Human interactions occur in a variety of contexts. When interactions are marked by conflict, misunderstanding, bias, or aggression, 2024-25 CASBS fellow Katy DeCelles illuminates the micro-sociological and social-psychological dynamics that contribute to the sub-optimal interaction outcomes, enabling the formulation of corrective solutions and better organizational design. DeCelles discusses a sampling of her innovative work in conversation with Pulitzer Prize-winning journalist John Markoff (CASBS fellow, 2017-18).
Human interactions occur in a variety of contexts. When interactions are marked by conflict, misunderstanding, bias, or aggression, 2024-25 CASBS fellow Katy DeCelles illuminates the micro-sociological and social-psychological dynamics that contribute to the sub-optimal interaction outcomes, enabling the formulation of corrective solutions and better organizational design. DeCelles discusses a sampling of her innovative work in conversation with Pulitzer Prize-winning journalist John Markoff (CASBS fellow, 2017-18).
Katherine (Katy) DeCelles: Univ. of Toronto faculty page | Google Scholar page | Poets & Quants profile |
DeCelles work discussed & relevant resources:
"Scale Dichotomization Reduces Customer Racial Discrimination and Income Inequality," Nature 639, 19 February 2025
"Racial Bias Eliminated When Ratings Switch from Five Stars to Thumbs Up or Down," Nature, 19 February 2025
"How Gig Platforms Can Mitigate Racial Bias in Ratings," Harvard Business Review, 14 March 2025
"Different or Impartial? Actor-Observer Asymmetries in Expressing and Evaluating Sociopolitical Neutrality," Journal of Experimental Psychology: General 154(11), 2025
"Understanding the Dynamics of Workplace Violence Can Improve Employee Health and Safety," Rotman School of Management, Univ. of Toronto, 2022
John Markoff: website |
John's latest book is Whole Earth: The Many Lives of Stewart Brand (Penguin Random House, 2022). His next book (forthcoming, 2027), will be published by MIT Press.
Narrator:
From the Center For Advanced Study in The Behavioral Sciences, this is Human Centered.
Most of our everyday interactions occur interpersonally with friends, acquaintances, and strangers, as well as within workplace teams and organizations. Sometimes the interactions are mediated, if not influenced, by online platforms and other technologies at our disposal. While most interactions occur without incident and seem unremarkable, others can be marked by misunderstanding, bias, discrimination, tension, or even aggression.
A variety of micro-sociological and social-psychological dynamics underlie or fuel the processes that contribute to those interaction outcomes. Step one in elucidating the mechanisms and generating potential remedies requires astute researchers to notice the phenomena in the first place, either through visual observation or detection of patterns in available interaction data.
Today on Human Centered, a conversation with 2024-25 CASBS fellow Katy DeCelles. She holds the Secretary of State Professorship in Organizational Effectiveness at the University of Toronto, where she focuses on the psychology of conflict, emotion, aggression, and inequality – particularly in organizations and criminal justice contexts. All those contexts feature human interactions of some sort, and Katy is here to describe a small sampling of her innovative research and methods.
Joining Katy DeCelles is former New York Times tech journalist and Pulitzer Prize winner John Markoff, a 2017-18 CASBS fellow and an occasional guest host of our show.
As you’re about to hear, Katy unpacks her 2025 coauthored cover article in the prestigious magazine Nature. The article proposes a more effective evaluation process of online labor market platforms to reduce interaction bias and better identify quality of service. She also reveals findings of fascinating new work on people’s skepticism and morality evaluations of others’ expressions of neutrality over contentious social and political issues. Understanding why people choose to express neutrality in their interactions with others sheds light on why a neutrality strategy to manage impressions usually backfires. We’ll include links to articles related to both lines of research in the episode notes.
And, like many fellows from year to year, Katy launched new cross-disciplinary research projects with members of her CASBS cohort. How do her interests intersect with an English literature professor and a geographer?
Let’s listen to Katy’s conversation with John and find out…
John Markoff:
00:02:56 - 00:03:16
Welcome, Katie. So if you go back to the, to the sort of founding of CASBIS and the organization, it was about Fostering Interdisciplinary Research and Thinking. And I was trying to place you, and I kind of get the sense that you're very interdisciplinary in your interests. So how do you describe yourself?
Katherine DeCelles:
00:03:16 - 00:03:56
Well, I often call myself a mutt, that I'm a fit nowhere and everywhere. And I didn't really have a plan going into this. If I could look at my work now and say what discipline does this best fit in, it would probably be microsociology. Sociology, so social psychology, but more interested in topics that sociologists are interested in, such as inequality, racism, crime, prison, large societal problems, but really looking at individual-level processes and individual-level understanding and psychological processes.
John Markoff:
00:03:57 - 00:04:09
And I saw that, at least in some of your work, you've played with behavioral economic techniques. I noticed nudge bubbling up. Which is stuff that I love. And so you've used that as a, I don't know, as a methodology or as a—
Katherine DeCelles:
00:04:09 - 00:04:21
Yeah, I think I would probably not call it nudge. I would probably call it organizational design, that both physical design and structural design features that affect psychological processes.
John Markoff:
00:04:22 - 00:04:43
I'd really like to focus on your February paper that appeared in Nature. It was titled "Scale Dichotomization Reduces Customer Racial Discrimination and Income Inequality." And I might add, it was on the COVID of Nature. Nature magazine. Could you start by sort of laying out the background and what drew you to focus on this question?
Katherine DeCelles:
00:04:44 - 00:08:05
Sure. Well, initially, I didn't study issues related to race and discrimination until I had an undergraduate student who was interested in this. And I said, "Oh, sure, I'm happy to. This sounds interesting to me, too, but I have no expertise in it." And I started collaborating with her and with some other co-authors. Just on the topic. In my first sabbatical here in management science, you, of course, can't get away from Bay Area, and at the time it was gig work. Everyone's studying gig workers. And I had started thinking about evaluations of gig workers, specifically when it related to gender. And I had a conversation with a friend of mine about hiring a male nanny. And would you do that? And what sort of, what problems How do processes enable that, especially when you're selecting from a bunch of people? What are your attributions about people's motives? But I wasn't able to get data to study it. So I had reached out to a couple of companies asking to sort of, if they would be willing to share data, and I got nowhere. And then also during the pandemic, I got a phone call from one of those companies I had reached out to saying, you know what, this Black Lives Matter issues keep coming up, and we wanna make sure that we're doing everything we can for racial equity. And would you be willing to study that? And I said, well, yes, absolutely. So companies bring me ideas that seem really interesting and important. And so that company was open enough to share data with us, which is extremely rare, extremely rare, right? Being willing to see where the problems lie and then thinking about how they might fix those things. As we started just getting background on trying to understand in this organization, what are the various points at which discrimination could be occurring, sort of in selecting people, in algorithmic replications, in income, all these different sort of, we refer to as pinch points in the process of gig work, where might those be producing inequalities by race? And so we were understanding first, of course, evaluations, right? Ratings. The company says, "Oh, yes, but at one point we changed our rating system from this 5-star standard rating to a thumbs up, thumbs down because the distribution of the ratings was so skewed that we thought this would simplify the process for customers." Really, the underlying distribution was a smattering of 1s and then a ton of 5s and basically very little in between. This is called a J-curve. It's pretty consistent across online evaluations. And so we thought that that was really fascinating, especially when, from more psychological perspectives, you expect discrimination to occur when there's more ambiguity in evaluation. So simply removing the ambiguity of those 2 through 4 values and just having thumbs up or thumbs down should reduce expression of bias, which is what we found, which is really exciting. And—
John Markoff:
00:08:05 - 00:08:07
So you were looking at historical data in this case.
Katherine DeCelles:
00:08:07 - 00:08:34
Yes. Archival data. Yes. But I thought what was really interesting with that is that we're so well-powered. So we have 100,000 observations. So the difference in rating between your average white worker and your average non-white worker is extremely small. So 0.02 difference on a 5-point scale.
John Markoff:
00:08:34 - 00:08:35
But significant.
Katherine DeCelles:
00:08:35 - 00:09:16
But statistically significant. Because we have so many observations. But if you found that mean difference in a lab experiment or even a well-powered online experiment, you need hundreds of thousands of observations for that effect size to become statistically significant. But what was, I think, really most interesting for us is that it then related to wealth inequality. So the organization had looked at, you know, what's the average rating in our 5-star scale Well, let's put this compensation benchmark there. And so whoever receives maximum compensation is just above that line versus below.
John Markoff:
00:09:16 - 00:09:19
You were able to assign a dollar amount to the difference.
Katherine DeCelles:
00:09:19 - 00:09:19
Yes.
John Markoff:
00:09:19 - 00:09:23
How did you get to that? Where did you get the dollar data from?
Katherine DeCelles:
00:09:23 - 00:09:25
From the company. We're not allowed to—
John Markoff:
00:09:25 - 00:09:30
Oh, they know what people got because they were passing— the money was passed through in terms of the interaction.
Katherine DeCelles:
00:09:30 - 00:10:19
Okay. Yes, and again, this is why having the actual data is so important. So a lot of the— wealth inequality, even on gig work platforms, has been done through scraping publicly available data and estimating, you know, based on price differences and racial categories, maybe people's income might vary, but you don't actually know. You don't have that actual income to say that. And if you just looked at the rating difference, you might say, well, that's just normal error, right? But the fact that it then— the organizational design piece of putting the compensation benchmark at the average meant that your average your average white worker just made maximum compensation and your average non-white worker just missed it for inconsequential difference. So that's the practical significance that really intrigued us.
John Markoff:
00:10:19 - 00:10:23
Did that mean 9 cents on the dollar? Was that the number that you said?
Katherine DeCelles:
00:10:23 - 00:10:31
Yes, so yeah, I think it was 91 cents made versus a dollar made for your majority, Yeah.
John Markoff:
00:10:32 - 00:10:36
Is this a lesson that the company had already learned that you confirmed, or how—
Katherine DeCelles:
00:10:36 - 00:11:15
No, they had no idea. They had no idea, so I think they were pretty happy to see that that inequality had been remedied. But we are still working with them on an ongoing project that is probably a bigger problem. So the Nature paper is about evaluations for jobs that have actually occurred. But if you think back to the initial process of a customer searching for a worker to do a job, they're matched with a worker, and then they can potentially cancel on that worker right up until the job starts. So there we see a much bigger disparity by race of cancellations.
John Markoff:
00:11:16 - 00:11:18
You mean, so someone of color shows up, they cancel?
Katherine DeCelles:
00:11:18 - 00:11:18
Yes.
John Markoff:
00:11:18 - 00:11:19
You see that frequently?
Katherine DeCelles:
00:11:19 - 00:11:43
Yes, and it's a bigger problem. So when you think that downstream, those people with the ratings disparity, those are people who were not so offended, let's say, by the worker they were assigned to cancel at that moment. That might be a very different sort of subtle psychological process, whereas the cancellation decision could be driven by other things.
John Markoff:
00:11:44 - 00:11:47
It just occurred to me that you probably have geographic information.
Katherine DeCelles:
00:11:47 - 00:11:48
Yes, we do.
John Markoff:
00:11:48 - 00:11:51
That's fascinating. Have you explored that yet?
Katherine DeCelles:
00:11:51 - 00:13:11
Yeah, we generally control for it in the Nature paper. We don't see much variance in part because the company's customers do not vary very much, right? It's really, you know, an app for wealthier people in cities, which tends to be white people. But we don't have actual customer demographics. So we couldn't say— we could intuit from the demographics of the area. But I think with the cancellation project, it's more complicated because you don't want— as an organization, you don't want to send sending workers out to experience racism to their face, right? Hate crimes. So there's this also selection process happening, whereas if we still sent a worker to one of these places where the person preferred to cancel, they might experience different treatment. They might be mistreated or harmed in some way. So even if we were to apply some kind of intervention to reduce the disparity in cancellation, is that Necessarily the right decision, and we don't know yet. So I really, I would always hesitate to doing an intervention in the field before we really understand the implications of that, and how to manage that process.
John Markoff:
00:13:12 - 00:13:19
Separate from your interaction with that particular company, did the Nature paper generate any kind of commercial reaction? Did you get that kind of interest?
Katherine DeCelles:
00:13:19 - 00:15:21
Yes, yes, and our response really has always been, show us the data. You know, if you really wanna solve the problem, a quick fix of changing the evaluation scale might not be right for all companies. So, really, the company we were working with didn't know the disparity that they had or where it was happening or why it mattered. Other organizations might have disparities in different locations. It might not be a problem in a 5-star evaluation. Maybe it's in selection. Maybe it's somewhere else in the process. So, applying our solution might not fix their under— on their problem, but it also could potentially exacerbate things if we really, in that paper, are looking at high performers. So the organization we worked with does select people for their platform. So you can't just sign up and start working on it the next day. There's selection. So it's highly skilled laborers, often small business owners, highly skilled people who are generally doing well in the interview. We don't know much about poor performers, so it could be the case that in organizations where you have a wider distribution, not just a J-curve, but you have more distribution in the middle, that dichotomous evaluation might actually increase discrimination. If people were hedging previously and giving slightly better ratings because they had they felt guilty or they felt bad about something. And now this, I think of the dichotomization as this like a clarifying effect for people. Like, did this person perform well or not? And if they didn't perform well, then you could actually see a bigger problem, which may or may not be accurate, right? So we don't know much about that side of the problem. So I really think the first step in working with organizations is to let's just understand what's happening first before we apply a quick solution.
John Markoff:
00:15:21 - 00:15:43
Yeah. Let me ask you to step up to the sort of 30,000-foot level. And did you come away from the research with any new understanding into racism or racist beliefs? Are there any kind of generalizations you could make about our population or the population of people who use these services?
Katherine DeCelles:
00:15:46 - 00:17:44
Okay, so I will answer this. I was surprised in some ways and not surprised in other ways. I was surprised that it was very subtle. It was a very subtle difference. I really expected a much bigger ratings difference. So that can give you hope, right, that we've made some progress as a society. But then when you start thinking about the selection mechanisms and what's happening in cancellation, I hadn't thought through that implication. So, what did not surprise me, I guess, is that racial disparities do exist and that inequality is a problem for income. But this is hard— this is hard studying in online lab paradigms with large samples because those individuals are very different from a customer using an app that doesn't think a social scientist is going to see their data and make certain attributions about their behavior that's sort of inconscientious. Consequential, people taking online experiments know exactly what you're doing, right? And so they will often react in a way that makes themselves feel better. And so combining the two approaches, basically field data that shows these small effects, but practically meaningful, with lab studies where you're trying to deal with this social desirability and people who are aware of what you're trying to study is challenging. And my sort of The takeaway from that is that both methods are really important, but studying things like racism, sexism, you really need to observe things in real life. Because there are still problems. There are still problems, even if in our sort of standard psychological method right now, it might be hard to show because of how our methods have evolved.
John Markoff:
00:17:46 - 00:17:57
I mean, I guess you'd call this algorithmic intervention in a sense. Well, I mean, it's certainly not nudge. I mean, going from the scale to an up and down kind of choice, that's not a nudge. That's a—
Katherine DeCelles:
00:17:57 - 00:17:58
Some people would call that a nudge.
John Markoff:
00:17:58 - 00:17:59
They would call it a nudge. Okay.
Katherine DeCelles:
00:17:59 - 00:18:03
I don't know if I would. It's a structural change, an organizational design change.
John Markoff:
00:18:05 - 00:18:13
Does it say things about the ability to reduce bias? Just more general. I mean, using algorithmic approaches to reduce bias?
Katherine DeCelles:
00:18:16 - 00:18:56
I think using structural— structures and design is important and probably more important than trying to change people's attitudes and views. So, it's a very interdisciplinary team, of course, on that paper, but sociologists would say bias and they mean discrimination. Psychologists say bias and they mean attitudes. Attitudes. So we didn't change people's attitudes as far as we know. But we did change their behavior to be more consistent with what's fair and equitable. And I think that as researchers and as humans, that that's an easier problem to tackle than changing people's beliefs and attitudes.
John Markoff:
00:18:56 - 00:19:06
[Speaker:DAVID] Could you sort of walk me through how you got into the neutrality issue? What attracted you to that as a topic? This is something to focus on?
Katherine DeCelles:
00:19:06 - 00:20:24
[Speaker:KATHRYN PAIGE EASTMAN] Sure. Well, a lot of my work is phenomenon-driven. It's from an observation in life that makes you go, "Hmm." And I tell all my students that first you have to be an anthropologist. As any social scientist, you have to observe your social world. And rather than going straight to the literature, just look around you. What is happening? And so neutrality had come up as we were all in lockdown during the pandemic. And I was working with my colleague, Rachel Rutan, and we had just had a sort of a conversation about like, huh, you know, some people are neutral on vaccines. Like, how does that work? What do people think of when you say you're neutral on the vaccine issue? And we kept seeing it come up in various social media posts and just sort of sending it to each other and sort of thinking through, Why would people use this tactic? I mean, our intuition was this tactic would not be very successful, and yet people seem to be doing it quite often. That just kind of naturally led to, "Well, let's study this. Is there any work on this? What is neutrality?" from then. It was really very much just an observation of the social world.
John Markoff:
00:20:24 - 00:20:30
How did you set up the research framework? How did you frame the problem to actually study it?
Katherine DeCelles:
00:20:31 - 00:22:13
I would say this is one of my more social psychological projects in that it's very experimental lab-based, whereas most, if not all, of my other work has some field data. And that's in part because it was during the pandemic, and in part because it's a new construct. It's very messy in the field. So people can invoke multiple tactics, including neutrality, or they can do something that what some people might consider neutrality and others won't. In order to be precise, we really relied heavily on experimental situations, such as you come across this post on your Facebook feed, and this person wrote that they're neutral on the issue versus this person wrote that they're supportive of this issue. How would you react to that? What are your perceptions of that? Much of psychology has now moved to online platforms for collecting data, in part because it gives you more representative samples, but it also gives you larger samples. Psychology, after this transparency revolution, realized, you know, we're not well-powering our studies for good conclusions. Now, whereas before you could run an experiment in a couple weeks, now in order to run it well-powered, we're thinking like 200 people per experimental condition, sometimes more depending on the effect size. You really need many more people than you could even get in a laboratory, so using online panels like Mechanical Turk or Prolific Academic allows us to have better data in that sense.
John Markoff:
00:22:14 - 00:22:21
I have no idea what using Mechanical Turk costs. Is it cost-effective generally? I mean, you don't need a massive research grant to do this?
Katherine DeCelles:
00:22:22 - 00:22:54
Well, you do need massive research grants. It costs— tens of thousands of dollars to do that paper, for instance. Yeah, and so often with this type of work, you have multiple collaborators. You often have multiple grants. And now with the pressure for additional studies and convergence and replication, yeah, it costs a lot of money. Whereas before, you know, undergrads can get course credit and you can eke out an experiment in a couple weeks. Now it's, It's considerably more expensive.
John Markoff:
00:22:54 - 00:23:13
You saw this in the COVID— or during COVID Were there public events or controversies that also around neutrality that you stumbled across in the, you know, sort of larger than life things that came up as people taking a posture of neutrality that you could cite, or this was just real world kind of individual behavior?
Katherine DeCelles:
00:23:14 - 00:23:58
Well, real world, but often high profile. So a lot of politicians and organizational leaders that has really only increased Since then, it's hard to recall exactly what topics were happening besides COVID. It was just so top of mind back then. But now we've observed many, many instances of celebrities and social media trying to express neutrality on something that's highly complex, like Israel-Palestine, and then just getting really punished publicly for doing that. And what are the mechanisms of people's sort of vitriolic responses to neutrality that makes it different from even people who oppose them on that viewpoint.
John Markoff:
00:23:58 - 00:24:08
So you mentioned Greta Thunberg in particular as someone who got pilloried for attempting to stay away from taking sides on Israel-Palestine.
Katherine DeCelles:
00:24:08 - 00:25:05
Yes. And that was pretty early on. And I think observing these incidents in the news— so another incident was with the Disney CEO who had publicly said that they were neutral on some of the legislation, the anti- gay legislation in Florida. But internally, they had released a memo that said they were supportive of the LGBT community, but externally, we're going to be neutral. And if it's one thing that people really hate, it's hypocrisy. And so I think with Greta, that people have expectations about her approach, and a neutral approach was not was not expected, and so they, I think, saw that as a form of hypocrisy given her hard stance on many other liberal issues. So the hypocrisy is sort of a piece of it in the field that we don't tease out necessarily in the experiments, but I think is really part of the social phenomenon.
John Markoff:
00:25:06 - 00:25:11
So is this ongoing research? Where are you in the arc of— you did multiple studies here.
Katherine DeCelles:
00:25:11 - 00:25:11
With neutrality?
John Markoff:
00:25:12 - 00:25:12
Yeah.
Katherine DeCelles:
00:25:13 - 00:25:31
It is conditionally accepted at a psychological journal, but it's something we're gonna continue working on, especially when it comes to more naturalistic conversations. Now that we've got the basic phenomenon defined and figured out, there are many more questions for follow-up work that we'll be addressing.
John Markoff:
00:25:32 - 00:26:04
You only mentioned, I think, you only mentioned journalists and journalism once in your talk, but this really hit me hard because I think a lot about journalists as neutral observers. And, you know, I come at this personally as someone who started as an activist journalist and then later in my life, I decided there was a value in terms of what we expect from the fourth estate of having a neutral observer. Heinlein's notion of a fair witness in society. Have you guys— have you looked at journalists or have you thought about looking at journalists at all?
Katherine DeCelles:
00:26:05 - 00:26:37
We've thought about it. We've thought about looking at journalists and also about other social roles, and this is one of the biggest ongoing questions we have is how does social role affect people's reactions to neutrality? You can think about jurors or judges or anyone in the legal field, so that's where my mind goes initially. Those people are expected to be neutral, that we should see less punitive responses based on social role. Religious practitioners, so their social role is a big piece of this that we've yet to tease out.
John Markoff:
00:26:37 - 00:26:58
[Speaker:DR. HEFFNER] And I mean, I think journalists have been particularly in the crosshairs. You know, they've been hammered so hard with the allegations that they're hypocrites. But I know from being in the field that many people who were journalists, who are journalists, have this almost religious devotion to the idea. And so it would be an interesting research question.
Katherine DeCelles:
00:26:58 - 00:27:18
I think so. And the difference between objectivity and neutrality, I think, is also interesting. You can still hold views and be objective in a job, in a role. I think it's very hard to do, but it must be possible. And sort of figuring out how it's possible is also an interesting question, I think.
John Markoff:
00:27:18 - 00:27:31
Yeah. So you talked about the asymmetry of people attributing positive motives to their neutrality, but negative ones to others. Could you talk about that asymmetry a lot?
Katherine DeCelles:
00:27:31 - 00:29:58
Sure. So this is a theme in my work as well. Is what psychologists would call actor-observer differences, so an actor-observer effect. So when I, as an actor, am doing something, I make very different attributions about that, my own behavior, that are different from your behavior doing the exact same thing. So this initially started with a research project with Gabriel Adams and Leslie John, and we were looking at expressions of anger. So preceding this neutrality was a different project on— that was again inspired by news events where people who are accused of doing something unethical respond with angry denials. And we thought that was a really interesting phenomenon because when you are accused of something you didn't do, you're naturally angrier than if you are actually guilty of what you were accused of. The attribution and sort of social construction of an angry response to an accusation is that that's a guilt indicator, but there's no evidence that that's the case. And so highlighting that, especially for, you know, the legal system and my dad had some help with that project, right, with understanding that lawyers know this and they can take advantage of this this understanding. So thinking through that difference really then translated into the neutrality project as well. It's like, okay, well, my neutrality comes from this place of wanting to learn more and needing more information before I make a complex decision. Like climate change research is always evolving and you can't possibly know, you know, what's happening in that. But other people saying that they're neutral on climate change, the attribution is more negative, like self-interest or or even apathy, right? And we really see this effect most commonly on these polarizing social political issues like climate change, gun rights, abortion, where people's attribution for neutrality is, "How could you possibly be neutral on something so strong? You must just be an idiot, right? You must just be uncaring as a human instead of just having an opinion one way or the other." And flip it.
John Markoff:
00:29:59 - 00:30:04
I think the idea was strategic avoidance. Could you look at it from the point of view of the actor?
Katherine DeCelles:
00:30:05 - 00:30:49
Yes. So you're good to pick up on that. So strategic avoidance could be framed also in multiple ways. So a benign attribution for that could even be, "Gosh, they're really caring and they don't want to upset people. They're just trying to be agreeable and make the group smooth process. But, you know, strategic avoidance, they're just trying to get out of a conversation without looking bad. That's a negative attribution for the same behavior. So I think there's more in that category specifically that could also, I think, flip either way that we didn't get at in the paper yet in a satisfactory way.
John Markoff:
00:30:49 - 00:31:01
Real-world consequences that things that spill over into hiring, firing, social inclusion, exclusion. Did you look at that in any systematic way, or is that relevant?
Katherine DeCelles:
00:31:01 - 00:33:24
[Speaker:KATE] Again, we collected most of this data during the pandemic. We actually couldn't even really do the in-person lab experiments to get some kind of behavior. Right now, we're looking at large language texts of conversations to see reactions to neutrality that are beyond just that initial response as well, like how do things unfold. It could be that neutrality, while it's seen as less moral than taking a position, it could be that it also reduces conflict. It could be that it sort of quells the aftermath, and we don't know that yet. So we're looking at that. In the paper right now, we have one project where we measured behavior. We have one study that measured behavior, and it was in the supplement. And that we also look at the topic of marijuana, marijuana legalization, and at the time At the time, the state that we were doing this research in had marijuana legalization on the ballot. And it later passed. So we felt that most people probably were pretty supportive of this issue in this state. So what we did was we created 3 different posters. And this is on an academic, you know, university campus. So we said, "We see these posters all the time of panels, panelists, you know, book talks, what have you." So we made 3 versions of posters. A poster for a fake panel on marijuana legalization. And in one of those posters, we had a supportive quote from one of the panelists, and one we had an anti quote, and one we had a neutral quote. And we put them up, hundreds of them, around campus. And we had sent RAs out to display these and then to come back in 4 days. So the organization had a rule that all posters after 4 days will be taken down. 'Cause they're always taken down or else you end up with posters everywhere. So we came back at 4 days and our RAs then took photos of the posters. And our hypothesis is that more people would take down the neutral posters. We also looked at defacing or we even had some little tabs to take if you were interested in following up. We didn't see any differences on taking of tabs or defacing, but we did see clear differences with the number of posters that were taken down.
John Markoff:
00:33:25 - 00:33:32
In response to, how did it, give me the spectrum. People tended to take, oh, they took down whatever they disagreed with.
Katherine DeCelles:
00:33:33 - 00:33:43
That would be my theory, although we don't have data on the human side, the psychology piece there. We just have the behavior, which is something we're missing in the experimental data.
John Markoff:
00:33:43 - 00:33:59
I wonder if, I was thinking about the experiment design, maybe you could have a different quote and seeing how many people showed up, were incented to appear at the talk with a particular different, I would love to do that study, but I am suspicious that not a lot of people would show up.
Katherine DeCelles:
00:33:59 - 00:34:03
Oh, interesting. Unless you told students they had to be there. Yeah.
John Markoff:
00:34:05 - 00:34:07
What about institutional neutrality?
Katherine DeCelles:
00:34:07 - 00:35:53
That's also another topic. So organizational or institutional neutrality. We have some pilot data where we started to look at this, because it's always a question. And at least in pilot data where you're asking lab participants or online participants to evaluate a response, and you're comparing that response to a supportive or anti or neutral reaction to some kind of divisive topic. People seem to be more understanding of organizations than they are of individuals. But that said, this isn't a very hypothetical situation for, I think, a phenomenon that I see in real life that is very different. So I hesitate to make any conclusions yet based on that initial data, but people do recognize that institutions and organizations are a collective of people who might feel very differently. Whether or not they think that they should have an opinion is something that varies, and varies by age. Certainly, when you look at research on, on boycotting and social political attitudes, younger people think that organizations should have an opinion on things, should have a voice, whereas older generations less so. They think, you know, this is— these are separate things that don't interface. So I think we're going to continue to see this tension, especially when it comes to, you know, how relevant is it for your business, right? Is this topic really something that you should be speaking about or not? Or how relevant is it for your consumers? Customers are in red and blue states both, so it's a tricky issue that we don't have a good answer for yet.
John Markoff:
00:35:54 - 00:36:08
Let me go back to the actors and the range of things you found. When somebody says, "I'm neutral," what was the range of reasons that people, I mean, were you able to tease out why they chose neutrality as a strategy?
Katherine DeCelles:
00:36:09 - 00:37:33
Yes, yes. So at least why they say they chose it. We did surveys, I think with a couple thousand participants, where we asked them, "Tell us about a time that you were neutral on a topic," and we gave them a list of topics. You know, why, what happened, why did you do that, and what did you say? We had another condition where we asked people a time they observed someone else being neutral on one of these topics and why "Like, why do you think they did that?" et cetera. So a lot of text data in that. And then we started just categorizing people's explanations into different attributions or explanations for why. And I think we have 7 in the paper. I'd have to check exactly how many. But, of course, there are many more that come up idiosyncratically, but these are the ones that came— kept coming up repeatedly across issues, across different voices, speakers, and actors, and observers. So strategic avoidance is one of them. Needing more information, complexity is another one. Not wanting to hurt people's feelings, be agreeable. Also, yeah, and I can't remember the other attributions.
John Markoff:
00:37:33 - 00:37:42
Would personality— I don't know if you type people by personality, but I was curious, do one type of people people take a neutrality position more than other type?
Katherine DeCelles:
00:37:42 - 00:38:26
[Speaker:DR. KATHRYN KAPLAN] It's a good question. I think we have some follow-up studies with personality measures. So you would expect, if at least you're thinking about the 5 primary dimensions of personality, that one of those is agreeableness. So the tendency to want to keep social functioning smooth and And also, a sort of disagreeableness being the opposite, right? Professors tend to be rather disagreeable. But these things also vary cross-culturally, personality. And so there's a lot more research to be done there. But you would expect sort of agreeables are more likely to do conflict avoidance, right, to try to smooth things over.
John Markoff:
00:38:27 - 00:38:40
Let me conclude sort of on your next frontier. You know, you talked a little bit about research ideas that came out of your year. Did you go in any new directions based on conversations or the stuff that CASBS is supposed to spark?
Katherine DeCelles:
00:38:41 - 00:41:08
Yeah, and I think, really, this is kind of— I think of it as an incubation year, that there are many seeds that have been planted that will develop over time, and in part because relationships and projects take time. So I've set up at least two projects that will start with people otherwise I wouldn't have met. So one example is Dylan Connor, who's here, who's a geographer, a social geographer here. He and I are starting to collaborate to look at how wealth inequality relates to mass violence, so school shootings and those sorts of mass violence. I think there's— unfortunately, there's now enough data to start studying school shootings and mass violence events, historically, and then pairing that with other historical data. Sort of looking, thinking more about community structure, social structure, and where things go terribly wrong. Can we learn from that? Can, a dream study would also be looking at the physical interaction with physical space. We have no idea, you know, how schools should be designed. We're really, our implications for training for schools come from the military, and that's a very different context. And so I would like to learn more about keeping people safe in these rare, these rare events that can have massive catastrophes. Mark Algie Hewitt, who is here, who does, is an English professor but studies big data and word evolution. He and I have started a project on understanding how does hate speech evolve. And change meaning, so especially in the presence of algorithms. So I sort of took a bit from management science and engineering. You see sort of technology kind of creeping into my work from that. And he has been studying compound phrases and how those evolve and change meaning and which terms take off. And I said, well, that's a question that I've had about hate speech, especially in online content. When algorithms are released to detect hate speech, hate speech, people still often imply the same things using different words. So what words then become the new hate speech? And so that is something we just started. We have data to study it, but it just started.
John Markoff:
00:41:08 - 00:41:14
Yeah, that's neat. And so basically the control mechanisms that are put in place actually have an evolutionary function.
Katherine DeCelles:
00:41:14 - 00:41:34
Yes, yes. And is that good? Yeah. Because now people are just sort of saying nonsensical things and aren't using the loaded words that we all know we shouldn't say, or does it just perpetuate the problem and sort of chase it underground, whereas before it was more obvious who was the problem? So it's an interesting question, I think.
John Markoff:
00:41:34 - 00:41:43
Is this a research project which will be done entirely online, or does this stuff go on in the physical world as well, or are you interested in that side of it as well?
Katherine DeCelles:
00:41:43 - 00:44:39
Yes, so I'm also interested in— very interested in communication patterns. Especially in interactions. This project idea had spun out of an ongoing project of mine with collaborators at my home institution, University of Toronto, so Rachel Rutan primarily. We call that our Breaking Bad Project. We're looking at how do— the big question there is, how do people with deviant attitudes such as racist attitudes, how do they find each other and know that it's safe to express those things to each other in the absence of social groups? We know social groups exist, like certain groups on the internet. You go there and you can express deviant attitudes, but how does this happen interpersonally? How do people go from just meeting each other to robbing a bank, to buying drugs, to expressing racism? How does that communicate? Communicative process occur. We have done research on that outside of the online context and definitely in interaction studies. If you ask people, for instance, you move to a new city and you're interested in buying drugs, what would be some ways about— how would you do that? There are certain tactics that we've learned from that project about things that people do to try to navigate safety and deviant expression. One example would be they invoke a third party. Say I'm interested in— Marijuana is legal, so I feel like I can talk about it. Let's say we're chatting about it and I can bring up a third party and I could say, "Well, you know Bob Sutton, right?" Bob Sutton "Definitely smokes a lot of weed," and I can judge based on your response how you might judge me. If you're like, "Oh, that's awful," and I say, "I know, right? We need to help him," or if you say, "Oh yeah, I'm down with that," I say, "Oh, okay, cool, yeah, do you know—" It's a starting process of a very subtle communicative process that happens between people where there's a bid that has cover. Another way that people do this is they use symbols. So this is, you know, in secret societies in the history of the world, there have been, you know, physical symbols that people use to communicate membership in a deviant group that is not visible to most. So when LGBTQ was, you know, much more controversial, gay men would wear a bandana in their left pocket, right, which might sound inconsequential to your average person, is not going to trigger hateful attitudes, but is meaningful within the group that can signal that the deviance in that category is safe. Isn't that true about the Castro?
John Markoff:
00:44:39 - 00:44:46
I actually don't remember this. I live in the Castro, but I don't— or I have lived in the Castro. Which side your keys are on, isn't that supposed to be a signal?
Katherine DeCelles:
00:44:46 - 00:45:15
It could be. Right, there are lots of these, lots of these secret groups and symbols, but we're We're more interested in sort of the language behind how people communicate those things and how successful they are of navigating the tension of risk and reward, right? Making that connection and getting what you're interested in, finding similarity and homophily, but navigating the risk that that comes with.
John Markoff:
00:45:16 - 00:45:19
Yeah. Thank you very much, Katie, for your time. It was wonderful talking to you.
Katherine DeCelles:
00:45:20 - 00:45:20
Thank you for having me.
Narrator:
That was Katy DeCelles in conversation with John Markoff.
As always, you can follow us online or in your podcast app of choice. And if you’re interested in learning more about the center’s people, projects, and rich history, you can visit our website at casbs.Stanford.edu.
Until next time, from everyone at CASBS, and the Human Centered team, thanks for listening.