Ted Audio Collective.
HBR presents. Hi everyone.
You're listening to After Hours.
I'm Yang Mi and I'm here with Felix and Mahir.
Hey guys. Hey Yang Mi.
Hey Yang Mi. Hey Felix.
So do you know what this is?
It's After Hours. On the HBR network.
This is our 50th episode.
This is number 5 -0.
Oh wow. We should celebrate.
Yes. We should celebrate.
So does it feel like one of those things where it can't possibly be 50 because they've gone by so quickly?
Or actually it feels like we've done 300.
A little bit of both, I have to say.
I can't imagine that it's 50.
It feels like 10 or something.
It's been great. You said it yesterday?
Exactly. So both of you guys brought topics in tonight, right?
Yeah. So I was hoping we could address the Sackler controversies.
The Sacklers are the family behind Purdue Pharma, which is the company behind OxyContin.
And there's been a number of efforts to hold them, not just the company but also the family, responsible for a number of things associated with the opioid national emergency that we have.
And so it's a really tough situation, and I really want to get your perspective on that.
Wow. That's a good topic.
Yeah. A hard one, I think.
Well, we've been doing this 50 episodes now, so we've got to step up the pace a little.
Sure. And I would love to talk about artificial intelligence and particular applications in the human resource space, where lots of companies are now using AI to determine who gets to show up at an interview.
You know, there are so many companies that are moving to screening applicants using technology now.
Yeah. It's really surprising.
It happens in all kinds of interesting ways.
I think that will be a great conversation.
Okay, Mahir, you want to get us going?
Yeah. So as I mentioned, Purdue Pharma, which is owned by the Sackler family, has come under fire for all of its behavior associated with the propagation of opioids and specifically OxyContin.
And, you know, without going into too much detail on this, it's been a real national disaster in the United States with maybe 200 ,000 deaths associated with the misuse of painkillers.
So what has happened recently is that litigation has been stepped up against the family and the company for playing a role in propagating the opioid disaster.
And so specifically, there's something like, you know, 2 ,000 lawsuits that are now in the works against the family and the company to hold them accountable.
And, you know, the issues are kind of twofold.
First, there's an issue about whether or not the practices that they engaged in selling and distributing it were beyond the pale.
And specifically, there were distributors and doctors who were really running mills where they were just prescribing this stuff like crazy.
And whether they kind of overstepped the bounds in overlooking that and possibly even propagating it and building on it.
And then the second thing is you have now owners and a board of directors being held accountable for actions that seem beyond the pale because they're ultimately the ones who are supposed to oversee management.
The claim is the Sacklers, as the family, were actively involved.
So I'm curious, is what you see happening in Purdue and Sackler very different from what you see in other parts of the pharma space?
Is it something that is beyond the pale or are we actually starting to kind of step over the bounds of kind of holding people responsible for things that they couldn't possibly be held responsible for?
What do you make of it?
So obviously, many of the details, what management did and didn't do and what the board of directors didn't do.
And then in particular, what the Sackler family, how much involvement there really was.
I think all of this is a matter of dispute.
So we will just have to see to really learn.
I don't think we have a good sense of the family's view of the specific allegations.
And so we will have to see how all of that plays out.
The two things that seem pretty clear, I think, is one that at a point in time in the 1990s, the industry has its share of responsibility for underplaying the long run effects of taking these paid medications.
And I think that's a first set of really problematic behaviors.
And then the second set, there were very specific incentive schemes that reward the salespeople for, quote unquote, success with doctors, meaning the more likely it was that you would prescribe the kind of painkiller that would get people addicted, the greater the rewards.
And so I think that's one of these instances where you can tell a story about how the industry contributed to the current epidemic.
And let's just pause on that, Felix.
That sales practice.
Does that strike you as wrong?
I mean, don't we see that all over the place?
I mean, the part that strikes me wrong is that you would have an incentive system that essentially is perfectly aligned with the likelihood that someone would become addicted.
And I think that's hard to square with best medical practice.
So, Youngmi, what do you make of all this?
And in particular, have you thought about the Sacklers in particular?
Yeah. So I'll say a few things.
One is the focus on the Sackler family.
They're just one piece of a much larger ecosystem that includes not only other pharma companies, but the entire pharma supply chain, which includes wholesalers and distributors, as well as doctors.
The second thing I'll say is that I think the specifics really matter here.
So, for example, if you market, say, an anti -inflammatory drug and you know there is some sloppiness in the supply chain, such that, you know, 2%, 3 % of the sales of your drug, probably shouldn't have happened, that's one thing, okay?
But if, let me give you an example, so McKesson, which is the nation's largest drug distributor, they shipped nearly a hundred million doses of opioids over a six -year period.
There's a tiny county in West Virginia with a population of fewer than 25 ,000 people that received 1 .2 million doses in that period.
There's another tiny town with 400 residents.
McKesson shipped three million prescriptions to a single pharmacy in that town in a 10 -month period.
So the details really matter here.
There is a gray area, but there's also some areas that are more strongly black and white.
But, you know, the McKesson example you just pointed to, they are gratuitous and they are, you know, really horrible.
And I think managers should be held accountable for that.
Once you get to the board level, it's less clear to me, and it's less clear to me that the Sackler's new, and, you know, it's all getting parsed out and it's quite complicated.
So why do you think there has been so much focus on this particular family?
I think there's a couple things going on.
One is Purdue is very important, and they are the owners of it.
And so Purdue was the OxyContin family, and they have literally reaped billions and billions of dollars off of it.
Simultaneous with that, the Sacklers have attained a great deal of social prestige by donating monies to the non -for -profit world.
And there's a sense in which people feel that they're having it both ways.
And, in fact, all the philanthropies that have taken their money, including our university, are somehow complicit in that laundering of their reputation.
And I think that makes them just an enormous target.
So they are the natural embodiment and manifestation of the entire opioid crisis.
And they've, what appears to be, tried to put a sheen or veneer around who they are via philanthropic efforts, and that might have masked what they've been doing in the product markets.
You know, the second piece of this that's really interesting, in the pharmaceutical industry, they are also prominent for a reason that's less well -known to the public, which is that Arthur Sackler, one of the original Sackler brothers, he came from the advertising industry.
He created a pharmaceutical company.
His company was the first one that began marketing to physicians, first one to put them on the payroll as consultants, first one to bankroll academic studies, first one to send doctors on all expenses, pay trips to conferences.
And so a lot of the tactics that many people consider to be unsavory in this industry originated with Arthur Sackler.
But what do you make of all that?
So, I mean, that was, you know, in a way a very searing indictment.
Or was it meant as a searing indictment?
Which is, is this family one who has changed pharma in a way that is highly problematic?
And should they then, as a consequence, be held responsible for that?
Yeah. So, I mean, I think this is the challenge with management in general, and this is why credibility in management matters so much, whether we're talking about Wells Fargo or talking about pharmaceutical industries, which is that techniques that you could say are really constructive techniques, techniques
that are designed to make sure that the right drugs get in the hands of the right patients can then become destructive when they are overused or used without the proper judgment.
So many of these marketing techniques, you could absolutely argue that if you have a drug that you really believe in, why not try to spend a whole bunch of time with doctors and enlist their help in making sure that those drugs get in the hands of the right patients?
Exactly. The tactics become unsavory when you begin to cross this line, but this is judgment, right?
And this is why I think the details, the specifics really matter and why, at the very least, our system should be able to pluck out the most egregious examples of bad behavior and hold those executives and companies accountable.
I agree. Sometimes it's more great in black and white, but it's also true that the pharmaceutical industry is just one of these industries where it's not hard at all to find companies that seem completely toned to what is happening in their surroundings.
So I give you one current example.
We're trying to deal with this terrible epidemic.
The main drug that is used to bring people back once they overdose has just tripled in price.
Even though this is a drug that has been around for a long time in an environment where there's some culpability of the industry, I find these kinds of tactics, I find just completely inexplicable.
What are you thinking?
Yeah, I take both of your points, which is this is a very fact -based inquiry.
We need to understand better what's actually happening and people have to be held accountable.
I'm curious if you think, on average, we are erring on the side of being too generous towards management and boards, and actually we should be going after them harder.
Or if you think, on average, if anything, we're being too soft.
So I actually think we go back and forth.
It's astounding to me that there were all these claims about, oh, the drugs are not that addictive, and that that didn't lead to almost immediate backlash among researchers, also the medical community.
And then I think once we see the catastrophe, we very often overdo it.
So you now have a system that has become far more restrictive when it comes to access to opioids, down to now many instances where, say, cancer patients cannot really get the pain medication that they really need because we have tightened the restrictions around access so much.
So the pattern, I think, is that we seem to be too blue -eyed to begin with, and then we crack down in ways that are sometimes inconsistent with thinking what's best for patients.
What do you say, Young?
I think, in general, we tend to be very soft on companies.
It's hard for me to imagine that there aren't other cases of bad behavior that are happening every day, that we just, because they're less visible, companies are able to kind of get by.
And I do think, in general, in the pharmaceutical industry, I think there is a lot of evidence that the marketing techniques have become so sophisticated, I'm using air quotes, so sophisticated to the point where you could argue that it's having a counterproductive effect not only on patients but the overall
cost structure of the entire industry.
What do you think, Mayor?
So I think, I mean, I think two things.
I think Felix is absolutely right at some high -frequency level, meaning on a year -to -year basis, we kind of oscillate between being too tough and then too easy.
And I think you're right at kind of a lower frequency, meaning there's been now a decade or two decades of, I think, not being tough enough on misbehavior at the corporate level.
I think what's problematic about this is I think what is happening now is also married to a larger kind of populist moment in our society and the wanting to hold powerful interests accountable.
And I think that's fine, but I do think there are all these collateral consequences that we don't fully understand.
When we start to hold, for example, executives and boards responsible for things that it's really hard for them to understand that they knew they were doing, there is going to be collateral consequences which are not great.
Okay, but you can't have it both ways, right?
I mean, executive compensation in this country is higher than in any other country in the world.
And so if we're going to reward executives when they do well, then we, you know, I think the converse has to be true.
Okay, but this is a really tough topic.
I would love to know where you think this shakes out.
Is Purdue around in five years?
What has all this litigation led to?
What does the world look like?
So I'm hoping for a tobacco settlement kind of outcome where two things really happen.
One is I think what the tobacco settlement did is we had a pretty good understanding of where the companies lied, where they told us the truth, where they didn't tell us the truth.
And I think that's really important to understand more generally, like how much can we trust these claims that come out of industry?
And then I think the second really interesting thing that happened with the tobacco settlement, which maybe we could do even better than we did back then, is that it created a stream of funds that become available to treat patients.
I think in general that this is an industry that I think for some time is going to be saddled with helping to pay for the remediation of the situation that they helped create.
What do you think, Yungmi?
Yeah, I agree. I see this heading toward some kind of consolidated settlement.
It's hard for me to imagine a scenario where they're not able to demonstrate that these companies engaged in deceptive marketing practices, that they engage in negligence.
I mean, I don't know.
We'll have to see. But the difference though, I think the master settlement for Big Tobacco was about $200 billion.
The size of the settlement, I think, if it comes to that will be interesting to see if it's in the dollar amounts approaching that of Big Tobacco and whether or not it's in dollar amounts that would actually make a meaningful difference in helping these states somehow combat this national health catastrophe.
To throw my two cents in, first, I think some estimates of the dollar costs of that disaster are upwards of half a trillion.
So there are estimates out there that are just incredible.
The second thing to say about this is I think actually the better analogy might be asbestos more than tobacco, which is I can see this leading to the end of Purdue, and I think that might be okay.
And that kind of asbestos like situation I think might be better than tobacco.
I think the ultimate signal has to be a little bit stronger in a way than we did with tobacco.
And remember, tobacco, in some sense, was saved by that settlement.
So I don't know. I think I'd like to see something a little bit stronger, maybe a little more along the lines of what happened with asbestos way back when.
Okay, well, that was that's a tough topic, but thank you guys and help me think about it.
Felix, so you wanted to talk about robot interviews?
Yes. So, you know, the conversation about AI is everywhere these days.
And then one of my intuitions is, who knows, I mean, the three of us, will we really ever see fully autonomous cars?
Maybe, maybe not. Will you ever really look forward to speaking with a chatbot?
Maybe, maybe not. So my general attitude towards AI is like all these stories out there, and I don't really know how many of them are real.
And then you have this corner of interview preparation and selection of candidates for interviews, where I think the number of companies and the names of the companies in this context was really just super astounding to me how quickly this happens.
So this is kind of AI meets HR?
Yes, yeah, it's one way to put it.
And so I should maybe say, in the typical process, it's essentially often used to select the candidates that you're going to see.
So I'll give you some numbers.
Vodafone, they interview about five candidates for every job opening that they have.
That means in their case, they have about 5 ,000 interviews annually.
And now they have AI look at 100 ,000 people who apply for these jobs.
They use a company called HireVue, where you go online, you apply for a job, and they will videotape you while you talk about yourself and the job and why you think it's a good match.
The software will look at micro expressions, sort of these small little expressions in your face that you can't really control.
It will look at whether or not you smile if you apply for a customer service job.
And all of that is then used to determine who are we going to meet in person.
So it's almost like an extra step that we didn't have before.
Now, if you had a scale from, oh my God, this is horrible.
This is the best thing that happened to HR in a very long time.
I know your academics are going to come out somewhere in the middle, but are you closer to the left or are you closer to the right?
I would, uh, seven.
And you haven't even said which is higher or low.
I decided to make it a 10 point scale because the whole left right thing I was confused by.
So Unilever uses HireVue as well.
And as you described, they use this as the first step in the job application process, but not as the only step.
And there are other steps that of course involve human beings.
So I think there can be unintended negative consequences, which we should absolutely get into, but there are some unintended or maybe intended positive consequences as well.
So one of them is it sort of dramatically increases the size of the opening funnel.
And for a company like Unilever, what happened once they moved to this process is that they started to get much greater diversity in the pool of people applying.
And the diversity was not just around gender or ethnicity and that, but socioeconomic diversity.
So now is it perfect?
No, it's not perfect.
And we can talk about all the reasons why humans absolutely still have to continue to be part of this process.
But it has opened my eyes to the potential of this kind of technology.
Yeah. So me here, YoungMe was a seven.
Where do you come out?
I'm going to be even worse, maybe a five.
So I think, you know, the other piece of it, YoungMe is obviously it's great that it makes the funnel bigger.
That's terrific. You know, the part that I worry about, I just feel like this is a recipe for kind of flattening out and homogenizing populations of people, which is, you know, what we like is we like people who smile and we like people who smile this way when people say funny things.
And that to me is really depressing.
And this is not a rational, I just feel like, well, yeah, I kind of do.
I mean, I think there's just this flattening.
I don't know how else to talk about this, which is just, it feels like, oh yeah, there is a right way to behave to get a job.
And that, to me, is kind of like the worst aspects of corporate behavior.
The interview process is already fraught with a game where, oh, what do they want?
I need to give them what they want.
And now we are going to encode that.
And then it becomes a game that is even more prone to manipulation.
And God, it feels depressing.
There's actually two radically different approaches, and I think they bear on this concern that you just pointed out to me here.
One is, if you look at companies like Talind Bin, Pharma, Interviewed, they're really trying to take information about you and make it accessible in a way that serves the HR manager.
So, for instance, I test you on skills.
Do you really know how to use Microsoft Excel?
So some of it I think is just a much more sophisticated provision of information about candidates.
The second approach, which I think is actually a more interesting one to think about, is where companies like Roundpeg, companies like Kuru, what they're trying to do is they're trying to match applicants to, say, the cultural environment in the firm.
So they would say, who are the most successful people in a particular organization?
And then you train the algorithm to figure out the reasons for success.
And then you look in the applicant pool, who matches that level of success?
That, I think, is probably closer to what you're concerned about me here.
But what I like about this space as a whole is that there's real competition.
There's different approaches.
And some of these will prove very useful and some of these will prove more problematic.
Whenever I think about these deep questions of whether or not AI can be a help or a hindrance, I'm always reminded of the story of IBM's Deep Blue.
So in 1997, IBM's Deep Blue became the first to beat a reigning world champion, Gary Kasparov.
And this was huge news because now computers were smarter than humans at a game that required a really unique kind of intelligence.
What a lot of people don't know is the aftermath of that game was that Kasparov then came up with this idea for a new form of chess in which humans and computers work together.
And he called this advanced chess, or it's called cyborg chess, or centaur chess.
There are lots of words for it.
But basically it involves combining a human that plays alongside a machine.
And today, in 2019, this is now considered to be the strongest chess playing entity possible.
When you combine a human with a machine, you can beat a machine or you can beat a human.
So one of the reasons I like this story is I think it reinforces that, number one, you need this combination.
You need the humans to make the machines smarter.
But you also need the combination to make the humans smarter, too.
In other words, we do have biases as humans.
Now take us back to our classroom.
In our classrooms, we evaluate our students based upon how they participate in our classrooms.
And as professors, we have biases in how we listen, who we call on.
And so we introduced a tool that enabled us to track our own biases with respect to how we are interacting with the students in our classroom.
And what it has done is it's created a feedback loop that I think has made us as professors just much better, much more sensitive, much more aware of our biases.
But Yangmi, if you want to, let's go to our classroom.
Let's go to our classroom.
Let's use AI practices to hire faculty.
Are you good with that?
We spend an enormous amount of time interviewing PhD candidates.
We spend an enormous amount of time bringing them to campus and lots of things.
How about we do this?
Let's videotape them in a classroom, and then we'll score people, and then we'll allow that to be the screening mechanism.
Are you okay with that?
Well, if you came to me with that idea, I would come back to you with a different idea, which is, right now, when we recruit faculty, where do we look?
We look to all the top schools, and we look at the top candidates from the top schools.
Now, you could argue that there's the real logic behind how we do that, but it's hard for me to imagine that there haven't been times in our history as an institution where we have completely overlooked candidates because they emerge out of places we would have never thought to look.
So imagine we had some mechanism that was AI -based where anybody in the world could apply to be a faculty member at the Harvard Business School, and we used it as sort of an initial screen.
We still did all of our other things, but we used it as an initial screen, and if it looked interesting, we would then invite them to our campus, and then we would go through a more human -based approach.
Okay, but just to be clear, that goes to like a funnel approach, and I get that.
You wouldn't widen the funnel, but what about videotaping everybody who comes and teaches and parameterizing what it means to be a good instructor via an algorithm, and then scoring people on that algorithm and making decisions on that basis?
I think more information is better than less information.
Why wouldn't you use what you can learn from a videotape of someone teaching a class?
If that's somehow correlated with students learning, why wouldn't you use that piece of information?
I mean, I think the concern is that hard information, and you've heard this before, hard information crowds out soft information.
So that's just a way of saying, oh, so now I have some data on micro ticks that Felix says is very predictive, and that will begin to crowd out the soft information of the community's consensus about what certain values are and what kind of people are good people to hire on the faculty.
I hear you, Mahir, but if you think about it, there are so many industries that have a history of relying on the soft stuff, and the soft stuff has worked in favor of a particular kind of individual.
The truth is, the soft stuff that you refer to is often a euphemism, in many cases, for bias, for people being able to use their discretion to hire people who are just like them, that they're comfortable with, that look like them, that act like them, that talk like them.
And so having some counterbalance to that might not be a bad thing.
Right, but in our example, Young Mee Moon, fantastic instructor that she is, is charged with designing the algorithm to screen our candidates, who are going to be potentially good instructors.
Now, what do we end up with in the faculty?
I think what we're going to end up with is a bunch of Young Mee Moons.
And what's the problem with that, you ask, right?
I'm liking this part more.
Exactly. But don't you see the possibility that this flattening happens?
And that is homogenizing.
So I think there's a really interesting aspect of AI here that is important.
One of the big advantages of machine learning is that you discover all these second and third order correlations that you as a person have no chance of ever seeing.
So for instance, if it's the person who smiles, and at the same time is really tough, and at the same time speaks really fast and in ways that are even challenging to follow.
If it's like that combination, if that ends up being what students experience as a really valuable experience, AI is much more likely to pick up on these combinations of attributes really that can make a difference.
And so that I think is a reason to think that we will see these really unusual candidates where we can't even understand why on earth would you say this is a winning combination where it doesn't really make intuitive sense for us.
And yet AI may have found something that we wouldn't easily see the way that American orchestras got rid of bias.
So these were male dominated institutions unlikely to hire any women musicians.
And the change that really created diversity in orchestras was that they then had people who applied for these positions perform behind the screen.
So you couldn't make out the gender of the person and think about for a moment what that means.
That means the way humans make decisions, our best chance is to throw away information.
Because the moment I see that you're a woman or that you're a man, I'm just incapable of just listening to the quality of your sound.
But the relevant analogy here, Felix, is that it would be better to have a computer listen to a violinist than having someone who's a human being because the computer would be able to better parameterize and weigh what it meant to be a good violinist.
Are you willing to go to all that way?
Because my concern is that what we define as music and what we define as high quality performance, especially at really high levels of performance, is going to get flattened.
You don't worry about that in the music analysis.
I don't really worry about that because in all of the processes that we have today, this is just a first step.
But where this is going, Felix, is you are going to pick the violinist based on what the algorithm is.
Once you have hard data coming from algorithms that's generating the pool of people in, that information and that mode of analysis will come to dominate.
You are jumping so far to the other extreme.
The status quo is not working.
Actually, let me amend that.
The status quo is working for some people.
And for some people, it's really not working at all.
There are underrepresented segments of the population in industry after industry as a result of entrenched calcified forms of bias embedded in our HR systems.
And so this is a mechanism to shake it up a little bit.
I don't want to defend the status quo and I don't want to defend discriminatory practices that have been embedded through soft judgments for decades.
And I think this is a powerful tool to remedy a lot of injustice.
But I have a hankering feeling that there's a dark side here that is not fully clear, both because I think biases get embedded in algorithms.
And I think we really perhaps miss value human interactions and human judgments at final stages.
Do I believe the funnel should be wider?
You bet. But the rubber hits the road with these harder things and I'm not signing up.
But look, I'm going to be replaced next week with a robot on this podcast.
So it's been great knowing you guys.
Yeah, exactly. It's been great knowing you.
And thanks for playing along with an imperfect human over the last 50 episodes.
Okay, pics. You guys got pics for us?
I got a pic. So there is a fantastic new history by Anne Harrington of psychiatry.
I'm always fascinated by psychiatry and kind of how we switched very quickly from like a Freudian approach in the late 70s and 60s to like a very medicalized approach and a very biological approach in beginning in the 80s.
And she tells the history of how that happened.
And it is fascinating because I'm always interested to just think about that world, which is a world where some people have enormous problems and are helped incredibly by these drugs.
And then I have a simultaneous sense that these are massively overprescribed.
And there's a whole bunch of mood stabilizers that are somehow again to relate to another part of our conversation kind of flattening us all out.
I can't recommend it up.
It's a fantastic history of science, specifically of psychiatry over the last three decades and how it became biological and how it has failed.
So for example, after these three decades, we have no biological markers of some of these underlying problems, even though we've been prescribing biological agents.
And it is so fascinating.
And she writes it in a really even tempered way, but she's got a point of view.
So Anne Harrington, it's called the Mind Fixers.
And it is just a spectacular history.
So fascinating. Okay, interesting.
I have a recommendation for a Netflix series and the show is dark.
The first season was aired a few years ago.
And now on June 21, actually, it's time for the second season.
It plays in a small German town.
And if you love sci fi, it's just, you know, too good to be true.
The story is around kids disappearing from the town.
And then there's sort of this revelation that there's a wormhole just right next to the nuclear power plant station.
And you have time travel where these characters go back and forth in time in ways.
I always find it, you know, how every show that has time travel runs into some inconsistency at some point in time.
But here it's done in a really clever, smart way that actually has you think about the nature of time itself.
So super fun. The date, it's going to be June 21, which is relevant for the show itself, because that's an important date in the show.
And so season two on Netflix, dark June 21.
That sounds really good.
I got to say this whole episode has been a little bit dark.
We got kids disappearing.
So my recommendation is let us do something light hearted.
Good young. I'm going to recommend an app that I've actually used for a long time, and then I sort of stopped using it.
And then I found myself falling back in love with it again.
And the app is called Trello.
Oh, yes. I showed it to you one time.
I think so. It's just the most fabulous list keeping app organization app for organizing small things.
And so if you're the kind of person who finds yourself writing little notes to yourself on yellow sticky pads and putting them everywhere, this is the app for you, because it enables you to organize little snippets of information in lots of different ways.
And then it's super easy to move those bits and pieces around in an interface that is so simple, so clean and really fun to use.
So the app is called Trello.
And that's my recommendation.
Sounds great. And a nice note to end on.
Okay, so those are our picks.
Thanks everyone for listening.
This is After Hours from the HBR Podcast Network.
Thanks for watching.
