*APAC Heritage Month* This is your Tech Leader Speaking
Today, we're discussing data ethics with Laura Summers, Founder, debias.ai, and creator of Ethics Litmus Tests .
Hi there, I'm Gretchen. Thanks for joining us for today's session of This is your Technologist Speaking.
This is a series of talks with our local APAC industry leaders.
We've been super lucky and I've curated a guest list filled with technologists that have both knowledge and leadership experience.
Our guest today is both a holder and a sharer of great knowledge, as well as a proven leader.
Her skills range from UX design to ethical operations and pretty much everything in between.
She isn't afraid to forge new paths and has been passionately working to change a landscape of digital bias.
Somehow along the way she found some time to create ethics litmus tests, which are the most practical way I've seen to, I'm going to use her little tagline on it because I love it, get that little voice out of your head and start having some tough conversations around ethical considerations in your workplace.
So welcome Laura, it's so great to have you with us, especially at this early time in the morning.
Thanks Gretchen, I was just saying I'm not much of a morning person, I'm more of a night owl, but it's always a pleasure to get up to a conversation with you, so no worries at all.
Well I was thinking about you as I wrote it, that's one of the things Laura does and has done and you're such a leader around ethical bias and thoughts and I know everyone in the tech industry here when they're stuck on these problems, the conversation always goes, call Laura, she'll know, and so I think it's really good to call that out and let it be known because maybe there's some people that don't know it's you they should be calling at this point.
Anyway, tell me about ethics litmus tests. Well they're basically a little card pack I put together, I'll try and show it and see if the image thing can, you know it's hard to see but you kind of get the idea, they look a little bit like litmus tests from chemistry and I basically built them for myself because I'd been looking around, this is January just before COVID hit, and I was looking around for tools to sort of support my work in the space of thinking about the potential for bias, the potential for, you know, unintended discrimination or unintended harms coming from either fixed deterministic systems or machine learning systems that have, you know, an element of change and unpredictability to them because they keep evolving and I just couldn't find something I liked and there was a lot of stuff floating around that was very academic and very dry and like very long and it felt really sort of painful and hard to pick up and also because I have a design background, as you mentioned, I love beautiful things and I wanted something that was beautiful to touch and look at and, you know, would try and like soften the pain of having to ask yourself or other people like potentially awkward or uncomfortable questions so.
I mean to that point I think on my black, oh you still can't quite see, these are beautiful cards, they actually are really pretty.
I had a lot of fun designing them, I did a lot of like watercolor painting, that's what's on the label, it's like those little splotches and I had a lot of fun thinking about like, I mean this is such a nerdy thing to say so please like don't judge me in the morning, but I was thinking about things of like balance and like data and stochastic, the stochastic universe, like the world where we have like, you know, chaos and entropy and like sort of anti-entropy is like, you know, programming and design and balance and geometry, so I was thinking about like that sort of tension between things falling apart and things coming together and trying to find the balance between them and then like all of the watercolors are about like, you know, litmus tests and that like beautiful bleed that you see when you dip your litmus test and you get those like colors and it's green or yellow or blue depending on how acidic your test is, so yeah, I'm a total design nerd and I had a lot of fun thinking about design references for these packs.
It makes it, because what you're actually asking people to do is get really uncomfortable and have a little beauty along the way, kind of smooths the path, right?
Yeah, well like there's definitely a thing which is like it's, if it's attractive and easy to use and not too confronting, that's that certainly like helps you get over that barrier and I wanted something that could sit on your desk and just like, you just like pull it out and be like, okay, I'm just gonna pull one out and say, oh, I got a blank one, perfect.
Let me, let me pull it.
I did put a lot of blank ones in because I felt like, you know, the provocations I wrote came from like kind of questions that already floated in my head a lot, but I don't want to pose myself as like a moral philosopher or a major expert in terms of like what is the best way to elicit these moral intuitions or the best way to elicit these conversations, and I'm sure everyone has good questions, so I wanted to make sure everyone had space to put their own.
Ah, okay, so I pull a card and it says, does this represent me well?
Oh, that's pretty good.
I can't read that at all. It's not blind.
It has a way to go, but yeah, like the idea is that these cards in this like activity can range from like exactly that much, like you can have a little, you know, tickle in your stomach that says, I don't really feel great about this moment or something feels wrong, and you can just like ask yourself a question, just get out of your head, you know, so sometimes the job to do is just to like, you know, reframe it to ask yourself a question that helps you like, you know, firm up whatever it is that's going on, so it might be like, oh yeah, that's not okay because, you know, like does this represent me well?
Like I wouldn't want someone to know I worked on that.
That's not okay. Why is that not okay? Okay, maybe it's because, you know, if this was seen on like Platformer by Casey Newton and he wrote about it, it would sound really bad in public, and why would it sound bad in public?
And you kind of like take yourself down that journey of, you know, what's going on, and why are you feeling uncomfortable about it?
And those sorts of questions, like, you know, we don't really necessarily make space for them in classic product like rituals or agile rituals or like the sort of day-to -day conversations.
We tend to focus really hard on things like feature priorities or, you know, thinking about risk more from like vectors of attack or security point of view, but we don't necessarily talk about like the ethics risks or the risks to our customers.
And in an increasingly like complicated world with lots of systems going on, another big thing that happens is there's not just our systems that we're building, but the interactions with all the other technological systems, and that really like explodes the complexity, right?
It does, and I like that these take it from, like you started off there with, oh, I've got a niggly feeling, and it's really hard to qualify or get a niggly feeling past any kind of commercial, I guess, structure and system.
So when the cards let you set it up and think it through as an actual, you know what, this does have risk because potentially if it was on the front page of a newspaper, that would do business harm or harm to our brand.
So it gives you a way to justify, well, not even justify, but to work through your feelings in a less emotive way and structure them and have the conversation, right?
Yeah, exactly. And it kind of helps you go, well, I couldn't just take this niggly feeling to my friend sitting to my right and say, hey, you know, I have this weird niggly feeling.
But if you say this maybe doesn't pass the 60 minutes test or, you know, like, I think that there's something going on here that like could cause brand harm and you start talking in the language of business cases and, you know, the things that help you like determine whether or not there's like a risk that needs attention or time or product thinking.
Like it, yeah, it's, so yes, as you say, like it's hard to like just talk to your gut feelings and, you know, depending on how emotionally intelligent you are, that may feel more or less comfortable anyway.
That's a really good point.
Sorry, not to throw shade, but you know, like, we're not, not all of us have a lot of fun talking about emotion words anyway.
So yeah, like the idea that, that like, it's a skill and we can develop it and we can continue to like practice it by like practicing these, like, you know, these little questions and asking themselves to ourselves.
And, you know, like it's, it's not about being right or wrong.
And it's not about saying like there's a correct outcome. It's about saying this is a skill to develop and, you know, we want to try and like make it as small and approachable as possible.
Have you, have you heard of Atomic Skills by, what's his name?
James Clear, maybe? James something? Atomic, Atomic Habits. Sorry. It's called Atomic Habits.
Yes, but I haven't read it. It's on my, you know, the anxiety causing list of books you should read.
Oh, I'm sorry. Atomic Habits is one of the books that like was in my mind when I was making these.
And he talks a lot about either piggybacking new habits on existing habits that have established like themselves in your routine.
So like, you know, if you're already brushing your teeth every day and you really want to be flossing your teeth, you should just do it at the same time that you brush your teeth rather than stopping your day going back to the bathroom.
And like, that's an obvious example, but like a lot of the things he talks about are like how to break down habits to the atomic level, the smallest level, and to try and adopt them at that smallest level so that you don't make it this like huge, big behavior change that obviously, you know, is kind of annoying and feels bad.
Right. Like big change is hard. You kind of go and look for all the reasons why not.
Exactly. So I think that like this also has that kind of thinking behind it that like, you know, you don't have to do much and you can just ask yourself a few questions.
And I did design some activities that you can do like for yourself or in groups.
And I was using very much like kind of retros and, you know, like a lot of agile methodology, like kind of thinking in it.
So it's hopefully not activities that should feel too different from what people already do on their day to day basis.
But with this kind of framing. So it's a way of helping us like continue the conversation if we do decide it's a bigger thing and it needs to have some attention.
How can you, you know, elicit these moral intuitions or these concerns for a larger group?
And then how can you come to some kind of, you know, next steps, whether that is research or product, you know, thinking or exploration or, you know, like circling back to it in three months and seeing how it is, you know, like some of that is like, let's do nothing, but let's continue to listen out for this and like see what has, you know, changed for us in terms of our thinking after we've developed a bit further, after we've heard a bit more from our customers.
Yeah, academic stuff earlier, you said, you know, there was, there is so much work and research in this space.
Like it's, it really is a big, you start looking into it and you will just never, ever, ever stop reading the research.
But I've had the same experience that you alluded to that was, it's very academic and not very, I don't know, it's quite sterile and not applicable.
You can't take the research and go, hey, it worked today, let's do this, because that would be a really useful way to integrate these massive concerns about ethics that I've got.
It kind of just has this gaping chasm between the two. Yeah, it's like theory and then there's practice.
Was it that kind of massive chasm that led you to develop the cards or was it, did you truly just go, hey, I need something for me and here we go?
Yeah, well, honestly, it's funny you ask that because I was, I was trying to think about what made me write, what sort of was the trigger that got me to make these.
And I think it was, I had been asking myself this question in my head a lot, which was what would, what would happen if this happened to my mom?
So I, like my mom is kind of like an interesting user in my head because she's like very enthused about technology and, you know, she's very supportive of me.
She loves that I'm a techie and she like, likes to know what I'm doing, but also gets annoyed when I use words she doesn't like understand or have.
So, you know, like, but she's, she's enthusiastic and she wants to be involved, but also like, she's maybe what you might think of as like a little bit naive or a little bit vulnerable.
So like, she doesn't always like understand the difference, say between a pop-up, which comes from the actual native operating system and a pop -up, which is some scammer website trying to like, look like the native operating system.
So I've been asking myself that a lot. And, you know, like to the point where it's like, oh, how do we put these things into practice?
I was like, oh, well, actually these little questions I asked myself are quite useful.
And they're often the way that I start framing a conversation when I'm guiding or facilitating a group of, you know, product people or devs, like that's often how I would start the conversation anyway.
And it was like, maybe this is just a thing. Maybe I should just try and like come up with an approach.
And then the other thing, which this is going to sound obvious once I say it, I was watching The Good Place at the time.
Oh, what a great TV series, right? I'm sorry to be that person, like to be such a cliche, but I was incest and I loved it so much.
And yeah, so I was watching The Good Place and early on in the show, they have this thing where they're like, oh, we have this quiz and it's like an ethics litmus test.
And I was like, oh, that's a great idea.
So I totally stole it from The Good Place. But like, there's just this dumb quiz that was stuff like, you know, have you ever had like bad view license plates?
And, you know, yeah, I mean, it was like, it's obviously for comedy.
And I was like, well, that's not really useful for me in terms of the discipline I'm trying to build.
But I thought it was a really evocative and like, obviously, these these cards are trying to like reference the actual chemistry, the litmus tests.
And I think like, I mean, obviously, not to try and like over philosophize this, but I did, you know, I'm a sort of very atheist diagnostic, like don't really not really into religion person.
And I'm not, I'm not saying that religion, you know, is or is not valuable in the space.
But I wanted to provide a tool that kind of reference science and scientific methodology as opposed to religious like doctrines or iconography, because obviously, that's so loaded.
And that's going to be really welcoming to some people and really unwelcoming to other people.
So I thought, which is not the point, right? You want everyone to understand, we don't we don't want a tool which has this whole like, you know, loaded gatekeeping thing, which is like immediately out of the bat, half of the people are like, no, that's not for me.
So here we are with these beautiful, beautiful cards.
And I still love that they come in a little test tube. I mean, when I got mine, I was super excited.
I use them all the time. Can I just say I got the cutest question from one of someone who found me on Twitter, they were like, um, when you get the card pack, do you get any like, solutions for them?
And I'm like, they're not actually litmus tests.
I'm sorry. I thought that was like the cutest thing I'd ever seen.
It was, I felt quite bad that I didn't actually design some kind of cool, like solution situation where you could like dip it and it would turn red if it was bad.
Like, I thought that I had a Barbie that like, you know, you comb their hair and did that.
Is Barbie good? You could dunk it. And there is no, they're never good.
Barbie is not good. That doesn't even just sit in my head a little.
Um, so is it anyway, we've briefly talked about how they work and you've said you've got some kind of case studies.
Can we do a dummy, dummy run a little demo?
Yeah. So, so, um, with the activities, the way that you, I sort of designed them to be like anything from, you know, quite a small, um, group of people.
Oh, look, you can tell this is a prototype because it is a piece of paper. It was like, where's my examples?
That's hilarious. Luckily for everybody, I have multiple versions of this on my desk.
That, that was literally like a prototype version.
I was like, how, how small do I have to fold my instruction sheet down to make it work?
Um, uh, here we go. Here are the activities. Um, yeah, so, so I have a number of activities and they, they range from like pretty sort of small stuff to like more like a group activity.
But what we could do is maybe the two heads, which is, um, an activity that's basically just trying to like, let us like elicit our own responses to a scenario and then compare them with a friend.
So like phone a friend it's like, oh, you know, I'm still not sure whether this is a problem or I'm still not sure how big a problem it is.
So, you know, let's just do this quickly together for five minutes and then see, like, if we think we need to take it back to the broader team or towards direct manager or to someone else.
Um, so that's like, you know, it's like a little bit bigger than a single, a single litmus test.
So the way I ask people to do this is to start by like doing a little bit of problem definition.
So like often, often you sort of like have a gut feeling and you're like, okay, what is this in response to?
And it's, you know, it might be like the product, or it might be like a new feature, or it might be to like a press release or to a piece of marketing.
Um, or, you know, like, so, so part of it is about kind of like trying to like zero in as much, um, to what exactly is a thing that's like causing that uncomfortable gut feeling.
Um, because, you know, as we, we talk about with problem definitions all the time, like defining it well is half of the battle, right?
So, um, so we do have some scenarios that we have defined ahead of time.
You know, like here's the cake, the cake is a problem.
Yeah. So I've been, I've been like gathering a bunch of scenarios that I think of as like motivating scenarios to try and practice doing this stuff.
Um, and I also, another thing you mentioned with like, why do it with toy scenarios first is that it is good to practice doing this stuff in no stakes environments or low stakes.
It's like the higher the stakes, the harder it is, right?
Like the more, the more it actually matters, or it might impact someone, you know, like the company's income, or it might be like actually a problem.
The harder it is to be the person to call it out or to like insist that you take time for it.
Um, so I think that there's a real value in practicing in like low stakes environments.
So you can just like, you know, get a feel for doing this and building this vocabulary for yourself and like thinking it through logically, but knowing like that there's nothing for you to do at the end.
As you're saying that I'm thinking you'll need for someone to run these really well in a high stakes environment, they're going to need to have enough confidence and ability to hold space for some really awkward conversations potentially.
Yeah, I could be hard. We'll just need practice. Does it, is there ways people can learn how to do that more that you know of?
Um, look, I'm actually, I'm learning and growing into the space right now.
Like the question of how we can hold space for different opinions and like respect others' voices, but then also need to make decisions under a certain time frame is like, I think kind of like the big existential question of our time.
Right. And it's extra hard because, you know, as we know, like everything is fraught in high stakes and like social media is just like, you know, like two teams barricading for other sides, like going each other all the time.
Incredibly tribal. Oh yeah. It's literally like, I think of it like two football teams, just like yelling at each other.
And like occasionally there's like a punch up at the end. That's kind of what social media feels like to me half the time.
So I guess it's work in progress. You have to practice.
Yeah. I mean, look, this is, this is my attempt to offer an approach into doing this.
And I think, you know, a good starting point is just saying, Hey, we're going to do a litmus test activity, and we're going to assume that we have differences of opinion.
And we're going to start from saying, let's make those clear.
Let's elicit them. Let's make them explicit and not leave them implicit in here.
And let's be okay with that. So like just practicing going from like their inside, and we haven't talked about them to their outside and they're up on a wall on sticky notes is like a huge win.
Right. Cause that's, that's the thing we mostly don't do.
And I think it would help DNI outcomes as well, because you're really talking about being inclusive at this point of all people.
So that's, I think we're winning by not just building better products and features that aren't causing harm potentially, but also you're bringing the whole team into the game, which can only help.
Yeah. Look, I will say that I've been thinking that the activities I've designed are very kind of like design activity focused.
And like, I love post-its and I like talking in rooms with a group of six people, but that's not for everyone.
So I'm aware that there might need to be some adjustments to these kinds of approaches for people who feel shy or like English is not their first language, or they're not used to being like asked to, you know, really speak up and participate in this way.
So I think that there's certainly cultures and personality types that this might feel confronting for.
And I I'll acknowledge that while saying I don't have a perfect solution for it yet, but let's do it quickly.
I'll take us through a scenario. Please. The scenario I have here is price versus value.
And I'll just, I'll read it out. It's sorry, it's a bit of a school teacher thing to do, but this is an example that I think will probably everyone will like listen to it and say, oh yeah, we know who this is about, but it's the thing that happens a lot across a lot of different industries.
But you know, toy scenario, as part of a large multinational online retailer, your work focuses on optimizing sales across regions, cultures, and users.
Your product manager wants to explore opportunities to adjust prices based on browser device and location.
And this is like an attempt to make products whose prices have sufficient margin, more affordable to lower income people or people in like, you know, lower income areas.
So people who might have lower salaries or might not be able to like buy things at the same price as Western Western countries.
So they're looking to overall increase their sales when there's sufficient margin on the products to allow for like this adjustment in the pricing.
They're also thinking at the other end of increasing prices for users that they identify as well off, but who have a high appetite for the product.
So that's kind of a classic, you know, what the market will bear pricing strategy, right?
Like, it's like, if you want to buy that thing for 10 bucks, but they charge you 12 bucks, you're still happy to pay for it, then that's a good outcome for the seller, at least maybe not the buyer.
But oh, yeah, it depends on what value they perceive from the product.
And it depends on like how price sensitive they are.
So how much disposable income they have, and how much that extra two bucks might hurt.
And so they're proposing their product prices could vary up to 20%, based on whether a potential customer was identified as higher or lower income.
So obviously, like this is, this is going to be, you know, still within like a pretty fixed range of countries, you wouldn't assume that someone who's making like the equivalent of like one USD a day is going to be like included in this in this market.
So that's, that's an example of like a thing that, you know, companies do often try and do.
And it, you know, like, it's, I think it's, it feels a little funny, but maybe we're not sure if it's okay or not.
So it's kind of like, we'll say that we had that gut reaction of like, I'm not sure.
Do I like this? Is it okay? So what we're going to do now is I'm going to pull a litmus test.
That was our that was our motivator. And we're, we're just gonna like pull a litmus test.
And then you and I are just going to have a little think, work out what our response is based on the litmus test, and then we'll just discuss it.
Oh, this is a good one, too. If I don't discuss it further, will it play on my mind?
So this is a litmus test that I think is a good one, because it helps us like work out that question of like, okay, well, say that this is a done deal moved on.
No one's talked it through.
No one's thought it through. Is this something that's gonna like, you know, keep me up at night?
Or is it going to play on my mind? Am I going to think it through, you know, in another few weeks or something?
Am I going to feel like I should have spoken up or said something about this?
I'm going to invite you to go first, Gretchen.
Do you have any responses? I had a lot of responses all at once, and they're really hard to then pull apart, right?
Fair. Yeah. I think it would play on my mind, I'd be questioning, as some of it depends on the product, right?
So if we were talking about, at the moment selling oxygen tanks, then that's a completely different thing in my mind than if, for example, we were selling luxury vehicles, right?
Or, or like cars or something, something, yes, really classic commodity products.
So I think it is really dependent on that. But then also I go, oh, I did economics 101.
And they just said, if the demands there, you know, charge the rate, the market will pay.
That makes sense for everyone. Everyone's winning.
So I think if I was in an organization, and this was the case, I'd really like to talk it through.
And the justifications of probably, if we do charge way more for people, at what point does that create harm for them?
So I'm trying to think of a case where it would be annoying, like imagine your taxi, all of a sudden was $150.
And I needed, I fell into that market of, yeah, I could afford it.
And I'm not that price sensitive. And I could do it. But actually, I've got these other mitigating factors of why I desperately, desperately need to get home.
And I don't, it feels callous to be charging that extra, extra money.
The idea that, you know, you might have no bounds on how much more you would charge based on demand.
And like, looking at markets is like kind of surging and falling and thinking, is it okay to, like, continue to increase the price without any acknowledgement of the circumstances?
And like, I mean, as everyone knows, we live in a volatile universe.
And, you know, we've had, we've had instances of, you know, attacks and, you know, like terrorist attacks, and like, natural disasters, where people are trying to get somewhere get away from somewhere else.
And, you know, those are instances where you think, well, something like commodity, like getting in a car and driving, maybe that's, maybe that needs to take into account the circumstance.
I'm certainly, I would, I would need to talk this through with someone.
And I'm, I think, like the thing you mentioned, the beginning kind of is exactly where my head is at is like, if you're a company selling a ton of products, and you're thinking about applying this strategy, like across a whole bunch of them, I'm really worried.
If you have specific product types. And you have, like, you know, thought through whether or not like, you know, these adjustments to price are going to, like, negatively impact those customers.
Financially, if it's gonna put them under financial stress, is this like a necessity?
Is it something that it's nice to have? You know, like, how, when you talk about, like, you know, not price sensitive, is it that they have to have it, and they will pay for it?
Or is it that they can't afford it? Because there are different questions.
They are two different questions. Hey, I've done a standard issue me and I really wanted to touch on one other thing really quickly.
I recently saw an Australian company that didn't do any litmus tests, I suggest, and they were asking job candidates for after their ethnicity, they would then ask their skin tone color.
Is there any legitimacy in these questions? And if there is, how can it be done?
Well, and we only actually have about a minute and a half. I'll try and smash out the answer fast then.
Yeah, it's it certainly played across my social feed.
And I was very aware of this incident. Because if you're not familiar with the term, this this has this uncomfortable callback to colorist and racist practice called paper bagging, or paper bag test, which is where you ask a person of color to stand next to a paper bag and to determine how light or dark tone their skin is.
I think there there is a legitimate need to know these things, if what you are trying to do is improve the diversity of faces and a training data set, which is I believe, I mean, you know, I'm reading between the lines, but I believe what the agenda was for this company.
But I think that you have to do this with a consent driven approach.
So first ask someone if they would consent to participate in a data set for that reason.
And then if they do consent, maybe find some less invasive ways to work out.
You know, I'm gonna get you back on to talk about this more another time, because we have to go now.
But thank you so much for joining us.
Sorry. It's okay. Thanks, Gretchen. That was a lot of fun.