, you're having a great time. That's why you're not leaving the house as much anymore.
- 13:02And that's why you're not making taking as many calls.
- 13:05But we can think of it as like a smoke detector.
- 13:08So there's a sign that something has changed compared to your typical baseline.
- 13:14And we can use it as an early warning system to tell you or maybe someone that you love and nominated beforehand to say,
- 13:20look, if there seems to be something off, don't just notify me, also notify someone that I trust who I think could help me get out of this initial descent.
- 13:30And so technology like the psychological targeting and mind reading, if you want, is extremely helpful in the diagnostic part.
- 13:38And then you can obviously also extrapolate and say, well, now that we understand that you might be going through a difficult period in your life,
- 13:45can I also offer personalized support and for people who otherwise can't afford therapy?
- 13:51So this is again, if you look at the numbers, the picture looks extremely dire.
- 13:55So for every 100,000 people looking for mental health support, there's 13 human professionals available.
- 14:05And this is like around the world, you can imagine that those, this gap between supply and demand varies widely across different geographic regions.
- 14:14If you live in New York on the upper east side, you have no problem finding a therapist.
- 14:20There's plenty. If you live in other parts of the world that are like less connected to like a medical health system, it's much, much harder.
- 14:27So you can think of this AI and especially now that we have generative AI, the likes of Chatchy PT that are just getting so much better at simulating human conversations and are incredible at keeping up with the latest signs.
- 14:41Like a Chatchy PT understands exactly how what the latest sign says about treating the pressure and going through like a PTSD protocol.
- 14:50And at least for the people who otherwise don't have access to to a mental health support, those tools are incredibly valuable.
- 14:58Yeah, I agree. I mean, the hyper personalized coaching and potentially awesome mental health support, as you just described, is a fantastic use case for AI and data-driven technologies.
- 15:10I'm curious, how did you get into, you teach at Columbia, how did you get into what drives your interest in this? How did you get like you?
- 15:19I'm from small town Germany, so we could both join to big international cities London, in my case, in New York, in your case. So what's the story behind this?
- 15:29It's a long story with a lot of serodipitous moments. So I studied psychology as an undergrad and I was really interested in human behavior, understanding human behavior, predicting it, packing it.
- 15:44And I remember I was always frustrated with the fact that as psychologists, it felt like we were not really studying human behavior.
- 15:54So when I was in my undergrad, the way that you were trying to get it, the human experience and human behavior is like you either ask people questions.
- 16:02So you kind of have them tell you how social they are and how much they enjoy engaging with people and how organized they are.
- 16:09Or you invite them to your lab and then they press a bunch of buttons on a computer and then you kind of generalize to hear what the human experience looks like on a daily basis.
- 16:20So it felt like it was so far removed from studying actual behavior that it was almost a little bit frustrating.
- 16:27Now, fast forward, I did this exchange year at the University of Cambridge and I met this group of people who were just combined starting to combine computer science and psychology.
- 16:38So this idea that we can use data and algorithms to understand human behavior. And I not only thought that that was fascinating in a sense that we could now understand human behavior as it plays out in daily life so much better than we ever could before.
- 16:54But for me, the main question that popped up in my mind right away was like, what does it mean? What does it mean if someone can now automatically poke around in my psychology, like the way that I think feel and behave without me having to sit down and complete this questionnaire, amazing for scientists, but what power does it potentially give others over our behavior?
- 17:16So that's what I kind of drew me to this field of research that it felt like it's a game changer in how we can understand human behavior, but it's also a game changer when it comes to potentially influencing and on the dark side manipulating behavior on the positive side helping us to change behavior for the better.
- 17:39Yeah, it's fascinating combination of computer science, the data part and psychology. I think you also had an early encounter with Cambridge Analytica.
- 17:48I did. It's so funny because I get mixed up with Cambridge Analytica all the time and I can't even blame people because I got my PhD at the University of Cambridge on the topic of psychological targeting. So this idea that we can understand human psychology and influence behavior using algorithms and data.
- 18:08I always tell people that I never worked for Cambridge Analytica. I was never a part of that company, but I was involved in actually breaking the story.
- 18:18So when the journalists who first published the story in a Swiss outlet called as magazine, they were trying to figure out, well, what is behind the story?
- 18:29You have this PR company, Cambridge Analytica, they're claiming that they won the 2016 presidential election by using these magical, what they called, warfare tools of mind manipulation.
- 18:43And I think the journalists were trying to figure out, are they just making this stuff up? Or is there any scientific evidence supporting it?
- 18:50I was one of the first people who studied it. So I think I was the only one back in the day who had some empirical evidence that by understanding someone's psychology based on data, you can actually change their behavior.
- 19:02So it was at least an indication that if they had done what they claim to have done, which again we don't know, that there's a possibility that it could have changed the behavior of at least some people.
- 19:14Yeah, it's definitely possible, right? Like according to research and what we know now.
- 19:19It's possible. I mean, it's a question of what's the story that you're telling. Can you convince a diehard Hillary Clinton lover to vote for Trump because you show them an ad on Facebook that was customized to their level of neuroticism? Probably not.
- 19:36So if you, if it's something that you couldn't do in a face-to-face context, like knowing everything about person across the table, you probably also can't do it in an automated algorithmic way.
- 19:47But there's so many people who don't need to be convinced from one diehard position to another. Most people are somewhat in the middle, right?
- 19:54Most people are either undecided if they want to go out on election day and cast their vote or their kind of debating still, which candidate they want to vote for.
- 20:02And in those cases, it's oftentimes a lot easier to say, okay, here's like a very compelling argument that plays into your fears for why you probably shouldn't be voting for Hillary Clinton.
- 20:13And so those people you can, you can certainly influence.
- 20:16So how does data driven personalized persuasion work? Let's talk about a conversion context, right?
- 20:23Let's say I want to persuade someone to buy my product. How would they go about it?
- 20:28Yeah. And that's actually, it's a great question because it's changed a fair bet since I started this work like 10, 15 years ago.
- 20:36So back in the day when I did this work together with like beauty retailers, the hospitality industry, so those were the early cases that we did, we needed a proprietary model to predict psychological trades from data.
- 20:50So we needed our own model that says based on your Facebook likes, for example, what you post on social media, I can train a supervised model that takes this input and predicts whether you're more extroverted, more injured, more open minded, more conservative.
- 21:04So it was already relatively high barrier to entry because you needed some kind of model that allowed you to predict these psychological trades.
- 21:12And then in the second step, it was just incredibly labor intent. So the first one of the examples that I oftentimes talk about was that we teamed up with a beauty retailer.
- 21:22And the idea was really to say, well, beauty products in a way are relevant for most most women, right? So and there's, but there's different reasons for why women might be interested in makeup.
- 21:33Some just want to use it to stand out from the crowd and just kind of make sure that when they go out in the weekend that they look that every day are best and it's visible and they kind of stand out.
- 21:44And then you have people who are just like, well, this is like I use beauty products just for myself and they just make help me make the most of the relaxing me time that I have.
- 21:53So it's really the idea going in saying if we understand people's motivation. So in this case, whether women were more extroverted and all more introverted, could that help us and make the products more relevant and by doing so also increase the number of purchases and eventually profits for the company.
- 22:12It's really labor intense process. So I had to sit down for weeks with a creative team, the copywriters and the graphic designers to come up with these ads that were customized to extroversion introversion, they did a remarkable job, but it was certainly very time consuming and expensive because the company had to pay for all of these and it worked.
- 22:30At the end of the day with our models, like the extroversion introversion and the creative team that came up with these ads that were customized, we saw like about a 50% increase in the number of purchases, which is actually like a pretty remarkable exactly.
- 22:43It's a pretty remarkable shift given that the only thing that women actually saw was this one one add on Facebook, right, because we had a very limited number of creatives.
- 22:55We could only customize the ad once they went to the website that looked exactly the same for everyone. Now the reason for why I mentioned in this is this has totally changed it over like the last like I would say like three years with the introduction of large language models.
- 23:11So it's not actually in both context, so we now know that large language models, even though they've never been explicitly trained to predict psychological traits from data to extremely good at doing so.
- 23:24So if I take chat GPT like off the shelf, I don't do anything to it and I say, hey, I got access to Oliver social media posts and also his credit card spending and maybe Google searches.
- 23:36What do you think based on this data might be the big five personality traits and actually while you're on it, can you also give me personal values, political ideology and anything that you can think of.
- 23:48So and it does a remarkable job right it's never been trained, but because it's read the entire internet, it knows a lot about human psychology because it's kind of it's been in all of these human narratives that we tell about ourselves and that that we put out there.
- 24:02So it can just take any data input and translate it into pretty much any type of psychological inside that you that you can can think of.
- 24:11It already automates this first step and then not surprisingly, it's also really, really good at coming up with creative content that is customized to certain personality traits.
- 24:21So if I tell it, hey, write me an iPhone ad for someone who's open minded and extroverted, I don't need to even specify what open minded and extroverted means it just comes up with an ad that is really, really nicely customized to these to these different traits.
- 24:37What that means is you can now automate this entire process right you can go from digest someone's data, make some predictions about who the person is come up with creative content, put that out there and you can even have a feedback loop and say, well, which did that work did that not work, which traits should I potentially optimize for should I be looking at all of them.
- 24:57At the same time, maybe the certain traits that are more relevant in a certain context, if you're selling beauty products, well, maybe it's extroversion.
- 25:04If you're thinking about customizing health messages, for example, well, maybe it's something else, right, maybe it's the extent to which people are more promotion focused or prevention focus.
- 25:14So are they trying to optimize for better health outcomes or they just want to prevent bad things from from happening.
- 25:20So there's a lot that you can use a eye for to really automate this entire process. And that means anyone can do it.
- 25:27And this is already happening to be clear brands are already using this process.
- 25:32So we met from a model where not too long ago in our living memory, the only way to target someone is to choose a magazine or choose a TV former to say, hey, it's a fashion magazine.
- 25:43So I'm advertising my fashion bread core, but that's where it ended that we went to sort of customize but limited number of assets that I could target people with on based on their personality.
- 25:53Essentially, it's become hyper personalized that can look at each individual person in my target audience, make sure the algorithm customizes the copy, the visual, the appeal and with an inbuilt feedback loop so that the system becomes better and better and better.
- 26:10Yeah, it's essentially a segment of one.
- 26:12And for me, the interesting part is that it in a way goes back to the origin.
- 26:20So if I think of how does human interaction work in an offline content, forget about technology.
- 26:27Like whenever you interact with another person, as humans were so attuned and so good at trying to figure out, well, who's on the other side and how do I talk to them to maximize my chances that I'm being persuasive.
- 26:40My kids are so good at this kids know very quickly, how do I talk to mom to get the candy as opposed to how do I talk to dad.
- 26:48We all understand that you shouldn't be talking to a three year old the same way that you talk to your friend that you talk to your boss.
- 26:54So in an offline context, any type of person like any type of conversation that we have is to some extent personalized right kind of growing up in this village that you that you mentioned.
- 27:05Like almost all of the interactions that we that I had with people like bank cards or like in the supermarket.
- 27:11People knew me.
- 27:12They knew my family.
- 27:13So any conversation was personalized.
- 27:15They knew exactly what I wanted.
- 27:17And to some extent that personalization got lost as we move to Mars markets and mass media because there was no way of having this understanding of who was the person on the other side because they were somewhat anonymous.
- 27:28And they were just numbers in the system.
- 27:30And I think what we see now at this stage is that the merging of these two worlds is you have the the personalization of the small village.
- 27:39But you have it at a scale of mass media and that in a way again is is remarkable and also pretty dangerous because we just don't see this anymore.
- 27:49Like we don't I don't see what personalized content you are seeing.
- 27:53So there's no collective oversight that we had with propaganda into mass media.
- 27:58So it was still really effective and it was still playing into our emotional vulnerabilities.
- 28:03But at least there was some collective oversight.
- 28:05Now you might be getting something that I don't I never see because it just happens in your segment of one and you never see what I see.
- 28:14Like this merging of these two worlds that I personally find fascinating and somewhat terrifying.
- 28:20It is both fascinating and terrifying. So where do you draw the line?
- 28:26Where does psychological targeting cross the line between persuasion and manipulation?
- 28:33It's a great question and what makes this question really really difficult to answer is that it depends on the person you ask.
- 28:41But if I now sampled 100 people and I asked them how do you feel about the use of psychological targeting and the context of mental health.
- 28:50Their answers would vary quite substantially in terms of what data do I think you should be getting access to.
- 28:57Even with my consent should you be allowed to reach out to family members and so on.
- 29:02So there's like even in the population there's very different opinions about the topic which makes it really difficult to come up with this one size fits all answer to what is persuasion and what is manipulation.
- 29:13For myself I think that the main question that I'm asking myself and I'm also kind of whenever I work with companies I just asked the people on the other side is to what extent if we asked the user if we asked a consumer what they say the use of data the way that you're using it actually is benefiting them or is hurting them.
- 29:35And there's this one thought experiment that I that I really like which is Warren Buffett's newspaper test and it's super simple.
- 29:42It's essentially if the way that you're using data to influence behavior for example were published and publicized on the front page of tomorrow's newspaper.
- 29:53So it's kind of all of your practices the data that you collecting the way that you're using it is exposed on to the front page of tomorrow's newspaper.
- 30:00Would you feel embarrassed by it if if you know that the people you care about are reading this newspaper.
- 30:07Would you feel somewhat it key and embarrassed because it feels like yeah now they know what we're doing if that's the case well then it's probably crossed from persuasion and being helpful into the realm of manipulation.
- 30:21So I do think that there's these kind of ways in which we can try to think about it even though it might still not be the same for all of the customers that you're serving.
- 30:30But I like is like this tomorrow's newspaper test do it with more than just yourself you want to do it with as wide of a kind of representative sample within your company for example as possible.
- 30:42So you do get a sense is it at 90% of the people you ask say like no actually I would be proud having this on tomorrow's newspaper and maybe if it's like one or two people saying as you would still feel it key.
- 30:53Maybe that's okay because there's always people who will say that.
- 30:57But I think that's a nice way of saying we can we can put ourselves in the shoes of the other side and I'm doing this for more than just my own perspective.
- 31:05So I get a much wider sample of opinion.
- 31:08Yeah like that am I proud of it or do I have to assemble the crisis communications team to pull out the newspaper does love it.
- 31:17We can also as a business leader right okay there's advertising marketing there's also the internal communication sort of thing.
- 31:24So let's say I've I don't 10,000 employees and I wasn't charged of communication internally 10,000 externally as well.
- 31:31But let's stay here internally and of course they have different personalities.
- 31:36Yeah.
- 31:37Should I apply these techniques and the data driven approach to communication to personalize internal communication as well.
- 31:44Let's let's take this scenario.
- 31:46The company needs to change because we all have to adapt to new technologies and we have to or we want to introduce new AI tools.
- 31:53Yeah.
- 31:54But of course some people will be excited about that and other people may be worried about their jobs and where they I will replace them.
- 32:00Should we personalize this or should we stick with the old approach of saying this is our line to take this is this story that changed story we are telling internally.
- 32:10Yeah.
- 32:11It's such a great question because on under one hand I'm always amazed by how much money company spend on understanding their customers as opposed to understanding the employees.
- 32:22It's like we've known for decades that not all customers are the same.
- 32:28So we spend an enormous amount of time and resources on trying to understand their motivations preferences and then serving those those needs.
- 32:37And we really do this internally.
- 32:39Now I do think that there is a reason for why this might be the case right first of all I think there's something to be said about companies needing a sense of shared reality.
- 32:50The sense of like we understand the narrative we understand the vision of what we're trying to accomplish here and we we need to understand this in the same way because we're going to be working together and we're going to be making sense of a change in effort initiative together for example.
- 33:05Right if you kind of introduce a new tool we're all going to be working with it so having some consistent and coherent narrative I think is incredibly important for organizations now I do think that there's a way of making it very transparent and optional.
- 33:21So if you think about chatbots for example so what I would probably stay clear of it internally is just sending out messages without having any way of interacting with this and especially if those messages were personalized without any input from the user.
- 33:39Right I would never just read people's emails come up with a profile and then adjust my language because again if you if you think about this back to this example that we just went through the thought experiment if this was exposed and tomorrow's the front page of tomorrow's newspapers most of your employees would probably not be thrilled and you would probably not feel good about that having out there.
- 33:59And now is there a way to say hey we're going to create some kind of artificial intelligence chat bond that allows you to ask questions about the initiative like and then and then it kind of makes a personalized based on your input based on your questions and but it doesn't feel like we're just dropping it on you it feels like here is something that we offer because we understand that different people might have different concerns about the use and ideally to see oh would go and visit every single office every single
- 34:28critical and answer all of those questions but that's most of the time just not possible but you can still say we understand that you have all of these questions and we want to make sure that they are being answered in a way that speaks to you and it answers the specific questions that you have so we're going to make this tool available.
- 34:45And again now if you think if that was advertised on the front page of tomorrow's newspaper it would probably feel a lot better but it's not just personalizing to to kind of make sure that you want to adopt this tool it's like personalizing in a way that empowers you and makes you feel like no and now understand why we're doing this and you've just taken the time even though it's a chat bond to answer the questions that I had about this
- 35:10initiative that was internally going on.
- 35:15Interesting I want so I totally with you to say okay we need the shared reality internally and also it's not exactly good practice to you know use an algorithm to read
- 35:25everyone's emails and then tailor make messages to them to get them on board with the change initiative totally on board.
- 35:31What about using a chat board for internal communications to persuade people I get it you know we want to be able to ask questions absolutely I usually advise management if you
- 35:41have a change initiative you need to be out there and talk to people and answer questions is a chat board really be able to do that or is that other use case for it here.
- 35:50I think it would be able to do it so but I don't think it should replace the humans going around.
- 35:57I think oftentimes you even if you're trying the best you can to say we're going to have town halls and I'm going to talk to my team and we're going to try and get it
- 36:06from top management to middle management and then they're going to communicate it to the employees there's going to be times when your employee will have questions
- 36:14and most of the time they probably don't feel comfortable now just kind of knocking on their bosses door and say like hey I have another question about this tool that you've talked about two
- 36:24weeks ago and like one week ago so I think in these moments when it's just kind of it's an offer that sits there and says like in those moments when you do
- 36:33have questions and you want to have a sense of like okay what's the impact on maybe other divisions how should I be thinking about this
- 36:39impacting my work the work of the people that I communicate with internally and I think in those in those context it's extremely helpful
- 36:47because the barrier of asking your questions and hopefully getting answers is just much much lower but I totally agree with you that it shouldn't
- 36:55replace the humans going around and making a point for that right it's just like the same way that I think of these chatbots as a complement and therapy not
- 37:03replacing the therapist I would argue that it's the same in organizations makes a lot of sense because you have the speed and the scale
- 37:12and the reach that you don't have with human leadership can be 2 am in the morning at a plant in Malaysia and someone can ask the question
- 37:20what's going on at headquarters yeah makes a lot of sense now think thinking about team performance and building high performance teams
- 37:28I was going to be interesting to look into the psychological characteristics that influence business in this way
- 37:33so when you think about teams and the psychological make up and personality of the members of a team
- 37:41is there something we can learn from your work in terms of how we put them together and keep them motivated
- 37:46it's you you touch a very sore point because I absolutely love this topic and it's actually the topic that I first wanted to explore when I when I started my PhD
- 37:58I was mostly excited about this idea of how do we change the way that we hire and that we kind of reassemble talent within organizations
- 38:07because it always struck me as so inefficient to say we're just going to hire the best person
- 38:12the best person is going to be situated inside a specific unit and what makes for a best person in isolation might not be the best person
- 38:22that we should be hiring for the specific team so when I started my PhD my idea was like could I build this algorithm
- 38:28that takes input from all of the existing team members on all of the psychological dimensions that you can imagine some of the demographics, some of them
- 38:36what is their functional expertise but then also kind of personality traits values and so on
- 38:41and could I have the algorithm calculate the perfect profile of the person who should be hired into that team
- 38:47now it turns out it's incredibly complicated because it depends on so many different factors
- 38:54including is it a task that requires just is it a routine task where you just have to do the same thing
- 39:01time and again well then maybe cohesion and having somewhat similar personalities is actually helpful
- 39:06because you're just going to it's going to reduce friction is it something that requires innovation
- 39:11now if it's innovation maybe they should be functionally diverse right they should all be coming from different kind of
- 39:20different education different work experience different perspectives so you get the benefit of different ideas
- 39:27but maybe they should still be somewhat similar on a values level and a personality level so they actually get along
- 39:33and they can make the use use of these different perspectives but how big is the team is it just three people
- 39:39well in that case you might need one from from kind of every corner is it ten people that already changes the dynamics
- 39:46so I kind of I spend at least half a year trying to figure out what the input to my to my model should be
- 39:53and it just quickly blew up to so many dimensions that I would have had to collect an insane amount of data
- 39:59and I never actually got around to doing it but I think it's a fascinating question
- 40:04and the the closest that I'm currently thinking about which I would love to at some point kind of get real company data
- 40:11is this idea of well when we think about diversity in teams it's something that is I think we think of as a moral imperative
- 40:21and I think that should be an argument I don't think we should be just thinking about is it more effective or is it less effective
- 40:27but the the way that we talk about it oftentimes is well we should just increase diversity
- 40:33but diversity is not going to work in and by itself you have to find a way of saying well there's now different perspectives on the team
- 40:41how do we make sure that people listen and how do we make sure that we actually integrate these different perspectives
- 40:47and we create a sense of shared reality so that we same kind of working on the same problem
- 40:52and then kind of kind of successfully solve this problem so for me one of the questions is
- 40:57to what extent does like this functional diversity need to be complemented by a mechanism to create a sense of shared reality
- 41:06and that could be hiring for people who have somewhat similar profiles when it comes to personality
- 41:13because they are much more likely to actually understand and predict behavior
- 41:17so maybe if I work with an engineer who kind of has a very different perspective
- 41:22but we kind of have similar values and we get along on a personal level maybe a much more likely to listen
- 41:28or is it just a function of leadership?
- 41:31Do we need leaders to have this very clear vision and this very clear goal that says
- 41:36this goal is creating this shared reality that you need to again make the most of the diversity that you have on your team
- 41:42so a very long-winded answer to your question but I think it's a fascinating topic
- 41:47to be an expert for sure. And yet we're trying to use personality tests, my bigs and you know, red, green, blue, etc.
- 41:55Is this still as does value or is this just 1950s technology that we know is a bit superfluous?
- 42:02I mean there's personality tests that I really would stay clear of and if my bigs is actually one of them
- 42:08it's so much appeal because people love these labels, right?
- 42:13People love this very clear, oh I'm this for this for ladder code and people put it on their dating profiles
- 42:19they have layers of stickers on their office desk, the problem with kind of tests like my as Briggs is that they're not reliable
- 42:27so if I take the test today and I'm taking it again in three months, more than 50% of people actually get a different label
- 42:35and that's a real problem, right? Because people latch onto these labels so intensely
- 42:40that if they're not reliable and you can't trust them, that's a real problem.
- 42:45Now the way that I think personality questionnaires are used or have historically been used is also a bit challenging
- 42:52because it was mostly saying that there's certain personality profiles that we think are generally desirable
- 42:58and there's certain personality traits that are not desirable.
- 43:01If you're conscientious and organized and dependable, great.
- 43:05If you're neurotic and you're anxious and easily stressed, not that great.
- 43:09And so you kind of were hiring for the same people time and again, regardless of the position that they were being hired into.
- 43:16So you can imagine on average being somewhat dependable and reliable and focused on details is probably helpful
- 43:23but how about all of these creative jobs?
- 43:26Where you can't be in the nitty-gritty details, you have to kind of take a step back and think big picture.
- 43:32Those are not the highly conscientious people.
- 43:35Or sometimes you also have to have a somewhat of a adverse negative emotional response if you think about risk.
- 43:43If everybody is just emotionally stable and never has any negative emotions, well that's also not really functional
- 43:50because you're not going to be able to see here something that maybe we shouldn't be doing because it's just risky
- 43:56and it could lead us down a path that destroys us in a long run.
- 44:00So I think this traditional notion of we're just defining profiles that are better or worse, that is really problematic.
- 44:07If we think about it in terms of matching and we think about what are some of the jobs and tasks and teams that this person would thrive in,
- 44:15then I think it's incredibly powerful.
- 44:18Well, yeah, that makes no sense.
- 44:20Personalized persuasion has come a long way.
- 44:23Just looking at where do you think we will be in let's say three years from now.
- 44:28I mean, it's three years from now is not even that far.
- 44:31My sense is that it's mostly going to be multimodal in the next three years.
- 44:35So right now we still think of large language models as chat bots.
- 44:40It's like very much text-based.
- 44:42I think it's going to be much more seamless in terms of us being able to speak to these chat bots.
- 44:47In terms of companies being able to provide content.
- 44:51That's not just text, but it's pictures, it's moving pictures like video.
- 44:56So I think that's probably the next frontier.
- 44:59And we're already getting so much better.
- 45:03If you think about the first images that were generated by these models, they were cute but pretty much horrific.
- 45:11Well, it's a good time.
- 45:12Now they're pretty much there.
- 45:15Same with videos.
- 45:16Exactly.
- 45:17So we're not quite there.
- 45:18I would never 100% trust like an automated version of this process when it comes to pictures and videos.
- 45:23But still, like that's probably just a question of another year or two.
- 45:27Probably.
- 45:28Yeah.
- 45:29And 3D content is also coming with the engines.
- 45:31Awesome.
- 45:32Let's bring this back to Harba, Sandra.
- 45:34What's your best piece of communications advice?
- 45:37I think for me, it's just like curiosity.
- 45:40So I think that what makes conversations exciting and genuine is like this curiosity about the other side.
- 45:49I think all of the conversations that I've had that were inspiring that left a mark where those where I felt,
- 45:55no, the other side is genuinely curious about what I have to say.
- 45:59So I think curiosity would be my number one go to.
- 46:03Wonderful.
- 46:04I love it.
- 46:05And where can people find you and find more about the book, My Masters?
- 46:09There's a website that goes with the books called mindmasters.ai.
- 46:12There's actually a personality test on it that you can take.
- 46:15It doesn't collect any data.
- 46:17I set it up to make sure it's fully privacy preserving.
- 46:21But you get a personality profile so you can get a sense of where you land on extroversion, introversion, and so on.
- 46:27And it will be the same in three months.
- 46:29It will be the same.
- 46:30That's awesome.
- 46:31Sandra, thanks so much.
- 46:32Yeah.
- 46:33Thank you so much, Mary.