Diversity and inclusion in data with Debbie Botha

Loris Marini: I'm pretty excited to introduce today the first of a series of Discovering Data Leaders.  We're going to dive briefly into why we're doing this before jumping into our conversation with our first leader of the series, Debbie Botha.

The mission and vision of Discovering Data we've gone through that a couple of times. It is about empowering the next generation of data leaders and the present generation of data leaders for that matter, by fixing the gaps that we see between business and technical. Getting that domain knowledge to flow and getting people slightly more interested, more curious, and hopefully use that momentum to solve problems more effectively. Not just build data products, but build things, solve real problems and impact people in a positive and the biggest way possible.

In a way, it is about gaining range, as opposed to getting specialized. There are many, many schools that do one thing really well. From data science to machine learning to data engineering, there are many creators as well lately coming up in the space. There are a plethora of options for those that want to dive really deep into one aspect, but I found that it's a little bit harder to find things that broaden the horizon, the spectrum of the different skills that you need when you're trying to do data in the real world, whatever that means.

That's what we're trying to do here. We're trying to increase the range. Discovering Data Leaders is one of the ways we do that. It's a series of short conversations with data leaders from all over the world, really. Instead of being the usual episode as the Ideas series where we take a topic and we dive in for an hour or more here, the intention is to have brief conversations that don't focus on how to do stuff but focus on the person behind the leader.

Can we have 15 minutes with them and imagine that we are a tiny little flies that are invisible and we sit on their backpack and see them around? Ideally, that's the intention. That's what I see when I think about the person behind the leader, but obviously, we can’t be flies and we can't hide.

We're having conversations with these leaders and trying to probe into their lives, into their motivations, and see what is data to them. What are the biggest problems they see?  it is a discovery process like all others, except the subject is the person behind the leader.

The goal here, if I was to explain it in just a sentence, would be to inspire and inform, and connect people that are interested in data through a human lens. They want to know more about the person more than just the title, more than the avatar you see on LinkedIn when you scroll through your feed and reach a more diverse audience. Like all things we do here, the primary goal is for me to learn and I'm so glad that you're here with me hanging around and following me in this discovery journey.

The first one in the series is Women in AI. Women in AI is a not-for-profit organization that is active across 140 countries.  It's about women and minorities, representing those that get less exposure and we hear less of. Again, the goal is for them to tell stories, inspire and influence data leaders and get more people to become a little bit more curious. Find mentors, find their tribe, and expand their understanding of what data is and how it impacts people's lives.

When I landed on the Women in AI website for the first time, I realized I had a sense that we were really aligned and then I met Debbie Botha, chief partnership officer at Women in AI. She’s spent more than eight years at IBM, working as an information architect and as a consultant for various other organizations. When I met Debbie that's when it clicked. I was like, “We need to do something together here because our missions align so much and diversity speaks to my heart.”

Debbie Botha, welcome to the podcast.

Debbie Botha: Thank you so much, Loris. I'm very happy to be here.

Loris Marini: Do you want to kick it off? How did you hear about Discovering Data in the first place?

Debbie Botha: Okay. I've been very busy on LinkedIn as an observer in 2021. I discovered your podcast in the beginning of the year and I kept on going back to your podcasts. Even if it's an hour, I was so willing to put it aside to listen to your podcast, because it's always interesting. It's really getting the information from so far and so innovative the way that you describe things.

I really enjoyed the human side of your podcasts and you really had very interesting people on your show. I thought it would be brilliant for us to come together and bring that interesting conversation to the table with Women in AI.

Loris Marini: Well, yeah. If you can see where my heart is smiling at the moment. The best thing that a podcast host can hear. The best feedback is from someone that really enjoys the content. Sometimes it feels like you've put ideas out there in a vacuum. Thank you for that. Really appreciate that feedback.

Debbie Botha: Yeah. Also, it is clear the amount of work that you've put into the podcasts. I mean that the content is so dense and the ideas are so vast. It's really a pleasure to listen to it.

Loris Marini: I'm so glad that it transpires miles and miles of physical distance. Let me deflate my ego a little bit now. I want to talk about Women in AI a bit more deeply. What is the vision of Women in AI? What's the mission? What are you guys trying to do?

Debbie Botha: Yeah. Women in, AI, as you said, is a nonprofit organization. We operate in about 140 countries. We have many, many volunteers working for us. There are ambassadors in the countries. The number of volunteers I think is about 180 by now with the ambassadors in the countries with their leadership teams. We have these chiefs identified and appointed late last year to really help the organization scale in a much more mature way.

We started in 2017 and we grew massively in a very short time. We now have 10,000 members with a very engaging Slack channel, newsletters, blogs, the web, and so on. 27,000 LinkedIn followers, we are scaling. We are growing and we want to scale in a much more mature way through partnerships.

What we aim to do is to increase the numbers, the presence, and the imminence of Women in AI, from where they are educated in school, right through to getting their first job, right through to helping them grow in their careers and pave the way for these women through various partnerships in academia, government startups, and so on. We launch these programs for innovation, awards, education, talks. We love doing panels. That's in a nutshell, what we do.

Loris Marini: Mentorship, in a way, is a big chunk of what you do. It's not just about diversity, it's about mentoring those people that are, in a way, left out of the conversation.

I see side effects of limited sample size or skewed sample size in data science. I do have memories of models that underperformed because the data sets we put in were not representative of the population of the real world. This is a very nonproblem in data science.

A court of law using software as an aide to decide whether to release someone or not. A doctor in a medical environment using computer vision and computer recognition to maybe detect the early stages of a disease. Those applications are serious. It's a matter of life and death or it's a matter of massively changing the quality of life of someone in the case of a court of law.

Those are examples that require the right ingredients at the input of the model. The right ingredients mean obviously the right data. How do you get the right data? It's a matter of training people to recognize biases.

Many people might say, “Well, what does gender have to do with the diversity of a data set?” One could spend 10 hours on this topic, but it's the topic of metacognition, the topic of cognitive biases. We all see the world through the lens that we developed over the years as we grow up. Different people from different backgrounds, not just gender, see the world in a different way. That lens creates a bias in our data sets and so it's important to have diversity.

From just a purely algorithmic standpoint, but it's also important to have diversity because in the real world, we don't have clones of ourselves. We interact with so many different people. The more we are exposed to different types of personalities and ways of thinking, the better equipped we are to actually solve problems in the real world.

Debbie Botha: Yeah. If I can go back to your point on bias in the data sets and why diversity and inclusion is so important is that you actually have to include women and minority groups in the start of an initiative. Not in the data itself or as an afterthought, but really around the table, bring those people at the table to help make decisions on what you are going to do, what type of algorithm you need. Your strategy must be designed by a diverse group of people. That's where you you'll get the major results.

In terms of mentorship, I remember most of the real pivots and bold moves in my career were based on my mentor. Female mentors in particular. I remember one of my favorite mentor in IBM Mandy Chessell. She was a distinguished engineer and she really took me under her wing and introduced me to a worldwide position. Number one, which was absolutely brilliant, she introduced me to the IBM Academy of Technology, which was a huge privilege to be a member there. Also introduced me to the worldwide community of information architects, where I became a leader in a team of about 800 architects in IBM.

That really made me realize the importance and the benefits of a community. That was that mentor. That's a positive side, but on the negative side, I have this desire to be loved by everybody and recognized for all the hard work that I do. I know it's silly, but that's inherent in my personality and I have worked hard over the years to not let it drive me. A couple of years ago, I found a huge wake-up call that there are women that do not love me that much. Particular women really. I had to struggle to not feel bullied or offended or put down in some way.  She was an executive that’s supposed to take care of women.

It was when I got cancer last year, where I thought, “What am I doing to myself here? Why am I allowing this to get to me?” I worked so hard, I worked hard to please everybody and to get recognized. I allowed this negativity to get to me. After the cancer, I decided, “Let's regroup, recover, become much calmer and just don't allow other negativity to get to you and rather learn from it and grow there”.

There's no such thing as a failure, it's just keep learning from it and grow.

Loris Marini: Absolutely. I think a lot of people in data and in entrepreneurship, in any field, need to hear a lot more of what you just said because it's so true.

It's easy to spread yourself thin. It's easy to get stuck with the wrong metrics. It's true in business, but it's true in our careers as well.  Getting that positive feedback and recognition, feeling that people appreciate the work you do. It's addictive. It's a form of dopamine cycle, right? Similar to sugar or other addictive substances.

Debbie Botha: Yeah. Before the cancer, I was always cognizant of what I say in public about vulnerabilities. I wanted to look perfect. Sound perfect. Be perfect. Nothing is wrong with me. Nothing really bad happened to me or everything is perfect. After the cancer, that's one of the things that shifted my mindset: it's okay to talk about the negatives and the positives because it also helps others grow.

I attended a conference late last year where the women were talking about these alpha females. I can't remember what they called them, but it was exactly this way. It was a whole hour on how to recover from experiences like that. It is a thing of women crashing and not uplifting other women.

I think that the positive to take out of here is I joined Women in AI as their chief partnership officer late last year and the wisdom, the compassion, the absolute privilege to be part of that group of women were phenomenal. I got my home. I feel appreciated. Not that I crave it that much anymore, but I really felt that that group of excellent women that are really doing so well in their careers, they’re for me, and for each other, and for themselves.

Loris Marini: You found your tribe?

Debbie Botha: I found my tribe and I want to do that for other women also.

Loris Marini: It's so powerful, right? Especially the last couple of years, we have all been forced to isolate from one another, but that really goes against to the biology of our brain. We didn't evolve in isolation, we evolved because we were able to formalize, to create these partnerships with other people in what we used to call tribes.

This word is kind of is an ancient word, but it's coming back more and more. We hear about communities online and in person and finding your tribe, but in the end, it is about giving meaning and purpose to what we do and also share the pain. There's a saying in Italian, "Mal comune mezzo gaudio” which translates literally to, “if you have a certain amount of weight and you share it with someone, it's half the weight.” Even if the person is not physically helping you support the weight or move it around, just the presence of someone else can make the effort a lot easier. Definitely it's a huge point of it.

I wonder like 10 years, nine years, eight years at IBM. Lots of time in information architecture. How do you get into being an information architect? What was your entry point to the world of data?

Debbie Botha: I love the day-ta and dah-ta.

Loris Marini:  I'm trying to mix it up. The audience is global.

Debbie Botha: I actually started in the early 90s by building one of the first data warehouses in South Africa at a short-term insurance company. That's where I started working with SaaS software. I was completely hooked. That was where I saw that you can't just go and throw a bunch of data sets into a pool and think you're going to get something meaningful out of it. You really have to follow best practices or principles that makes sense and have been tried and tested.

When I started consulting in 1998, that's when the Kimball university books came out and man, those books are brilliant. I started applying those techniques from 1998 up to now. I still refer to those books. These concepts coming out like the data mesh, talking about domain-driven insights or domain-driven data products and things like that. These are things that has been written in those books in 1998.  Distributed teams, all those things. It's nothing new, but I really love the spin that Jean-Marc is putting on the data mesh because it's now not just the data warehouse, but the data lake and everything is a mess.

Nobody is really following the books or know about the books. Fortunately, there are those that have read the books and, and can come to the party to weigh in.

Loris Marini: Yeah. Reduce the noise because it definitely feels like a high noise environment from the perspective of the practitioner that doesn't have those 20 years of experience.

Debbie Botha: Yeah. Actually 30.

Loris Marini: You jump around from one Medium article to another one that you’re like, “Who is right here? Can we establish a ground truth? What are you talking about ground truth, you're the expert? You should know.” It's a cycle.

Debbie Botha: These old dogs, they call themselves old dogs, like Bill Schmarzo. Yeah, he actually wrote an article about the comparison about the conformed dimensions of the Kimball methodology, where we compared that to the data mesh. It was really funny.

Loris Marini: I need to read that. I missed it.

Debbie Botha: It was brilliant. It was so good. Samir Sharma had a whole conversation with Bill and brilliant other people: Bruno Aziza from Google, John Thompson about the data mesh and about the Kimball conformed dimensions.

Loris Marini: I need to watch that.

You essentially started with a focus on building structures into data, using databases to store and organize data sets?

Debbie Botha: That's how I started. I actually started with data mining type of programs, development where we use SaaS procedures to call data mining functions, to predict fraud in the short-term insurance claims. That was as early as 1995.

I like to be quite broad in the things that I do. I did data mining. I did data modeling. I did data architecture, designing the whole big picture and then doing the data strategy, data and analytics and AI strategy. I helped organizations with what we call the chief data officer now.

There were always leadership teams that were formed to in all the organizations. I helped them form those data teams.

Loris Marini: You are the embodiment of someone that has range.

Debbie Botha: Yeah. When I mentor young people, many of them want to go into business analysis or architecture, even though they have a solid programming education and background.

I always tell them first, spend a good number of years getting to know as many, let's say, fields as possible. The data space is so vast that you can be a data engineer, you can be a data scientist, you can be an analytics engineer or AI engineer. There are so many things that you can do with the programming languages and stay technical for as long as possible, while growing your architecture and your business analysis and those type of skills and your project management skills. Do that but stay technical. That will give you this ability to stitch everything together when you do the more architectural and management type of roles.

Loris Marini: There you go. What concerns you about data today? What is the number one thing that we're not doing right and we should do as fast as possible?

Debbie Botha: If you look at the landscape of data, even data architecture, data tooling in all the various fields, we're all over the show. There are too many architectures. There's too many tooling. There's just too much. I think that’s why a lot of the newcomers come up with these new ideas that's ages old because it's just all over the show.

There is no common language in our field. There's lots of different opinions that's a lot of times overlapping and conflicting even. I think we should get to a common architectures, if you will, and common tooling.

I think a simple example is the data mesh versus the data fabric. There's a huge difference, but some people still say it's the same thing. Even the data fabric from its original meaning up to now, it also changed over time because of vendors’ perceptions and they wanted to infuse the technology discussion into it.

Another thing that I'm wondering about that I see happening is the whole thing of self-sovereign identity and privacy issues that we have and the regulation and everything that's going around that. If you ask any generation Z, which is born 1997, and whether they mind if Siri listens to every single word they say in order to come up with a very personalized location based on it, I have not come across one generation Z that minds that. They love it. They even come up with examples of how they were telling their friends that they feel like takeaways. All of a sudden there's a very personalized offer on their mobile phone about a takeaway place very close by. They don't mind.

That's our future generation that actually wants no privacy in exchange for extreme personalization and we old dogs have to consider that and maybe dig a common ground. There's always obviously the money laundering, the fraud, the serious things that can happen if we don't take privacy seriously, but we need to consider this.

Loris Marini: Yeah. Absolutely. Let's talk about the problem of the shortage of talent, how many students graduate and they’re just clearly not ready for the industry. They don't know what to expect. They have been trained to work with very clean data that's been already prepped for them and they can't wait to do the cool stuff.  By “cool stuff”, we mean the latest reinforcement learning algorithms, the latest image recognition software.

We both know that the largest majority of organizations don't need that really. They would be so much better off if someone was to walk in and standardize business terms with a business glossary and define what people mean with names like customers or products or order. That's not to say that all of their stuff is useless. It's just that I see a lot of focus on that and not enough on the management of data as an asset. What do we do with all these kids? How do we bring them up to speed?

Debbie Botha: Yeah, I think what we need to do is to really get into those schools and universities and teach them data literacy, data fluency. I think that the foundation is not AI. The foundation is data literacy. When you mature, you get into data fluency and data literacy. When you teach that, you can cover a broad spectrum of fields at the fundamental level.

Basically, you need to do that with the schools, the universities, first-time jobbers, even the executives. I think that's where the common language will come in and they have more accessible education and more short courses rather than these long university degrees. I don't say that it's wrong. It isn't.

Loris Marini: What's wrong is the expectation that at the end of the course, it's irrelevant. Yeah. Because the field within those five years moves so quickly.

Debbie Botha: I also think the apprenticeship idea is brilliant where you actually start working, get thrown in the deep end, and get courses as you progress. Learn and apply, learn, and apply, learn and apply. I think that's something that we'll see much more of in the future.

Loris Marini: Hmm. Very interesting. An idea for a business, for sure. Debbie, what's Trailblazers?

Debbie Botha: Yeah. Trailblazer is the new partnership type that I identified for Women in AI. That is for women that are really already very visible and eminent in their careers, visible in their organizations. They have proven themselves to be great mentors and inspirations to other women. I want to recognize them and have them as a partner of Women in AI.

Loris Marini: Man, I love my job. This is amazing.

Debbie Botha: I got the name trailblazers from two women that wrote an article for World Economic Forum. I think Beena Amanath and Kay Firth-Butterfield. They are already prominent women in AI. Beena is the head of the AI Institute, I think, of Deloitte, and Kay Firth is leading, I think AI in the World Economic Forum.  They are true trailblazers with lots and lots of followers on social media.

They said there are basically five ways to increase the women working in AI. The one was, they said, “We have to support the STEM education.” They said, “We have to showcase female AI trailblazers.” They said, “Mentor women for leadership roles and create equal opportunities for them.” We have to reward people in the same way financially and otherwise. That's where I got the name from and I really want to honor these women.

Loris Marini: Well, if Discovering Data can help at least with one of those five, the showcasing bit then that will be a fantastic 2022, I think, to contribute to that idea of getting people to see, “Hey, there is so much you can achieve. Look at these people. Look at what they've done and you can do the same.”

It's not a matter of IQ anymore. It's a matter of how you structure your learning, how you seek different point of views because we learn when we're forced to face the reality that our way's not the only way. Sometimes there is a better way and it might burn a little, especially for those that come from an academic heavy background.

Here's one when the system teaches you to pass an exam. There's a lot of focus on getting it right. What's the mark you got? Is it a 30 out of 30? Is it a 20? It's so pointless to measure that. The outcome is just a number and the number is a huge compression of information. There's so much that goes into passing an exam, absorbing knowledge, navigating a career, getting promoted, becoming more useful for the organization, changing career. There are no books for that. There are people that know and maybe have direct real-life experience doing it.

Man, if only I could access back when I decided that I wanted to become a data scientist and lead a team. It would have been immensely useful to have someone say, “Hey, check this out. What do you think? Am I on the right track here? Or am I completely missing the mark?” If we can create that space, imagine how many people can accelerate that progression because sometimes you just get stuck. We don't know who to ask and yeah, it's just a pity.

Two last questions I wanted to ask you, Debbie. What would you like to see more of in the data community and what would you like to see less?

Debbie Botha: I think more of is this common language that I spoke about. I think that will really help. Another example that I can give of that is we started with these concepts of DevOps and then we said, “Okay, then there's data ops, which makes sense.” There's ML ops, which still makes sense. Now, there's AI ops. Guess what? AI ops does not mean the same thing as the other three. It is AI for ops. It should be called ops AI, not AI ops.

Loris Marini: Oh, they got the order wrong. It's making operations easier by harnessing algorithms

Debbie Botha: Yeah, it goes into the infrastructure and the operations and put AI on top of it, you improve the infrastructure.

Loris Marini: Make things faster and more efficient and sort of optimize the system.  It's an optimization problem.

Debbie Botha: Yeah, exactly. It is called AI ops therefore it should mean, the ops side of AI. Ops for AI, but it means the other way around.

Loris Marini: There's an idea that I'm throwing out there because I did think about something similar in the past, but I did not follow up because I felt a bit of imposter syndrome and that stopped me from actually getting on LinkedIn like, “Hey, who wants to contribute?”

This idea initially came from the Master Data Marathon 2.0 with Scott Taylor. I attended the conference. I loved it. People were struggling to tell stories of how bad data impacted real businesses and people. I thought, “Why don't we create a master data table? We contribute the stories as much as we can share.”

Obviously, yeah, it gets tricky. Might get political. We kind of have to be careful, but if we do it right, instead of having one person crack the problem of data meaning of unifying, creating a vocabulary for all these terms and what they mean, we could have the community contribute to that and have a system of triaging and merging these things. I mean, it's still in the back of my mind that maybe we should execute it. What do you think?

Debbie Botha: I think so. I agree. Absolutely. One of the things that made me come up also with this frustration is when I joined IBM, there was this concept of a data reservoir, but the rest of the world was talking about a data lake. The data reservoir means something different from a data lake, but the vast majority of people inside and outside IBM had completely different understandings and interpretations of these two things.

Yeah, definitely. Please, let’s do that.

Loris Marini: I should start it here. It's going to be really tricky at first. Imagine how many debates on LinkedIn and Twitter threads that people will say, "What are you talking about, Debbie?”

It's fun, right? I don't care. I'm okay with looking ridiculously stupid. It's perfectly fine. As long as we reach a state that is slightly less confusing than it is today, then I've done my job.

Debbie Botha: You mentioned the imposter syndrome. I think a lot of us much, much more than one would think suffer from that. Definitely because if we didn't, then we would have all shouted out, “But hang on, I thought this is what it means. Now you come with this idea, oh, you must know better than me.”

Also, you asked what would I like to see less of, that comes back to this thing of newcomers or youngsters coming up with all these new concepts that’s ages old. I would suggest that they first at least Google if their concept isn't there before they go mainstream and viral with new concepts that the old dogs will tell you they've gone through it. There are books written about it.

Loris Marini: I'm going to say one good thing about science and research education. There are so many good things and bad things about academia. Those that tried both worlds know what I'm talking about, but there's one thing that I really like which is the mindset of putting things in context.

When you start any research, a new experiment, the very first thing you do before you buy the equipment is do a whole lot of literature research, see what people have done. You start from here, thinking that your idea is the smartest, the newest, the most amazing. You start Googling and after two weeks, that's the trough of disillusionment. When you realize that actually the entire world has been doing experiments on that exact same thing. You thought it was a niche and a banging idea, but it isn't. Now, you have to find a new spin or new angle, but that's part of the process.

It looks like in data, we don't want to do that. We just want to say, “Hey, this is new and shiny. Look at me, look at me!”

Debbie Botha: Exactly. You will have something special that you will be able to add to the existing understanding.

Loris Marini: For sure. Exactly. You don't have to go home. Absolutely agree.

Debbie Botha: Yeah, exactly. You don't have to throw it away, just add your sweet spot to it.

Loris Marini: Yeah, but don't screw up the history and the context please, because especially when you read the paper, you're like, “What did these guys do? Oh, okay. Oh, wow. That's a brilliant idea.” The first page I want to read is an overview of everything that's been done in that field to date. How does that thing that you're trying to do compare with the microcosmos of ideas that people already tried? Is it entirely new? Is it slightly new? It doesn't have to be new to be interesting, right? It could be a spin on something that already exists.

Debbie, I feel I took way too much of your time. I want to say, perhaps before we leave, where’s the best place to follow you? Where is the best place to follow Women in AI? Any last thing you would like to share before we moved to a cup of tea?

Debbie Botha: Me, I'm on LinkedIn. Debbie Botha. You can email me at debbie@womeninai.co. Women in AI's website is womeninai.co. Women in AI is also on LinkedIn, very active. I would like to just request the listeners to go and look at the website. Ladies can register as members. There is a place to request partnerships. We have commercial partnerships and non-commercial partnerships for communities and media. Please have a look we'll gladly collaborate with you.

Loris Marini: Fantastic. Yes. Call to action time. Absolutely do that. If you haven't subscribed to the list of Discovering Data, that's also a good time to do that. Cross-referencing each other through the show notes, sometimes it's at the bottom, sometimes on the left, sometimes you have to click, but you know where your show notes are. Click on that.

You'll find a link to Discovering Data and a link to Women in AI. I'll definitely include Debbie's LinkedIn profile URL there as well. We can keep the conversation going. Debbie, thank you so much for being with me, and enjoy the rest of your day.

Debbie Botha: Thank you. Thank you so much. It was such a pleasure.

‍