Science Friday - Contraceptive Access, Robot Bias, Story Structure. August 14, 2020, Part 2

Episode Date: August 14, 2020

Roboticists, like other artificial intelligence researchers, are concerned about how bias affects our relationship with machines that are supposed to help us. But what happens when the bias is not in the machine itself, but in the people trying to use it? Ayanna Howard, a roboticist at Georgia Tech, went looking to see if the “gender” of a robot, whether it was a female-coded robotic assistant like Amazon’s Alexa, or a genderless surgeon robot like those currently deployed in hospitals, influenced how people responded. But what she found was something more troubling sexism—we tend not to think of robots as competent at all, regardless of what human characteristics we assign them. Howard joins producer Christie Taylor to talk about the surprises in her research about machines and biases, as well as how to build robots we can trust. Plus, how COVID-19 is changing our relationships with helpful robots. Plus, contraceptives have been around since the 19th century, but for decades, more than half of the pregnancies in the United States were unintended. In recent years, that number has improved, but it’s still an astonishingly high 45%. Why is that? Family planning is a balancing act. Access to contraception, education on how to use it, and new developments that fit the needs of the public are needed. Even though there have been advances in all these fronts we somehow are still not completely hitting the mark. This is reflected in the high percentages of unintended pregnancies. How can we do better? Linda Gordon, a historian and professor at New York University and author of the book The Moral Property of Women: A History of Birth Control Politics in America and Cynthia Harper a professor in the department of obstetrics, gynecology, and reproductive sciences at the University of California, San Francisco join producer Alexa Lim to discuss this.  And, if you hear the words “once upon a time,” you might guess that you’re hearing the beginning of a child’s fairy tale. And if you hear the words “and they all lived happily ever after,” you know you’ve probably come to the end of the story. But what happens in between? Writing in the journal Science Advances, researchers report that by using computerized text analysis methods, they’ve been able to identify words that help indicate the structure of a narrative. The team analyzed thousands of stories—from fiction found on Project Gutenberg to the transcripts of TED Talks—and found some common rules that seem to apply to most narratives. During a story’s introduction and scene-setting parts, for instance, articles such as “a,” “an,” and “the” feature heavily. Conversely, during moments of crisis and conflict, words like “think,” believe,” and “cause” appear. The researchers wanted to find out if these patterns might function as a sort of signal, helping an audience follow plot lines. However, these patterns don’t necessarily make a story any better—the study did not find that stories using these rules were necessarily more popular. Ryan Boyd, a psychologist at Lancaster University in the UK, joins Ira to talk about the structure of stories and the rules we use when navigating a narrative.        Subscribe to this podcast. Plus, to stay updated on all things science, sign up for Science Friday's newsletters.

This is Science Friday. I'm Ira Flato. Later in the hour, we'll be talking about the role of education and accessibility in unintended pregnancies, and why we still don't trust robots to do the jobs they excel at. But first, if you hear words once upon a time, you might guess that you're hearing the beginning of a child's fairy tale, right? And if you hear the words and they all lived happily ever after, you know, you've come to the end of the story. That's a common structure we are used to hearing. But is there, a common story structure in the middle, too. Researchers are using computerized text analysis to key in on words that indicate the structure of a narrative. Are there words that indicate an author is doing some scene setting, or special words when some moment of conflict or crisis is at hand? And how do these structural words affect the story as a whole? Ryan Boyd is one of those

researchers, assistant professor in the Department of Psychology at Lancaster University in the UK, and one of the authors of a report on the research published this week in the journal Science Advances. Welcome to Science Friday, Dr. Boyd. Hi, thank you for having me. You're welcome. Now, I understand that you analyze thousands of stories looking for keywords. Give us an idea what were some of the words that you clued in on. That's right. So we looked at what we call function words. primarily. These are the small throwaway words that most of us don't pay any attention to, words like the and in and of. And what we know is that these words actually, even though they seem unimportant,

tell us a lot about what a person is doing psychologically. And so these are the types of words that we were scanning texts for. We were looking for articles and prepositions to measure how much staging or setting the scene authors were doing. And we also looked at pronouns to see how authors are referring to their characters and what their characters were doing and so on. So you could draw sort of a graph of how many of each type of words there are at a given point and kind of map out the structure of the narrative. That's exactly right.

We can take these words and we can look at how frequent they are or how dense they are in different parts of the story. and get a rough sense of how much an author is doing different types of storytelling behaviors. Give us an idea of what stories you looked at for this. We looked at a really wide array of stories. So we started off by looking at very classic novels, a lot of things like Mary Shelley, Jane Austen, just a big collection of what we think of as traditional literature these days.

but we also looked at things like movie scripts. We looked at short stories by professional authors and amateur authors. So it was really all over the place. We looked at a lot of different types of stories. So do you find that there is a common structure, that they have the same logic and layout and linguistics going on there? We do. We find that regardless of what type of story it is,

whether it's a short story or a long story, they all go through the same process in the same order, and the shape looks pretty much the same. And this is true, whether it's movies or books, if it's stories from the United States, from England, from China, from India, they all share, in general, a very similar structure. Can you say why that is? Is it something that's taught perhaps in school or something that naturally happens when writers write? That's a great question. We don't know exactly why it's the same. there. We think that's a little bit of both. We think that this is probably indicative of how people

naturally process information. It's really hard to understand the story and what's going on and why characters are doing what they're doing if we don't first have the background information. And I assume that over the years, people have learned that this is a really good way to tell a story, which is why we often teach this method for storytelling, exposition, then plot progression, So then we as readers, because we read so many of these stories, I guess we get clued in after a while to look for these keywords to say, oh, look, a plot change is coming, be ready for something coming up. I think that what a reader generally does is follows it in a more zoomed out sense.

So we don't think that readers are looking at the really low-level details, like pronouns and articles and thinking, oh, no, something bad is about to happen. But we do think that readers are kind of non-consciously picking up on these cues to get a sense of what it is the author is trying to do. Kurt Vonnegut once said that you could do a similar kind of analysis for the emotional arc of a story that maybe even a machine could do it. Is there any emotional component to your work? In our work, we looked a little bit at emotions, and we do see that there are some structures that emerge, but it's a lot more variable. What about the great stories that don't necessarily follow a linear structure, like my favorite movie, the flashbacks in Casablanca, for instance, or if you want to take it to an extreme memento, which runs backwards? Does your analysis hold up for these?

It does. And these are great examples in that when we look at how these words show us the structure of these stories, it maps on to how we intuitively understand them. So in both of these cases, Casablanca, Memento, they don't follow this very normal setting of the stage, plot progression, conflict arts. Casablanca has a very famous flashback scene in the middle where they're back in Paris and the Germans are invading. And we see the staging process really jump up at this part of the movie because the storytellers are giving us a lot of new. background information that we didn't have before. The same is true also for Memento, where when the film starts, we have no idea what's going on because we haven't been given that important background information yet. And we can see this in the structure of the film as revealed through these text analyses. Does following these rules then make something a better story?

Can we use this to improve our storytelling? I think that we can use this to improve our storytelling to a degree. If a book or if a movie follows this normative structure, that doesn't inherently make it a good story, though. So in some cases, we see that people really want a by the numbers type of plot structure. In other cases, maybe it's these variations from what you would expect that make for a really good movie or story in the same way that something like jazz music, for example, often a really good jazz solo is something that really defies your expectations. You know, we already have plugins in our word processors that correct our spelling and our grammar and tell us how to make a better sentence. Do you envision that there will come a time where

my word processor is going to tell me, hey, put this, your plot twist needs some sort of a little milepost here or you want to put on a little alert here to make it work better? I think that we're not very far away from that. What I would love, to do in the near future with this is to develop something like that that is more of a feedback system. So we can show you when you are trying to compose a story what it is you seem to be doing. Maybe you think you're setting the stage, but the numbers tell you're not doing it quite as strongly as you think you are, something like that. Do these rules still apply for works written in different languages that might have different rules of grammar or conventions?

They do. And one thing that we looked at with this was how these structures hold up across different cultures, across different times and so on. And we find that these same structures appear over and over again, regardless of these types of factors. When we look at languages that don't have, for example, articles and prepositions, the things that we use in the English language, we just adopt different language analysis methods. But we still find roughly the same patterns regardless of how we approach it. Interesting. Speaking of rules, you know, every rule is made to be broken. There are always exceptions to the rule. Are there texts you looked at that really stood out as just very different from the others? So we also looked at a lot of types of texts that don't follow what you might call a traditional stories, so things like newspaper stories and Supreme Court justice decisions and so on.

And we find that the patterns for these types of texts are very different. They seem to have their own unique structures that tell us something about the relationship between who the authors are and their audiences. Did you also find it true in works that you did not consider to be storytelling? I mean, were they still using the structural methods that you talk about, even when they think they're not telling a story? even when they think they're not telling a story, they're still engaging in a lot of the same underlying processes that a storyteller would use. So even a Supreme Court justice decision

still needs to give you some degree of background information and some degree of unfolding their logic. So even in cases where people aren't engaging explicitly in storytelling, they're still using some of the same techniques to convey information that a good storyteller would use. Are these linguistic clues something that our brains actually use to help decode the structure, or are they just an artifact? In other words, there are only so many ways to practically convey information? Probably a little bit of both. We think that, especially when we talk about very specific parts of language like articles and prepositions, these are more byproducts. There's something that we can look at to tell us what's going on behind the scenes.

Give me an idea of your takeaway message from your research. What surprised you? What things possibly did you expect? I think one of the things that really surprised us was just how consistent this structure was every type of story that we looked at. We thought surely there's going to be differences between cultures. We thought surely that a story that's only 200 words long is going to have a different type of structure than a story that's 20,000 words long. And the fact that we just kept seeing this same structure emerge over and over again

was really the most surprising factor for all of us, I think. I know you have a website where people can put in their own stories and have them analyze for these narrative structures. Tell us about that. So the website that we built, it has a couple of components. We have a part of the site where you can browse the films and the books that we analyze so that you can go in and look at the structures of a lot of different stories that are out in the public. And then we also have a part of the website

where you can put in your own stories and analyze them. And it'll show you the structure of the stories where the staging is occurring the most, where the tension is highest and so on. And tell us about what that website is, please. The website is www.orgofnarrative.com. And your arc of narrative was just terrific today, Dr. Boyd. Thank you so much. We've run out of time. Ryan Boyd is an assistant professor in the Department of Psychology at Lancaster University in Lancaster, UK.

We're going to take a break, and when we come back, what role does accessibility and education play in the contraceptive world? This is Science Friday. I'm Ira Flato. This year marks the 60th anniversary of the pill. You know, that fast and easy oral hormone? that helped to not only change contraception, but also family planning and women's health. But many of the other contraceptives we still see today were first developed way back in the 19th century. So what are some of the contraception conversations that are happening today?

Science Friday producer Alexa Lim has more. The pill ushered in more effective in convenient contraceptives like IUDs. But even with more contraceptive options, the current rate of unintended pregnancies in the U.S. is still around 45%. This comment came in on the Science Friday Vox Pop app from Valissa in Queens. I think what really needs to change other than the actual contraceptors itself is the discussion around contraceptors because I did not use the pill to prevent pregnancy. I used it because I had a medical issue. So how is the discussion changing? We're going to look at how contraception education and policy is evolving and how that could affect usage and access. My guest are Linda Gordon,

who is a historian and professor at New York University, an author of the book The Moral Property of Women, a history of birth control politics in America. Glad to be here. And Cynthia Harper, a professor in the Department of Obstetrics, gynecology, and Reproductive Sciences at the University of California, San Francisco. Thank you so much for having me. Linda, that rate of 45% of unintended pregnancy and seems really high. And it's also been 60 years since the pill was released. So how much progress have we made since then? Actually, the pill did make a big change because it introduced hormones. And that, of course, allowed many, many more people to have easy access to contraception, but it also created problems because putting hormones in your body carries risks. You use a diaphragm,

there's absolutely nothing you can do that would be. dangerous, right? The other great advantage to some people of the pill was that you do not use it when you're having sex. You can do it at another time, and so it is in some ways invisible, and it doesn't immediately associated with sex. Since then, I know, and I'm sure you'll hear more about it, there have been many, many hormonal methods developed. But what really hasn't happened enough is making sure that people who really want safe and healthy birth control to have access to it. Right. And Cynthia, you work in the area of education. I mean, how much does this play into it as well? Well, I couldn't agree more with Window. And we are at a watershed moment currently

in society where we just really need to change our inequalities and change. And to increase access for the people who historically have not had rights and access and who today don't have rights and access. And I believe that education has a large role to play because certain people are left out of knowledge and awareness, access to health care. And fewer people in our society really are able to choose when and how to have their children and then are able to bring them up safely. and with shelter and food. And Cynthia, you do do work in this area where you were looking at IUDs and you did a study providing education to health care providers. Could you talk about that study?

Like, what did that show? Yeah, we do a lot of work in the area of training physicians and nurses across the country in community clinics. And we have found that when nurses and physicians and their clinics are given the resources and are set up properly, then they're able to offer. offer quality counseling to patients who come to them and can enlarge the range of methods that they offer them. And I think it's a really important movement to have birth control methods available outside of clinical settings. I worked long and hard to get the emergency contraceptive pill available

in pharmacies so people could take it in time for it to be effective. And I really think it'll be a big change in this country, what we have in California, you can now get the pill from the pharmacy. And I think that is a change that both Republicans and Democrats agree with and would really help to increase access for people across the country who aren't able to make it to the clinic. That's become a big barrier right now during the COVID pandemic. And a lot of clinics are offering telemedicine for contraception. but the IUD, for example, or the contracept of implant require an in-clin visit, too. Can you talk a little bit about that, the move from emergency contraception to have it, you know,

available over the counter? Okay, so that took advocacy. It took research for, oh boy, I think it was over a decade that we conducted studies. It took the FDA evaluating all the data, and then it was really a big political process was involved, too. And it took the judiciary in the end where I think it was the Eastern District Court of New York, where they had a lawsuit where they decided that, yes, this would become available to all ages. But it really took a lot of different sectors in society to make that happen. The reason it was so important was that because its effectiveness is time limited. And if you have to go to a doctor's office and make a visit and get an appointment, particularly on a weekend or a holiday, then you've miss the window. And so, I mean, is there a bigger move to get things, you know, into clinics and outside of just having them to be prescription and prescriptive? Yes, there is a move for you the pill. And this movement is strong. The research now is taking place. There is advocacy there,

too. There are a lot of pieces to put into place. Our health insurance system has a lot to do with what's available and where. And in order for the pill to be a benefit in the pharmacy to people who need it, it would need to have insurance coverage and pharmacists would also need to be remunerated for dispensing it. Right now, we're not set up like that across the country. We have so many vested interests in our health care system and not to change. But I think now is a critical moment where we all have to agree that our health care system isn't working for us. We do not have the health outcomes that we really should be having in the United States and we could be having. And I'm hopeful that we can change to try to address a lot of these inequalities and care.

I mean, I guess it's a big question, but we do need an overhaul. But I mean, what are those individual pieces that need to be moved? In order to date contraceptives more available to women, the Affordable Care Act, went very far. And there was a mandate and there was a lot of agreement behind it that contraception is an essential health care for women. And I think that's going to be hard to remove going forward. There are pieces of that that are still there. Although there was this Supreme Court decision recently that employers who are religiously affiliated may decide not to offer contraceptive coverage, we do have more wide coverage now. However, with the increasing economic hardship from COVID, a lot of people are losing their health insurance because we have the system where health insurance is employer-based.

So that really helps the haves and really leaves out the have-nots. And I think our next step is to have a health insurance system that is not predicated on the fact that you have a well-paying job with benefit. Right. Linda, and maybe this is too on the nose, but is this just like history repeating itself? I mean, is this the same conversation we just keep having? Well, certainly there have been enormous changes in the last, say, 50 years. Let me go back to a couple of earlier dates. It was not until 1965 that the last state repealed its laws against providing birth control.

Then eight years later, when the Supreme Court ruled, that people should have a right to abortion, at least in the early stages of pregnancy. Just about everyone, all the journalists, all the experts, everyone assume abortion rights are now in a role, is no longer going to be controversial. By the time of the Supreme Court decision, there were already 18 states that had repealed their laws against abortion. What has happened since then, no one could have predicted, and that is a really massive case. campaign, a very well-funded campaign, not only against abortion, but as we're seeing more recently, against providing contraception where it's needed. There's a little bit of a contradiction there because I think most experts that I know of think that the best way to prevent abortions

is to make contraception more easily available, but then politics don't always follow that kind of logic. Anyway, I think we are also in a period in which we have to face that these issues of women's rights to control verse have been turned into a political weapon that has done a lot to change the politics of our country. And I think that's another thing that would have been very surprising in 1973 because, after all, this country faces many, many serious problems and issues. And it seems strange that this very personal choice should be one of those issues that are so controversial. I'm going back to that 45% statistic again, which came from 2011. It seems like there aren't these big studies investigating this.

I mean, why do you think there haven't been more recent studies on this? Do we need more scholarship on this? I think when we think about the 45%, which is very upsetting, we also need to think about the fact that this country has unusually high rates of maternal mortality and infant mortality. So we're talking about failures in the whole system that go together. I will add that we actually are conducting a study and a policy study. And I'd like to offer some hope, too, because we believe that there are ways that we can change our health care system and what we make accessible to young women to improve their health. And it requires looking at large social issues, too.

It's not just a health system issue. And we are conducting a study comparing young women who are in community colleges in California with young women and community colleges in Texas. And our intervention is to improve their access to contraception and care. And we also work with a network of community clinics around where these young young women live in order to train the clinics to open their doors and have quality care. We will see how the policies in California, including the recent COVID policies that are covering very innovative contraceptive care practices, such as the delivery.

There is a new formulation of Depo Pereira that can be administered at home. It can be self-administered or right now some clinics have drive-by, the plaza. for STI testing and for contraceptives. California is also covering, as I mentioned, pharmacy access to contraceptives. They have a lot of coverage in place that helps people to be able to choose a wide variety of methods and also to choose methods in this time of telemedicine.

In Texas, reproductive health care is more restrictive. People have a harder time affording the method that they may want, And this really gets to reproductive, those questions of reproductive autonomy and what they're able to choose when and when they want a child, the conditions in which they have the child. So we are following these women far out for five years to look at their reproductive health outcomes and also to see how they fare in their larger lives because all of these things are closely tied together. And we're looking at their educational pathways, also their entry into the labor market. And we're very hopeful that the kind of safety net that California provides, which is more similar to what you might see in a European country, will result in better outcomes overall, better life outcomes for these young women in addition to better access to reproductive health. the big question will be moving forward. I believe this data will be very useful if we have a change in the administration,

but it also can be useful on the state level for states that are considering whether to cover certain things and whether to pay attention to certain things, such as contraceptive actions for young people and maternal mortality, which is a burning issue in our country right now. I'm Alexa Lim, and this is Science Friday from WNYC Studios. And I know it's probably different state by state, but I mean, do we have overall guidelines about this? Or, I mean, do we need overall guidelines? We do not have overall guidelines. I don't know if overall guidelines are needed.

I do know that it will take all of us in all sectors of society to try to make essential health care available to our population at large. Right. Okay. And I know this has been a wide-ranging conversation. But, I mean, what are your general thoughts on the future of contraceptives and family planning? Like I said, it's been 60 years. Maybe we don't have to look exactly 60 years out. But, you know, what are the key issues? I think what is fundamental here is that people absolutely need the ability to control reproduction.

Part of the evidence for that is that in every society in the past, people have tried to do that. They're not always successful. But it's just too important. It's very important economically. It's very important in terms of the kind of family you want to have. It's been increasingly important as more and more and more families and children depend on women's employment and their wages. Over time, the closest correlation that I've read anyway, and I'm talking over a century of time, Over time, the strongest correlation with use of birth control is women's employment,

because this is something that just makes it really impossible to do a good job as a mother if you cannot control how many kids you want to have. Because of that, I'm actually fairly optimistic because I think so many people need and want this that they can't do without it for long. And we see that in the polls, which are that the... public opinion, even in the so-called red states, is much more favorable toward making contraception and even abortion accessible than the politicians respond to. But ultimately, I think it's just one of the most fundamental things that you do in life, that you have to be able to make these

decisions and have them a way to make the decisions that work for you. And Cynthia, what are your thoughts? I absolutely agree and I also am hopeful. I envision our society where all young people are able to choose in this very important area of their lives and are able to enter the labor force if that's what they want to do, are able to have the children they want to have and are able to have safe and healthy lives. Thanks so much for taking time to talk today. Yeah, thank you so much for having us. Thanks. Linda Gordon is a historian and professor at New York University, an author of the book The Moral Property of Women, a history of birth control politics in America.

And Cynthia Harper is a professor in the departments of obstetrics, gynecology, and reproductive sciences at the University of California, San Francisco. For Science Friday, I'm Alexa Lim. We're going to take a short break, and when we come back, how a roboticist looking for gender bias against robots found something else. Very different. Coming up after the break, stay with us. We'll be right back. This is Science Friday. I'm Ira Flato. Robots are everywhere. They listen to you speak, and they help you teach. Cars with increasing self-driving capacity. There are even robots that tell jokes. I kid you not. And of course, robots have been replacing humans in factories for decades. But as robots continue to add new jobs to their resumes, can we trust them to be

competent. Producer Christy Taylor talked to a researcher about why our biases may actually be the biggest hurdle. We've talked at length on this show about how artificial intelligence algorithms often come with hidden biases, biases that make them either less useful or even downright harmful for marginalized groups like women and people of color. But what about the biases we apply to AI when we encounter it? The robots, the virtual assistance. Those virtual assistants that, by the way are almost always given women's names and voices. Dr. Ianna Howard is a roboticist and chair of interactive computing at Georgia Tech in Atlanta. Last year, while investigating gender bias against robots, she found something unexpected about how we trust and relate to the robots that are

increasingly helping our world run smoothly. Welcome to the show, Dr. Howard. Thank you. I'm excited about this conversation. Yeah, me too. And I really wanted to start actually with some research that you published earlier this spring, you went out looking for, do people have biases against robots based on what gender they presume to be? And you found something quite different. Yes. So we know that in, you know, human, human interactions, there's this assumption that certain jobs are predominantly filled by women versus predominantly filled by guys. Example nursing is mostly female occupation receptionist and things like that. And so what we thought was that if robots had the same type of thing in terms of gender, that you as a person would be more trusting of, say,

a female robot in nursing versus a male robot in nursing and vice versa. So that was our original assumption, which we were trying to basically just grab evidence to see if this was true. But, well, what we found is that, In general, people don't trust any robot. So it had nothing to do with gender. It actually had to do with the robot itself. And so, female robot, nurse, male robot, nurse? It's like, it doesn't matter.

Robot, nurse, no. And so it didn't matter what kind of job the robot was doing. They just don't trust them? They don't trust. I mean, there was like maybe two occupations, but for the most part, they didn't trust. And so what we did find is or discovered is that people, people, when it comes to robots, trust is highly correlated with competency, right? Whereas I think in human-human interactions, we don't have that same.

Like, I see a nurse. I know that they can do their job. And then if, you know, they're male or female, then it's a secondary. Whereas here it was the competency issue was primary. Like, I don't believe that robots are competent. Therefore, I don't trust them. Therefore, I don't care what gender they are. And we threw in occupations that clearly there weren't robots.

And we threw in ones that we were like, oh, these are robot functions and there's companies making a lot of money selling robots in these spaces. One was in surveillance, a robot in the malls, a robot going around parking lots, looking for individuals and things like that. Even surveillance robots, people are like, yeah, no, no, I don't believe that they're competent in that space. Surgery is another one. We have like the Da Vinci system, like surgical robotics is a thing. And yet they were like, yeah, no, robots, surgery, not competent. Why do you think there's this disconnect between what robots can do that you know they can do as a roboticist and the people you did your research with, their trust that robots can do these things that robots can do?

Yeah, I honestly think it's about media. So I said there was one or two jobs where people thought they were competent. One being delivery robot, which, by the way, we don't have. have drones flying packages to us. But if you looked at the media, this is like right around the corner. We have delivery robots that are like at our doors, which isn't a thing. But yet when you see these other roles, media always portrays them in a mistake. So if you Google are being any type of surveillance robot, you'll see these articles about robots falling into pools in the mall. And kids like smearing ketchup on these surveillance robots and the robot like, I don't know what to do

with these kids. What am I supposed to do? Right? And so I think it's that, you know, like, look, these robots, they really aren't that good because they make all these mistakes. And so I really think it's about the portrayal of them in the media. Yeah. Well, so going back to that question you were asking about gender, does this result that we view robots as sort of universally incompetent? Does that mean that we wouldn't react whatsoever to a robot's gender? Or are you finding that that still has a role just in a different way? I think it has a role in a different way. So one of the things I believe is robots become more pervasive, which they are, I think it's

our role not to add gender because then we are adding in these biases and we're just enhancing it. Right. So if I assume that a surgeon is going to be male and I therefore make my robot that's doing surgery have a male gender, I'm just reinforcing these type of stereotypes, which I don't think is good, especially since people at the end of the day care about the competency of the robot itself. I honestly think that we should give people the choice. And they're going to gender, right? At some point, they may decide that, oh, this robot is male and this robot is female, but I think that

power should be given to the user, not to the developer. So you're saying that if we do gender robots, we're actually just exacerbating our own stereotypes? What if we made all those robotic assistants, male robots instead of female robots? So this is interesting. So as a developer, I can actually change the stereotype. But then I would argue we are also then not giving that power into people's hands. Because then I can say, well, why are we gendering based on a binary construct? Right.

So now I'm excluding the fact that we don't just have a binary gender category. Right. And so I still think that whatever we decide, we are not doing it right. We are adding in our own stereotypes, whatever those stereotypes and biases are. But also this is interesting because I think it starts a conversation. As an example, if I have, we'll use surgery, a surgery robot, and I go and I say, oh, I'm going to go see my robot. He is. Someone might say, oh, your robot is a male. Well, mine is a female. Like, that starts a conversation, right? That's like, oh, hey, that's true. Why did I enable mine as male and yours female? I think that provides much more value about inner reflection and having a conversation about biases than, you know, just a blanket. Every surgeon is a male. That's a robot.

Does any of this research you've done tell us more about how to make a robot more trustworthy in general? Yeah. So I honestly believe that robot. should have the ability to talk about their mistakes. Explainable AI is a word for this. And you can look at any program. There is this uncertainty in terms of accuracy. And we know it.

Like before we deploy any robot system, we know how accurate it is and what scenarios and, you know, it might be 10% inaccurate in certain scenarios. I honestly think that the robot should explain what they're doing and explain that inaccuracy such that it's understandable to the person. I think that will enhance trust because then it becomes, again, it puts the power on the user to then decide, oh, okay, I know the robot said it's almost 100% short. Okay, do I believe?

And if I don't, and I still follow the directions, I'm taking that on to myself. So if the robot makes a mistake, I'm like, oh, yeah, but I knew, right? It wasn't going to be 100%. So I'm not going to blame the robot. I'm not going to blame myself. Is that a higher standard than we would hold human beings to in needing to trust them? Yes. And I truly believe that we should hold robots to higher standards. So think about this. We as people, we are imperfect. We each have our own lived experiences, which again makes us bias in multiple, multiple ways. And so why should we give them the lived

experience that we ourselves are biased by. So I totally think that we should put them to a higher standard. And I think they can make us more human because of it. What do you mean by that? Say more. I think in a lot of times when we are called out, called to question about our own biases, a lot of times, not everyone, but most of us do an inner reflection. Right. Like, oh, I didn't realize that what I said might have been a sexist. Right. Oh, why is that? that I don't understand, let's have a conversation. But that only happens when we're called out about saying something. I think a robot has that ability.

To call out biases. To call out biases. So imagine I go into the store and I see a robot receptionist. And I go, hey, sweetie, how are you doing? Right? The robot is like, hey, I'm not a sweetie. In fact, that assumes that you assume I am a female. Why is that?

Not all receptions are female. I think you might need to think about why you thought, Right? Like, you could have like someone being kind of reprimanded a little bit. We didn't talk about racialized robots and bias. I haven't personally observed any robots where I'm like, that is a racialized robot. But is that something people are trying to do in any way or is that something that we have interesting research on? Yes, there was, and we actually had a paper that was an argument against this. But there was a paper about three years ago where a researcher had a racialized, a robot. He used black and white. And what he wanted to do was test the shooter bias scenario,

which of course ethically is like, why would you even do that? And what they showed was that if a robot was racialized as black, people were quicker at shooting the robot down. This actually caused quite a bit of an uproar. And when I looked at that, the researchers had prime the individuals to think about these robots as black and white. Even though in color, they had skin tones of black and white, but they had also primed it. Like they said, what race is this robot? And then they did the shooter bias test.

And so in my group, we actually said, well, what happens if you remove this priming? But people just look at it as a robot, not as a black or white robot. Well, they just look at it as a robot. And people looked at these robots as robots. as long as you didn't label them. Again, I believe that we as developers should not label,

whether that's with respect to gender, that's with respect to race, ethnicity, and all of those human attributes that causes so much interesting conversations, but also harms and biases as well. What do you think the biggest challenges in robotics for the future really are at this point? I think one of the biggest challenges is to make sure that the robots that are created

and the AI systems that are out there work for everyone and are accessible to everyone. I identify as black, I identify as female, and I develop robots. But my training is as a roboticist, right? I was trained a certain way. And so all of us really are trained in a certain way, and we're designing for a world

that doesn't necessarily have the same kind of training. And so our solutions, They're biased even through that, even though I may be diverse in terms of who I am. My thinking process, it still thinks through engineering and problem solving and things like that. That's what I worry about is that we don't have enough diverse, unique voices that are designing the technology, that we might have warmed the world to the thinking of how roboticness speak. I do an AI ethics class now where I'm trying to teach AI developers how to think, how to think about. some of these problems. The other thing that a lot of us are starting to do is bringing in community

input from the beginning, right? So it's not a, as an engineer, sometimes we'll go, I have a solution for your problem versus working with the community to say, okay, let's identify your problem together and develop a solution together. And that's also another way of bringing in diverse voices. I'm Christy Taylor and this is Science Friday from WNYC Studios, talking about how we trust robots and how we bias against them with Dr. Ianna Howard. I was going to ask if you have a favorite pop culture robot. Rosie. From the Jetsons?

Rosie is my favorite. Like if you'd asked me 10 years ago, it was always Rosie, but I didn't know why. Now I know why. So Rosie, she was there to make life better. She was everything to everybody. She changed based on whether it was, you know, George or the kids, right? She had different functions. but she was adaptable. But when you did something that she thought was wrong, she would actually

call you out and hold you accountable for your human mistakes. And at the end of the day, she was social and she was part of the family. And those are all the things I think about now in terms of robots of the future. Another thing I think about with the future is the pandemic. Do you see the pandemic changing the role of robots in our lives or adding more of a certain kind of robot to our lives. The pandemic has definitely accelerated the adoption and use of robotics in certain types of roles. So some of those roles are in things related to cleaning and disinfecting because you want to minimize human contact if there was an area that had some infection or prior infection, right, and robots are there. And there's a whole field now that designs or things about PPE for

robots. Wait, what? Why does a robot need PPE? Well, because If a robot comes in contact, as we know, the virus can stay on the surface. So you don't want the robot then transmitting it to a clean space. It is not to protect the robot. It's protect everyone else. And you can't clean it. Like you can't necessarily clean within the joints and all of that. You can't just dunk it in hand sanitizer?

It'll be very difficult to do. So the other is there's been an increase in the use of robots for, telepresence and telemedicine. And telepresence robots have seen an increase to allow a little bit more of a connection besides looking through a screen on a laptop or a tablet. Having that embodiment, even if it's a robot embodiment that looks kind of strange, gives you a better connection to the person on the other side. It's also been an increased usage in AI that's virtual as well. I saw your TED Talk, and I really liked one of the questions. that you posed in it, which was not just about us trusting robots, but can robots trust us?

Why do you ask that question in the first place? So that has to do with us as humans, us as people, and as our society. Unfortunately, I don't think that robots can yet trust us. And it's just because we're just too diverse in terms of opinions and what we want for the world. And I would say as humans, we are pretty selfish. And robots are the other. And I worry that we as people tend to treat others outside of our community pretty bad. You mentioned kids smearing ketchup on robots.

Exactly. Well, thank you so much for joining us today. Dr. Ayanna Howard is a roboticist and chair at the School of Interactive Computing at Georgia Tech in Atlanta. Thank you so much for joining me today. Thank you. For Science Friday, I'm Christy Taylor. And that's about all the time we have for this hour. If you missed any part of the program or you would like to hear it again, yes, subscribe to our podcasts or ask your smart speaker to play Science Friday.

Every day now is Science Friday. Oh, on our Science Friday Voxpop app this week, we'd like to know how has the pandemic affected your hygiene routines. Tell us on the Science Friday Voxpop app wherever you get your apps. and you can also say hi to us on social media or email us. Our address is SciFri at ScienceFri.com. Send us your feedback, please. We'd like to hear from you. We'll see you next week.

I'm Iraflato.