- 18 minutes ago
Category
🛠️
LifestyleTranscript
00:00And I hope that the information that we provide to you this morning is not only useful, but
00:05helps you in transforming your lives moving forward.
00:09Thank you again for joining us.
00:11Ladies and gentlemen, please welcome Dr. Kimberly Ellison.
00:31Good morning, ladies and gentlemen, and esteemed guests.
00:48It is with great enthusiasm that I welcome you all today to this discussion as we embark
00:54upon this thought-provoking journey in the world of artificial intelligence and bias in
01:00healthcare and employment.
01:03This is an important discussion that impacts the two critical areas that AI holds.
01:09It holds immense potential, but it also raises grave concern.
01:14So our purpose here today is to explore the artificial intelligence as it relates to using
01:20keywords and the keyword-identifying tools, and then can it either exasperate the existing
01:26healthcare disparities and or exclude those black candidates from employment opportunities.
01:32So we aim today to shed light on these potential pitfalls.
01:36And more importantly, on today, we are going to identify strategies to navigate them effectively.
01:43Well, I can't do that alone.
01:45So to guide us through today's exploration, we have brought together a remarkable panel of
01:51esteemed business executives.
01:54These individuals have extensive expertise in their respective fields, and they possess,
02:00again, the insights to the present conditions as it relates to the future implications and
02:06the current conditions of AI and bias in healthcare and employment.
02:11Today, their collective knowledge and experiences will undoubtedly provide us with the invaluable perspectives
02:19about this pressing issue.
02:21So if you will, help me welcome to the stage on today, Sule Sandy, Associate Partner at McKinsey
02:28& Company, Kelsey Ruger, Chief Product and Technology Officer at Hello Alice, and last but definitely
02:42not least, Brenda Darden Wilkerson, President and CEO of AnitaB.org.
02:48Welcome.
02:49Welcome.
02:50Welcome, panelists.
02:52I'm excited.
02:53Are you all excited?
02:56All right.
02:57So through our discussion today, what we hope to gain, again, is a comprehensive understanding.
03:02All right?
03:03You all are experts in this industry, in this field.
03:06And so we want you to share with us today some of the current.
03:10But as we have, again, at GBF and in this conversation, we're talking about the future of health.
03:16So Brenda, we're going to start this journey with you.
03:22Okay.
03:24As an advocate for access, opportunity, and social justice for historically excluded communities
03:31and technology, what are the different ways that AI bias can manifest in healthcare?
03:38Well, it starts with the data that the AI is trained on.
03:44So the data that's been gathered historically has been based on the biases that people have
03:50experienced in healthcare.
03:51And we know that black people in particular have had the hardest time of having been judged
03:58correctly on the level of their illnesses.
04:00And so if that data says that you're at low risk for illnesses, then therefore, many times
04:06in the past, you've had to be, in the current situation, you've had to be sicker to even
04:12be treated, right?
04:14So that AI has been trained on that data.
04:17Second to that, the people who are creating the AI, and the problem with it is it's black
04:24boxed, which means that others can't look at it and really figure out how it's coming
04:29to those answers.
04:32Those people who are creating that AI, they're pretty homogenous.
04:36And many times we're not represented there.
04:38So our experiences are not validated.
04:42The errors that are about our experiences are not validated.
04:45And so that perpetuates that bias and the data just keeps on ticking.
04:49Wow.
04:50Right.
04:51Wow.
04:52And you know, when you think about it, Brenda, considering these historical biases that black
04:56people have to be sicker, can take more pain, and all of those challenges that create those
05:02disparities in our communities, can you share with me what measures, what are the measures
05:08that are being taken to ensure that the implementation of artificial intelligence in healthcare does
05:14not perpetuate or continue to perpetuate or exasperate these existing inequalities?
05:19Well, here's what needs to be done.
05:22What needs to be done is that data needs to be checked for biases.
05:26And the best way to do that is to put people at the table who understand what those biases
05:30are and can take a look at them.
05:33For instance, if you have someone at that table who understands the illnesses, the historical
05:39illnesses of, say, black women, and they're looking at the data and can tell that that data
05:43has been mostly gathered about men, they're going to notice that unless they have the right
05:50to do and make those changes.
05:52And so we need that table where tech is created to be much more diverse, and it needs to be
05:59age-diverse.
06:00It needs to be physical ability-averse so that those algorithms can be corrected.
06:07And as it's executed in the healthcare industry, there needs to be studies to make sure that
06:13they study the differences between the different segments to see, okay, what's supposed to
06:18be happening and what actually happened with those segments.
06:21And then go back and take a look at that AI and make those corrections.
06:24It needs to be constantly measured and constantly corrected.
06:28Wow.
06:29Wow.
06:30And I mean, think about it.
06:31Yeah.
06:32When measuring data and having the facts to drive that data, again, will give us some outcomes
06:37or better outcomes than we've had historically.
06:39That's right.
06:40That's what we hope for.
06:41All right.
06:42Well, Kelsey, listen, I want to get your voice in this conversation, all right?
06:47Being a visionary leader in the tech industry for over 20 years and your expertise in product
06:53development, design, and engineering, in your current role as a chief product and technology
06:58officer, how can we ensure that AI-driven solutions do not inadvertently create new barriers?
07:05Well, one of the things that we talk about a lot is you can create those barriers without
07:11ever thinking about it because of the way we operate in technology.
07:15And so I always tell people it's a combination of best practices and policy.
07:20We haven't really gotten to the policy yet in the U.S., but a lot of what we've learned
07:25through dealing with PCI and SOC 2 and privacy concerns can be applied directly to what we're
07:32doing with AI.
07:33And so I think when you're starting, we talk about the three A's.
07:37We talk about whether the AI needs to just augment, whether it needs to completely automate,
07:44or whether it should just leave people to do their work autonomously.
07:48And when you build solutions, really thinking about which one of those categories it falls
07:53into and how you should build it, whether you're an engineer, a product manager, a designer,
08:00really understanding how you should implement the AI helps you build best practices.
08:05But even with those best practices, you have to have policies in place.
08:10We deal with policies every day for privacy concerns.
08:13It's why we have CCPA.
08:15It's why we have GDPR.
08:17I think some of the same measures will have to be put in place for AI just to keep the checks
08:22and balances in place because the technology can't outpace what we're either prepared for
08:28to deal with as consumers or what we've put the measures in to protect ourselves from.
08:32Okay, so you talk about the three A's, right?
08:35And so your automation, augmentation, and then what was the third one?
08:39Autonomous, and we don't think about that one a lot because sometimes there is an intrinsic benefit
08:46to remaining human and doing it yourself and not letting the machine do it for you.
08:51And that can bring a challenge.
08:52And so how can we then strike a balance between the benefits of these AI-driven solutions
08:59and then the potential pitfalls as it relates to deployment?
09:03So I actually borrow a lot from the world of accessibility to having worked with web accessibility a lot.
09:10One of the things that they say a lot is nothing for us without us.
09:15And I think when you're designing for people, you have to involve the people so that they can talk
09:21about whether they're comfortable with it, what might be crossing the line.
09:25And as we talked about earlier, really examining how the data has been trained
09:31and looking at how those biases might have been introduced.
09:35So thank you. And Sule, I would like to hear from you as well.
09:41In light of the increasing utilization of AI in healthcare, how do you foresee technology impacting
09:50the roles and responsibilities of black healthcare professionals?
09:53Yeah, I think it's a great question. So if you look, if you just look at the basic statistics,
10:01I think there's studies that say about 30%, somewhere between 30, 35% of the roles of black folks
10:08and are mostly roles that are going to be impacted by automation and AI, right?
10:13So if you're doing something that a machine can do or can help do, that's likely to be impacted.
10:19So for black folks, that's a big number, correct? And then you add to that the whole thing around
10:26generative AI, I think since last November, everybody knows what that is.
10:31That's not only limited to folks that do service jobs or jobs that are going to be
10:38hit by automation, but actually white collar workers, lawyers, right?
10:43Generative AI would also have an influence on that. So I think as it is right now,
10:48it will, it will severely impact black folks if we don't take the right actions to actually get people
10:55trained or get them into the right, you know, organizations and situations that would actually
11:01help them actually, you know, drive further growth.
11:06So when you think about the education and training for black professionals in healthcare,
11:14Kelsey, can you build on what Sule just shared and answered this as it relates to black male executive
11:21leaders in tech? You know, you both have this influence and you have voice on this topic.
11:27So what steps do we need to take or what steps need to be taken to ensure that deployment,
11:33training and the implementation of these technologies in healthcare are more inclusive?
11:38You know, how do we do that in representing the diverse perspectives or black healthcare professionals?
11:43Yeah, I think again, building on the nothing for us without us, I think you have to have people who
11:49are representative of groups that may be biased against involved in the process. And so one of the
11:57things that I think to explain about AI is it looks for like patterns. And so by default,
12:04there's going to be bias built in. And so I think you have to bring people in
12:08to enrich the data set. And again, also teach people how to build the prompts. And so that's
12:14really where we are right now is how do you prompt the AI to get the information out that you want
12:20and teaching people how to ask the questions the right way to get that information.
12:26Sula, will you add to that as well?
12:29Um, I know we're talking about AI here, but I think it's actually when you when you think of talent,
12:36it's much broader than that, right? So a lot of the work we do, I just I'll give a simple example
12:42here. And it's not AI related. But I was at a we were having this big kickoff with a client,
12:49like over 200 people in the room. And it was a healthcare, it was a life sciences company. And
12:55they talked about the different drugs that they had, and they had pictures. And you know, for
13:02for whatever reason, the picture about the AIDS drug was this really thin, malnourished looking
13:09black person, right? And I sat in that room, and I felt uncomfortable, but no one else did.
13:15But I think that goes to show that if you if we're going to build these AI solutions that are going to
13:20solve these disparities, we actually have to have black people at the table in terms of data scientists,
13:26the executives, you know, the black folks are about 12% of the working, you know, the workforce
13:33today, when you look at the data scientists, it's probably about four or 5%. I think that's critical
13:39to helping build these, these systems that can actually help us level their playing field.
13:44Because if not, that continues to perpetuate the bias.
13:47Exactly. As we talk about measuring the data. So Brenda, I'm coming to you, we've discussed the biases,
13:54we've discussed the barriers, right? And you know, let's talk about some solutions for the future of AI
14:01and healthcare. And how does that impact our community? So in the future, if you could look
14:06ahead, how do you think AI and these emerging technologies can be used to enhance cultural
14:13competence, as we were speaking of shortly here with Sule. So how can you feel? How do you feel that
14:20it can enhance cultural competence in healthcare delivery? And the overall patient experience?
14:27Well, if it's done right, then you're going to have more use cases that will be tested to be rolled out
14:34in the community. And so if you have people with various overlapping diseases, and the AI has been
14:41trained to recognize it on a cultural level, then you're going to have an increase in better
14:48diagnoses. Right now, black people are misdiagnosed at such a high level, and at such a high rate, that
14:56treatment becomes problematic. Also, many times, there are not medical facilities in our community.
15:02And so if the AI is enhanced, and it works better for us, then remote health can be one of the
15:08solutions that actually help increase our health. But one of the things that I want to add on in terms
15:13of having more black people at the table, when you think about, we've talked a lot about in the press
15:19about AI and facial recognition being a problem. But there's also a problem with voice recognition.
15:26And so in those times when you're using, for instance, I've had an iPhone for over 12 years,
15:33Siri still can't understand me. So if you've got an elderly person who you're trying to help
15:40medically, and they're talking to these systems, and there are not people backing it up, I love what
15:44you say about there's got to be people involved. It can't just be about the AI and the machines.
15:50But you many times have to get past the machine with your voice in order to get to those people.
15:55And so when those systems are better trained on various voices as well,
16:00then there'll be a better delivery for healthcare in our community.
16:03Wow. Thank you for that. Again, you know, Sule, I would love to give you,
16:10have you give us a look into the future. If you think about looking ahead, what opportunities and
16:16challenges do you envision for black healthcare professionals in leveraging the innovation?
16:21Yeah, I think fundamentally, in terms of opportunities, I think black folks have to be at the table,
16:29as we've talked about. Top executive levels, the people that can actually spend money on AI,
16:36but also the people that actually build the AI, plus the people that provide contextual information
16:43in terms of what the AI does. On that third point, as an example, if you're building an AI system that's
16:49about diagnosing a particular disease, you should speak to doctors. But I not only want you to speak
16:55to doctors, speak to actually black doctors, correct? So, you know, a black OBGYN and a white OBGYN
17:01have different perspectives in terms of the care of women, correct? So I think that's key.
17:05But what also needs to happen is, a lot of the recruiting today, whether we believe it or not, is,
17:14you know, people mostly go to the, you know, the Stanfords, the Harvards, and there's not a lot of black
17:22folks there, for obvious reasons. You know, various, you know, socioeconomic reasons people can't make it
17:30there. So I think what companies need to do, and one of the things that actually McKinsey is doing,
17:34we're going to the talent where they are. So we're going to HBCUs, and it's not only HBCUs,
17:39but you actually go to places where you think the talent is, and also have people in your
17:44organizations that other people can look up to and say, actually, there's a black person there
17:48that's doing this work. And I think there's also things in terms of policy, right? So how do you,
17:54I mean, the Supreme Court, you know, just had a decision and affirmative action. But how do we actually
18:00also get people into higher education, correct? Into STEM subjects? Because I think that's where
18:07the world is going.
18:10I'd love to add that it starts in education.
18:13And so we need more black professors in those schools, not just for the black students that
18:21are going to come through, but for everybody else to normalize the fact that this is where the knowledge
18:27base comes from, they have the experiences, they can train up the non-black doctors to recognize
18:33things that are key to our community. They also need to make sure that we do cross-functional education,
18:40making sure that the people who are doing tech understand these medical subjects.
18:45I think there's a, I'm going to forget his name, there's a medical student here who has called out a lot
18:51of the issues in medical school that the tech is being trained on that is not recognizable for black
18:57people. For instance, in our organization, we do pitch competitions. And one of the winners a few years
19:02ago was actually a black doctor who created a device to recognize bed sores in black people. Black and
19:11brown people are the majority of people who suffer and die from bed sores. Yet the training around it for
19:18those devices is generally done on white skin. And so getting those professors in there for both the
19:24black students and for the white students is going to be key. And Kelsey, as we wrap it up,
19:31what advice would you give to black healthcare professionals today as we look forward to the future
19:37of AI? So even though I'm, you know, very cautious about the bias, I'm actually very optimistic about
19:45where we can go with the AI. I think preparing yourself for a world where you have a digital
19:51assistant that's going to give you some of the answers to help make your job either more productive
19:57or faster, and just getting comfortable with that. Because I think today, when we think of AI,
20:02we think of the Terminator, right? So, but it's not going to be that extreme. I think there will be a time
20:08when it's a part of all of our lives and we just get comfortable with living with that assistant
20:14on a daily basis. Wow. Thank you. And thank you all for joining this discussion on this very
20:19important topic on today. It is clear that the conversations about AI has the potential to either
20:25exasperate the disparities or be a catalyst, again, to drive healthcare forward for all. So on today,
20:32we thank you, give our esteemed panelists a round of applause, and I encourage each of you to take
20:37these insights back into your communities so we can be more diverse and inclusive as it relates to
20:42healthcare and AI. Thank you.
Comments