What second by second phone data reveals about human attention with Byron Reeves

Listen

Watch on YouTube · Apple Podcasts · Spotify

Summary

Byron Reeves, chaired professor at Stanford University (communications, symbolic systems, and the Graduate School of Education), on why human behavior is getting harder rather than easier to study, what a screenshot-every-five-seconds analysis of phone use (the Human Screenome Project) reveals about radically fragmented attention, why AI simulations and synthetic personas should be judged on prediction rather than understanding, and why the most durable human contribution to research is the connoisseur's skill of asking the right question.

Guest

Byron Reeves — Chaired professor (communications, symbolic systems, and the Graduate School of Education), Stanford University · Academia

From this episode — top claims

• The Human Screenome Project — capturing a screenshot of a person's device every five seconds for up to a year — revealed that real screen behavior is startlingly individual and unrelated to the commercial media being presented; people stitch together radically different content (tutoring, video calls, shopping, private activity) into a narrative that is totally unique to them. — Ep. 64, Byron Reeves

aytm's take on this conversation

Self-report metrics like "hours on Instagram" are dead; attention fragments into seven-second shards stitched into a narrative unique to each person. Stop defending human studies as ground truth, run prediction-judged AI pilots, and protect the skill machines can't apprentice: deciding which question matters.

aytm's perspective, voiced by Stephanie.

Full transcript

Auto-generated captions — speaker labels aren't always available and wording may be approximate.

0:00 There's just a lot of individual variation. There's variation not only between people that may be even overemphasized, but there's a lot of variance for any one of us over time. That next shopping trip might have a totally different story about why for that same person, why the strawberries got chosen. The whole thing, media, human behavior, thoughts, emotions, and behaviors, and the media that we're interested in looking at in relation to them is just really complex. I like to say infinitely complex almost.

0:29 So, there's just a lot to know, and our job as researchers is to try to simplify that, but it just gets harder every day. Hello, fellow insight seekers. I'm your host, Molly, and welcome to the Curiosity Current. We're so glad to have you here. And I'm your host, Stephanie. We are here to dive into the fast-moving waters of market research, where curiosity isn't just encouraged, it's essential. Each episode will explore what's shaping the world of consumer behavior, from fresh trends and new tech to the stories behind the data.

1:02 From bold innovations to the human quirks that move markets, we'll explore how curiosity fuels smarter research and sharper insights. So, whether you're deep into the data or just here for the fun of discovery, grab your life vest and join us as we ride the Curiosity Current. Today on the Curiosity Current, we are joined by Byron Reeves, chaired professor at Stanford University with appointments in communications, symbolic systems, and the Graduate School of Education.

1:32 Byron has spent decades studying how humans respond to media and technology. His work has shaped the field of media psychology, and he's authored landmark books such as The Media Equation and Total Engagement that changed how people think about their relationships with screens and machines. Byron's research spans both academia and industry, from Microsoft Research to consulting for other global brands, and more recently through ventures exploring AI simulations and synthetic personas as tools for understanding human behavior.

2:05 Today, we're diving into one of the biggest tensions in modern research. The challenge of studying real humans in an increasingly complex digital world, and whether AI simulations might help us understand people in entirely new ways. Byron, welcome to the show. Thank you very much. I'm happy to be here chatting with you. Yes. Well, we are happy to have you. Um just to jump right in with you, Byron, you spent much of your career studying how people relate to media, technology, and increasingly intelligent systems.

2:38 Looking back, was there a moment where you realized that understanding human behavior was actually becoming harder or more complex rather than easier in this more connected world? Yeah, I noticed you said decades, spent decades. You made me tired doing thinking about how many decades that actually is. But when I started doing this, there were three television networks and maybe a newspaper that got delivered to my porch, and everything was all analog, and it was pretty corralled and easy to get to. And so, I mean, that suggests and that couldn't be less true right now. I mean, there's just an infinite amount of content to to get to.

3:21 So, the recognition that that this was a more complex world of media is certainly gradual. But they when we went from analog to digital, let's say 25, 30 years ago, and everything started to mush together, and there was a lot more chance for me to be in control, and different companies combining forces, and traditional media sources just kind of melting, and so there was a 5, 10-year period where that was happening, but it's just been a gradual kind of drip uh torture or uh change in in the media landscape that that has made it just really complex. So, it's really important.

4:04 There's one other thing about that question is is So, the complexity of research is something that that I understand we want to talk about, but it's really important to know why or let me congratulate you for having that be a first question because when we talk about AI and the role of technology, the first thing that people think of is oh, this is going to be faster and cheaper. And it's going to be easier. I'm going to lose my job. You know, things are this is just really a a revolution here. And I think we forget that one of the main reasons for me anyway, not not so much in the commercial products that are being in the going into this category, but one of the reasons that that it disinterests me is research is hard to do. This is a really complex scene, and and maybe we can do it better you know, with with AI not not only faster and cheaper.

4:58 Well, speaking of how research can be complicated, I think that in a lot of these pushes to go faster utilizing AI utilizing more technology, we sometimes forget how messy human behavior really is and how we're very nuanced and weird creatures. People say one thing, do another, and they sometimes don't even understand their true motivations for something. Um and from a previous podcast guest, we had Lori Dormont on and she talked about how she went on a shop along with a customer and she purchased organic strawberries but non-organic broccoli. And she asked the shopper why did you do that? And she was a mom of of young children and she said, "Well, the strawberries are for my kids, and I I think that it's really important to be cautious about what goes into them, but the broccoli's for my husband and he's kind of already grown up and cooked up the way that he is, so it

5:55 doesn't really so much matter if he has organic or not and the kids won't touch that broccoli, so I'm purchasing that for him, which was so interesting because on paper you wouldn't have gotten that insight because it's just a very weird human behavior that she observed. So, from your perspective, given all that context, what makes social research with humans particularly challenging today? Well, just the idiosyncrasy in the details and the complexity that you just told in your story. I mean, that that is but that's more in relation to what our interest is in terms of marketing. You know, you might be able to go into any one piece of that story you just told and simplify that a bit, but all those stories are relevant to uh the journey that that that woman was having with with the vegetables and the and the food choices. So, comple that So, it's it's complex behavior. It's there there are many layers to it. When you map it on to

6:52 media, especially, so if you if you if you want to know about food choices, you are biting off a complex piece of humanness right right there already, so you have to have all that all those different stories kind of lay layered in there. Um and then you throw on this increasing complexity of the media itself and then that makes the stories even more complex and so it's just this I mean, idiosyncrasy is really an important word for me in in thinking about what how how to do research in marketing and whatnot. There's just a lot of individual variation. There's variation not only between people, that may be even overemphasized, but there's a lot of variance for any one of us over time. Uh you know, that next shopping trip might have a totally different story about why the the strawberry for that same person why the strawberries got treasured. So, it's it the whole scene media human behavior thoughts

7:50 emotions and behaviors and the media that are that we're interested in looking at in relation to them is just just really complex. You know, we I say I like to say infinitely complex almost. So, there's just a lot to know and our job as researchers is to try to simplify that but it just gets harder every day. And I think that's also kind of fun though about it and what makes what makes human beings so interesting to study.

8:17 Yeah, of course and it it's uh that uniqueness and especially the specialness of any any given um uh instance. There's there's another bunch of answers to the question of complexity though that are um really important um in my university work just all all over the university thinking about doing academic and theoretical work on um on how people think and feel in relation to media.

8:44 Um and these are challenges that are that are really significant. They get under we don't look at enough in the marketing context. I mean, in the marketing context it's uh real humans or AI simulation which is best, you know, and and and one one will win. Um uh or it this is going to be faster and cheaper uh that that's enough of a solution for me but for for me one of my interests in these AI simulations is because we are having a crisis in research in a lot of domains that AI can help with. We're we're having uh a lot of issues with respect to reproducibility of findings. And and this is human research is not necessarily the gold standard or the ground truth and you know, in in evaluating AI. It's really hard to replicate human studies. If you do the same study twice, you're not going to get the same answer with humans, uh, there's a generalized ability problem.

9:39 You have all these complex media stimuli that you have like you have the the advertisement for the organic strawberries that you mentioned. Well, uh, that advertisement could be any of 10 a million different versions. Do they all work the same? No, they don't. So, how do you get all those special instances in there? We've got online fraud, uh, that's really a important problem right now.

10:06 We have a hard time getting to the samples of, uh, of respondents that we we'd like. We have a really hard time getting permission to ask really sensitive questions, which are increasingly important. You know, if it's about sex, drugs, or rock and roll, or finances, or relationships, or which are can be very important in terms of marketing, uh, especially products that deal with, uh, uh, uh, issues that are are products that are private and whatnot. So, it's just all of these issues really make research, uh, a lot harder hard to do and and the AI simulations have a chance of helping out in each of those areas, not just going faster and and for less money and what without those three jobs.

10:50 Kind of in that same vein, uh, it's you can a theme in this conversation already and and generally on this podcast is that paradox of having more data than ever but still struggling sometimes to explain, predict consumer behavior. I'm curious if you could talk a little bit about what do you think traditional research methods miss about the audiences that they seek to understand? Like even when they're done well. So, not shoddy work but, you know, robust methods. Like how are how are they, uh, not meeting the moment, would you say?

11:24 Uh, they just miss the complexity. Uh, and it's it's not a like an intellectual critique or you forgot to take three classes in college, uh, on how to do this better. It's just really, really hard to do to to have research that represents that complexity. You know, how many hours did you spend on Instagram this week? Those kind of questions are almost useless right now given the hugely fragmented nature of the way we were looking at all this content and seeing all the marketing messages within there. Just it just just very, very complex. So, you have to bring in new methods that that allow for the idiosyncrasy and the extreme fragmentation and um and they they just get harder and harder. More data and more data and more complexity, more complexity. So, it's the the biggest problem I think that we have is representing trying to simplify in a way that allows people to take action on the research that they're

12:22 they're doing but still stays true to this almost inevitable complex answer that we don't often want to hear because I'm not sure what to tell the boss about you know how to spend our money if it's if it's real complex. I want to take a deep dive into the mention that you had specifically about Instagram and then talk a little bit about your work as it relates to the human screenome project which was a moment-by-moment analysis of digital behavior which is so, so fascinating. What did your work reveal then about how different behavior that how does it show up versus how do people actually describe their screen habits?

13:02 Uh so, it it was a Well, first let me tell you what it is the the screenome project. What we did, this is pretty much unguided by a lot of great theory but just this is maybe seven or eight years ago we started this but just this this notion that we kind of thought people were on these smartphones and they weren't doing anything for very long before they did something else. So, we developed technology and this is a long story short, that allowed us to capture a screenshot of what people of what people were doing on their computers or their smartphone every 5 seconds that the device was turned on.

13:39 And then follow them for up to a year. So, my we take Molly's script smartphone, and we put some software on it. We talked to her a lot about privacy and sensitivity and permissions and whatnot. I signed the forms. Don't worry about it. I signed the forms. Molly's well comfortable. Yes, she loves it. Um and then you don't have to do anything. We're just grabbing a screenshot every 5 seconds off your smartphone, and we're compressing it and and transmitting it to a server at Stanford University. And we collect a million of Molly's screens over the course of a year. And now we can look at your and then do we could do the same thing for Stephanie and and hundreds, maybe not thousands, of other people. So, that's what we were doing.

14:24 And then what we did, and this is really an aha moment for a lot of us that were in this green home project, is we the first thing we did our idea was we can get the computer to look at all this stuff for us, you know, it it can recognize we can use machine learning and AI. We can get But the first thing we did was actually make a movie of a day of Molly's smartphone use. And and then just play that movie. Just watch the frames. And you play that movie, and it is um startling in how first of all, how individual it is to you, how unrelated it is to the commercial pieces of media that are being presented to you. Like Molly, like all the rest of the people in the world, is not looking at the whole news story.

15:13 That are not looking at you know, whatever you do I mean it's it's very TikTok-ish even before Uh, we were uh, uh, enamored by that or or even before that would actually took over as a format for it to be shown. So, people were spending maybe 7 to 10 seconds doing any one thing before they did something else. They were joining together incredibly, radically different content. You you you were tutoring uh, your kid in math and then you were at at a video conference for 10 or 20 seconds uh, uh, with co-workers and then you were shopping online and just doing all these radically different things and then plus some things that were really private.

15:55 Um, you know, in there as well. So, you get this story of this this narrative of what you're doing that is totally unique. So, that that was a watershed moment for us. So, this notion that you can that you could represent that with a a uh, general comment about how much time you spent with your smartphone or Instagram or or anything else uh, is is just just melted. It's just not possible anymore. And so, one of the things that we've been doing in in the research field and you know, this is highly relevant that for thinking about market research is is as trying to recognize first describe what it looks like. Uh, and then, you know, how to how do all these how does all this content get joined and what is this uh, big stew of uh, you know, that's happening in media. And and it's this it's so that that's a huge answer to this complexity problem and it's it's a way to look at the

16:51 complexity, uh, kind of celebrate the complexity and you know, your movie is different from my movie uh, but but really uh, take seriously what how detailed that is and try to figure out what that means for how you do everything. You know, how you how students at Stanford write a paper. I can tell you how they write a paper cuz we've looked at their screens. They write it about a minute at a time and then they'll do something else uh, in in between, you know, um, unrelated to writing the paper. Uh so it's just this you know this real fragmentation and joining of got kind of unrelated sequences.

17:26 Maybe for good and bad, you know, I get to be I get to be in charge of managing my own arousal and interest and engagement. Uh but I'm also you know flying all over the place and maybe not getting as deep into some types of thinking as I should or so so there there's things to be worried about and maybe things that are interesting as well, but that's what's different. Part of what makes that so interesting too is that I don't think that any of us or many of us intend to be that fragmented, right? It is just the nature of the the tech and the media landscape.

18:04 Pretty much, right? The technology certainly encourages that, but I think the technology has found a way to really I mean we if we worry about maybe we'd say to take advantage of this kind of attentional interest, but I think humans are built to survey the world in about this the units that we're uh working with now. And if I had to make any prediction, I think they'll be even shorter. Yeah yeah, we we have this this notion I mean this is kind of an ecological how to survive in the world a couple hundred thousand years ago that's that we don't need now, but we can't escape it because we're just really built to to look at novelty, to look at new things, to be concerned about what's over there and what am I missing and um all I mean you could even take you know what in any of the the typical effects that we think about fear of missing out on in social media as this this constant need to be checking and to be surveying

19:04 and to is there something more important? Is there uh a more interesting mate, some food, a place to live? You know, I in in the in the kind of cave person sense of of surveying the world. But so it's it's not just they've caused us to be that way. I think they've also figured out how we are built. That's such a good point. But like then how do you how do you think about that then in the context of like I think as a parent of a 10-year-old, I'm watching all of this news and these, you know, uh congressional hearings essentially right now around like tech out of schools cuz we have seen, right, that there's cognitive decline on the test that we use anyway, right, for the first time and that we're seeing this first generation of decline and it is being tied at least in you know, from what I hear from experts directly to the tech. So it almost seems like it's it's tapping into something that is our nature, but it is doing so to a degree

20:04 that is like reshaping what it you know, how we learn to a degree that is like I don't know. Like does the learning Yeah, no, I think you're I think you're right. I'm totally support getting the tech out of the classroom. Uh uh it because there's a natural engagement and and what the tech is doing, which is a little bit different than saying, "Oh, they figured out how to change us in a terrible way that is, you know, that has a Yeah, so I'm not sure I'm not sure I always buy that argument. But it but I've So early on when smartphones first came out, um never said anything about them in my classroom and I only in the last year or two said, "No, we're You just got to get rid of them. You just got to close them down because it's uh and me too is just too too engaging, too enticing and you know, we want to do something different that we couldn't couldn't do if we were all online. We

21:01 want to have something special happening here. So Yeah, I'm all in favor of that. What's hard, too, is at least for me, my son's a bit younger. My son is 9 months old and he sees me and my husband on our computers all day, every day. Not because we necessarily are choosing this, but we both work from home, but we have, you know, challenging careers and it requires us to be in front of our computers 8, 10, 12, 14 hours a day. And so, he we we got him one of the like a very like old computer that we had that doesn't even work. And he sits, even now, he sits in his little chair and he has a little table and he smacks all over that computer. He knows already intuitively what it's for. And part of me is like, "No! No, don't do that." But it's like, what else is he supposed to know? Mom and dad stare at this all day, so I must stare at this all day.

21:56 Like how we used to do what our moms did, right? My mom's working in the kitchen, I need my play kitchen. It is that name. Well, but you're both highlighting the uh a nurture part of it. There's a nature part of that as well. Bright flashes of light, fluorescent light from a screen get noticed in our world and because we're human and you better notice things that are you know, that are prominent in your environment cuz it they may eat you. Uh and and so so we're we're kind of built kind of built in as well. So, it's not only that you know, that it's happening, but um don't put that computer in his room when he gets one of his own.

22:36 Maybe not allow that. In the living room, you know, or No, this computer this it doesn't work. It doesn't work. It's not on. It's just for him to smack buttons, but then he gets really bored of it because to your point, he comes and sits over by me and he sees the flashy faces and the word documents going by and the slack channels. He's like, "How come my computer doesn't have that?" And I was like, "Oh, honey, no, So, want to be like me. Not yet. You have plenty of time to stare at a computer, it's not right now.

23:05 I I wanted to pivot a little bit uh to talk about more what you touched on at the beginning of our conversation about AI. So, let's talk about AI simulations. And to our listeners, no, you don't necessarily have to know exactly what this means or all about it yet because I sure don't know the context of this. So, Byron, for someone hearing that phrase for the first time, what does it actually mean to you to simulate a human response in market research?

23:33 Yeah, so can we get a computer to connect to a computer that has cataloged the entire history of the world via the internet and everything that's been said in social media and everything that's been published in the Med library and everywhere else on on campus. So, that huge volume of, you know, generations of knowledge, can we find a way to get a computer to jump into that space of information and via these large language models and AI that everyone is talking about uh and kind of journey around that space via the prompts that we give it and create a way to predict what people might do based on what they've done for, you know, 100 years as represented by information on the on the internet.

24:26 Well, when you put it like that Yeah, so So, we're trying to build a computer program or a computer instance or computerized instance of of a human that that simulates what a human might do based on all the information that we've gotten from this gigantic cataloging of of human experience via the internet. And then we can ask questions about that model. We can ask that maybe don't have that are not just looking up things that people have done in the in the past, but are actually developing some form of predictive extensions. You know, some people call it intelligence of some sort, but I'm we're not simulating human brains. That's not what what I'm I'm most interested in, although there's a place on our campus where people that's what people are using AI for, but can we make predictions about what someone might do next, you know, given prompts and and materials that we can actually

25:24 show. So, modeling humans to predict what they might do without having to actually find a human to talk to and find a a construct a conversation that they're comfortable with that that my university's IRB that the the people who are interested in privacy and sensitive information and whatnot and human subjects and whatnot that can live with, which is increasingly, you know, more stringent. Um so, trying to get a model of that predicts what people do that's that's kind of run by a computer.

25:57 I think it's I think it's important to understand the grand nature of these things, cuz maybe we say it in passing all the time, but to actually sit in what that means, I think it's important. Well, I was just going to say, you know, understandably in the field, there is curiosity, there's excitement about um AI simulations, about synthetic personas, but there is also, I think we all know, a healthy amount of skepticism. Um I would love to get down to just some practical tasks here. What kinds of questions do you think that AI simulations are well suited to answer?

26:32 Good question, and then the the counterparts of that I will be which one? Well, one thing to say is that they're going to be better at answering questions that the internet knew a lot about or knows a lot about. Um and that's not totally everything, but uh simpler questions, they'll be better at uh questions that get a lot of treatment in the internet uh and I think this is uh there's good news for market research in in that because there's a lot of uh information you know social media uh as well as you know gobs of millions of market research studies that can all be consolidated to allow a a model to be constructed where you can predict this.

27:14 So um that's I think the most important thing to say. I think they'll be good. So simpler questions is an important thing to say as well. So there are complex answers to questions about why people behave they way the way they do in the marketplace that that will be so I'm going to combine the pros and cons or so the complex answers that if you're of this personality type and have this uh uh kind of uh environmental ex- media experience and um have this kind of educational uh category but have been in these kinds of convers- just very complex answers to why people do what they do in the marketplace. The more complex that gets the harder it's going to be to create a model that's accurate. Um you know imagine a a model uh what would be a good example? A a model of an airplane. So we want to talk about how how do airplanes fly? Well here here's a model of an airplane and we you know we

28:11 can see it has wings and the more things that that it has the harder it's going to be that for that model to actually be useful in explaining flight um so that I think the most thing most important thing is the simple thing. Um yeah that's so let me let me start with that as the um the I think an important first answer to that that question. Here here's something it can't do that is responsible for a lot of you mentioned at the beginning of your question that you know some people are not terribly happy with this concept of uh personas or uh maybe are very critical of them.

28:49 There's a couple reasons why those uh are that have been put forward for those criticisms, but one is that the AI models are not simulating how humans work. They are not simulating the human brain and how the human brain experiences an environmental information, emotional experience, relationships, uh desires, and whatnot.

29:17 There I'd mentioned there are people on my campus who are very interested in having a computer do that because if we use the computer to do that, maybe we can actually model the complexity of the human brain in ways that that are have been really impossible heretofore, but I think for marketing in the marketing context, we're much more interested in predicting. In In In um I'm not sure how the human brain works, but I'm pretty sure if it's if this cookie package is offered at this price versus this price, people would be more likely to buy it, or there won't be any differences. So, I'm pretty confident in that prediction, but I'm not really talking to you about the different regions of the human brain that were involved in processing rational versus emotional information or some complex kind of neuroscience explanation there, but the prediction is is really pretty good, I would say.

30:13 So, prediction versus true understanding, but it also sounds like you're And this was one of the questions we had, and you've already answered it beautifully, but I was thinking, you know, in market research, prediction is good enough, but then there are these foundational fields like psychology, where it's really about understanding, but I suspect that these labs are neuroscience or psychology-related labs, right? Well, it's interesting. This prediction versus uh explanation uh has really been you know has a long history at in science uh and especially in technology. I mean one example I was director of an institute here at Stanford that was very was very active in the first technology related to speech recognition. So there was a group several decades ago that there was a group of linguists that thought the first thing we need to do is explain how human language works, how it develops, how people think about

31:10 language and theory theory theory and when you're done with that then come to me and we'll build some technology based on that theory. And then along come these computational linguists and said you know, we might not need to know how how language works. What we can do is just we can try all possibilities of how how this might be interpreted. Uh you know, we've got computing power to spare, we've got these new machines, we'll just uh you know, try a bunch of stuff and maybe we can match uh you know, what was said with with what it actually means and they won. Yeah, so they won and I think that's happening in a lot of a lot of different areas. So it's unsatisfying in the I mean the theory is really quite satisfying. It's really nice to be able to say why something happened. But it's also you can also develop systems of prediction that add up that you can build up into a theory or a at least a concept of how how something's working. So yeah,

32:10 prediction is is I I would argue is is is is enough to really keep us interested in in what AI simulations can do in in marketing and other areas as well. Uh you know, finding tumors in in x-rays, uh figuring out how organizations work. Uh uh the economists are really interested in in AI simulations of of people with a lot of them without too much interest in well, why did you choose that versus that? It's that you did and knowing that, I can then build a theory.

32:46 So, this is it's the same this is an important point by the way. I think that the these AI questions are the same across domains that you know, where AI is being applied to thinking about humans. I want to touch a bit on the adoption side. So, you mentioned economists, you mentioned that there's other people that are looking at this and they're excited about it and they want to implement it. And you've worked for major consumer brands for many years. When business leaders hear AI respondents, synthetic respondents, they can often times get excited at the shiny new thing that's going to help them the the promise of better answers faster, cheaper.

33:26 But what questions before they dive in headfirst should they be asking either of themselves or of these providers? Good that they're leaning in to asking questions. So, that's the first thing. I think what I think one first thing I'd say is that it's not AI or not AI or AI or humans is not a useful question. I would I would look into AI. I mean, I think there's that's quite well demonstrated that there's something to look in here. So, what what that says though that there's a huge amount of variance within this AI category of what they could look into and the ways in which you would look into it. So, this is not about using complicated software to do market research or you know, with humans or talking to chat GPT about you know, which cookie package is is better. This is not it's not I mean, I would that might be a fun thing to do to ask chat GPT which of these packages but it

34:24 is going to is going to work best but it's much more than that. There's a long list of best practices. So, when you look into this, know that there are people that have thought about how it needs to be made available for people to use, how if you're going to build AI simulations or AI personas, how many do you need in your research, what kinds of questions can they answer the best, how do you talk to them, how do you actually make individual models for that can be collected across different respondents. Um so, there's just a whole lot to know. So, so that's one thing. You're you'll need some help.

35:04 Um and it's I think it's important it's there's just I can't imagine what it would be like to I can't actually imagine what it would be like to be a business person having to wake up and get, you know, a thousand emails in your inbox about people selling you AI software that is actually turn turn your market research around and make you a fortune. Um but you've got to kind of wade through that. But there are people that have been doing it you know, I've I'm working with a group called BpointsAI that is that is trying to do this to build a comprehensive um package that can that can collect all this information and make it easy for you to actually do these simulations and to to catalog it and to to use it in the company. So, that's one thing is how you know, how look into the detailed look into ways in which you can get some help on doing this. What are the best practices and what how are the how can you get

36:00 access to the best practices? I I think the other thing I'd do and I've talked to a lot of marketers who are buying this or or thinking about how how they should respond to this kind of software, but and this may be true for market research even a little bit more than from for some of the other domains you know, like economics or organizational behavior medicine or something. But the people who are going to use this software in market research organizations should be and are going to be and are already worried about their jobs. What of this technology is going can be done that replaces me? Uh and not all of it by any means and not everyone's job is going to go away and there are these famous sayings about you know AI isn't going to take your job but somebody that knows how to use it might take your job. Uh you know I I like that kind of thing but but to be sensitive to that context. So

36:58 you need to as a business leader you need to figure out how are you going to get that that technology into your group and company in a way that makes everybody feel comfortable about exploring with it cuz that's what you want people to do. We don't know all everything we need to know about how to use it. Um it it there needs to be a lot of experimentation and somebody's not going to be a good experimenter if they're worried about worried about their job.

37:25 And and it also kind of chases some people in marketing to and here I'm being critical of one of my colleagues who have really interesting things to critiques to say about um these AI simulations but it it move pushes people off to I would say kind of minutia critiques of you know well there's sickofancy in you know AI and there's gender bias and you know that variance distributions for these kind of variables are different in AI than they are. Those are important things to look at but I think they're that big relative to you know what's going to happen um you know in in changing the way that we do research. And they're also things that once noticed there are chances to repair and I mean AI's never going to be worse than it is today. It's just going to be better tomorrow and next week. So there are a couple of comments about what I did I I like that question. I think it's really

38:21 important. You know, how would as a business leader, how are you going to get this in your How are you going to get experimentation done? But but to your point about how, you know, these tools are improving, um what do you think becomes the most critical human function? And we can ground this in the research process, just as we're saying, you know, cuz I'm sure it's variable, but yeah, what what remains human in this? Great question.

38:49 So, this is philosopher of science, pardon the professorial answer here, but this philosopher of science that was working on um what if science is this objective process, what's left, you know, that's really human about uh science? And he The answer in this one literature is the question. It is deciding what's important to try to know about. Um is the most important thing. And he he has this this a guy named Michael Polanyi, and he has this lot of really interesting literature.

39:24 To become a concierge of scientific beauty, or to become have good intuitions about what in marketing are the most important questions to try to answer. That's going to be harder for AI to answer that, although they may get there at some point, but it's going to be harder to answer than it than uh you know, create a sample of people that matched all the people in Iowa, and ask them, you know, who they want to vote for, or whatever. Or or what they what product they prefer. Whoever's doing that, and making the calls, and doing the interviewing, and and aggregating all the data, and running all the stats, that's that's going to going to go by the wayside. But but the uh Did I say concierge? I mean, connoisseur. Did I say concierge? I'm kind of I I meant connoisseur.

40:10 Yes, I I don't know what I was say I was thinking of being in a hotel this week, or being a connoisseur in the sense of um tasting fine wine and knowing which one is better. It is just something that's done as an apprentice as as somebody that's participated and and you become a connoisseur of excellence and Like a curator of of questioning almost, yeah. It's personal knowledge. It's not objective knowledge. It's totally mushed together with who we are as human beings. So that's that I think is one thing that that that that connoisseurship is really quite important to stay on top of that cuz you will be known and you can develop a great profession and be known as somebody who's asking the best questions. Forget how you're going to answer them for a minute, but just you know how are you going to ask the questions.

41:02 And I think the other thing I would say is that even with AI, you're going to have to answer the questions working with other people. And maybe you can ask ChatGPT how best to do that. But it'd be nice if you were also had some, you know, emotional intelligence to be able to work in a group with this kind of information. So you know, I think in in in our classes, so I if if we have a I teach a big course in media psychology and we have projects that we do. We do them all always in groups now.

41:34 You know, there's a tendency to think AI is well, I can go to my couch and do my study and you know, I don't have to talk to anybody and that's not true. You've got to talk about what's the question that we need answered. What's the best thing to ask the personas that we create? Um What do we make of the of the results? You know, is it important? Is it actionable? How much money are we going to spend, you know, on on that action? But doing that in a group cuz you've got to do that with different functions in your group. So connoisseur work in groups, play nice with others.

42:08 So Byron, for someone who's listening, Um, maybe feels both excited, but uncertain about where it where AI is taking the research industry. What would you say is one mindset or I feel like Byron just answered this though, right? By saying be a connoisseur and be able to work in groups. So. Yeah, the the larger answer maybe is lean in. Uh, don't run away. Uh, there are people that are kind of running away from and there are a lot of uh, places to go run away run away to that are not totally wrong. I like uh, if you think that AI is going to destroy uh, the universe in 10 years or 2 years, then then maybe we should all be running. But but to lean into experimentation and to change the mindsets and to really to experiment. To experiment. In a practical sense, pilots for software have never been more important. Figure out a way to and the

43:07 companies that are building the software are more than happy to find a way for you to inexpensively some of the AI slop that's being generated. Whatever looks the best, you know, whatever you can you can uh, evaluate may have a good prospect here and try it out. How does how how valid is is it on its face? Uh, did it make you think of things that you would didn't think of before? Uh, how consistent is it with information you've been collecting for the last uh, you know, couple decades?

43:38 Um, and just to lean into the experimentation. And and it's kind of disconcerting and it's a you know, change something that's a change, so it's it can be hard in that sense, but the lean in part I think is important. Yeah, that makes a lot of sense. Thank you so much for joining us today. I think one of the things that really struck me from this conversation is the reminder that in the age of AI and tech that humans have always been astronomically complicated and that adding more data about our behavior or what we do doesn't necessarily mean that that a load is going to make us easier to understand.

44:16 For sure and I think it another thing quite honestly that I am going to be thinking about for the rest of today at least is that this notion that prediction and understanding they're really not antithetical and that building enough prediction can get you in some cases to that theoretical space as well and I really have not thought about that and I think that's just terrifically interesting. So thanks for sharing.

44:42 Thanks for your questions. Yeah, so thank you so much for all of this wonderful perspective that you shared with us today and to everyone listening. Thank you so much for being part of the Curiosity Current. We'll see you next time. The Curiosity Current is brought to you by AYTM. To find out how AYTM helps brands connect with consumers and bring insights to life, visit aytm.com. And to make sure you never miss an episode, subscribe to the Curiosity Current on Apple, Spotify, YouTube, or wherever you get your podcasts. Thanks for joining us and we'll see you next time.

Produced by aytm

Curiosity Current is made by aytm, the consumer-insights company. Guests speak for themselves; the synthesis is ours. About aytm