Why We Need More Data Empathy
Phi Harvey is a big beardy AI geek who loves working with data and solving interesting problems. He is especially interested soft skills for technical people in data, empathy, ethics and in the impact of data on what people know and how they know it. Starting his career with a BA in AI, he has worked in a wide range of industries from surveying, to architecture, to advertising, to being the
CTO and technical founder of a data start-up. Phil now works at Microsoft as and Industrial Metaverse Architect in the Industrial Metaverse Core Team focused on Bonsai. Harvey is also co-author of the book Data: A Guide to Humans
Adel is a Data Science educator, speaker, and Evangelist at DataCamp where he has released various courses and live training on data analysis, machine learning, and data engineering. He is passionate about spreading data skills and data literacy throughout organizations and the intersection of technology and society. He has an MSc in Data Science and Business Analytics. In his free time, you can find him hanging out with his cat Louis.
Key Takeaways
Empathy is needed for both properly understanding data, but also for ensuring clear communication with internal stakeholders about data projects.
Empathy is a skill that can and should be learned by practicing, it isn’t something that you either have or you don’t. This means that anyone can increase in empathy.
Not every single role in a data team needs to be an expert with empathy. Data teams should evolve some roles into “bridge roles” that can support both the highly technical team members, as well as the non-technical stakeholders outside of the team.
Key Quotes
Learning is about practicing. Nobody came out of school being a great Python programmer. You don't just read a book and that's it. You need to develop the muscle by practicing and putting in the time to learn it. Cognitive empathy is the same way, and you can practice in several different ways. The first way is in your work directly, but you can also practice with media: entertainment, films, music, art, etc. All of these things work because you start to engage with them as a form of practice by asking thoughtful questions and being open to what they are showing you.
I talk a lot about people practicing cognitive empathy, data empathy, and technical empathy. It is correct that in some people's roles, these are not appropriate. For example, someone may be your best Python developer, so having them spend a lot of time with people is actually less valuable. This means that you need some role to be a bridge between the non-technical and technical people. Empathy will make everybody more successful overall, but in an enterprise scenario, when you're looking at your culture, you can look at evolving specific roles to become bridge roles, instead.
Transcript
Adel Nehme
Hello, everyone. This is Adel, data science educator and Evangelist at DataCamp. A lot of the time it's easy for us to think about data science or even data literacy as mechanistic, a process by which data comes in and products come out. And it's an entirely physical process. But as we've explored throughout the show, succeeding in data, whether you're a data leader looking to build a data culture, a data scientist ascending the ranks or even a policymaker looking to have an impact with data. The Human Side is crucial at the heart of this human side is empathy, empathy for your stakeholders if you're a data scientist developing a dashboard for them, empathy for your workforce if you're a data or learning leader, or empathy for the planet and your citizens if you are a policymaker. So how can we all practice better empathy, specifically, how can we all practice better data apathy? Here to answer this question is Phil Harvey, industrial Metaverse architect at Microsoft. Phil is a big beardy AI geek who loves working with data and solving interesting problems. He is especially interested in soft skills for technical people and data, empathy, ethics and the impact of data on what people know and how they know it. starting his career with a Bachelor of Arts in AI. He has worked in a wide range of industries, from surveying to architecture to advertising to being the CTO, and technical founder of a data startup. He also co-wrote the book data, a guide to humans on exactly the topic we'll be discussing today. Throughout the show, we talk about how he defi... See more
Phil Harvey
I'm quite strange in that I can say that I've got 22 years in artificial intelligence because I started with a degree a Bachelor of Arts in artificial intelligence in the year 2000. So strange degree in a strange amount of heritage. But at that point, AI was more of an academic concern. So I've worked in a wide range of industries, from building an architecture, 3d rendering to advertising I was the CTO and technical founder of a data technology startup. And now I work at Microsoft and my current role is an industrial Metaverse architect focused on up on-site technology. So you know a wide range of things. And yeah, the book was part of that journey because I needed to learn my way through and all the things that I met along the way when it came to actually making data useful.
Adel Nehme
That's really great and definitely excited to you know, parse the book and deep dive into your work. I had a pretty hard time preparing for this interview, because there's definitely a lot of different angles and ways by which we can approach our conversation and discuss the book. And this are usually a signal that the book is really dense with information has a lot of richness to it. The book does a really great job at straddling between, you know, the big picture philosophical implications of developing data, empathy, and the practical situations where empathy or lack thereof, can really impact organization's ability to leverage data effectively. So maybe as a primer to contextualise our chat walk us through the motivations behind the book and what led you and your co author and Amalia Martinez to write it
Phil Harvey
You're right. It's dense and wide ranging for what we hope to do was to find a balance between the context and some practical examples in that way. Now, Ali and I are both from a technical background and speaking for myself. I was in it I've done most jobs. In it. I've cabled under floors, I've built machines, I've done it support, and I was a programmer for 15 years. And people would hit me with the word empathy, sort of hit you on the head and say you should have more empathy, whether it's in IT support or in data or programming or those things. But nobody ever explained what it meant. I was sort of meant to know what the word meant because normal people know that right? And if I don't know that I'm not normal. So I decided to go and find out, go and learn my way through it and learn what it meant. And because I'm a fan of philosophy, I went that route and went as far up the tree as I possibly could. And I found out the data is deeply connected to a whole bunch of success with technology success was data. We mentioned breaking down silos in that way, and also found that empathy is not just one thing, you know, it's an area with lots of different facets and lots of different things. And so we talked about cognitive empathy, as a learnable skill, because as programmers as technical folks, everybody's on the spectrum, right, a spectrum of different neurological capabilities. And you know, if you'd like machines more than people and those kinds of things, and you start off thinking, maybe I'm blocked here, maybe empathy is a thing that I can't do. And so cognitive empathy was a sort of weigh out skill that you could learn. And that was the motivation to start. And then I've realised that so many aspects of what we do with data and technology are connected to how we understand other people, and that's where content comes in.
Adel Nehme
That's really where I really want to deep dive into all of the different motivations behind the book, but maybe also still setting the stage here. You mentioned here the different definitions of empathy. You mentioned cognitive empathy. You know, an important aspect of making sure that we have a productive conversation is developing a common understanding of what exactly is meant by data empathy, and you do a great job of laying down the different definitions within the book. So I'd love it if you can break down that term and how it relates to the varying definitions of empathy as a whole
Phil Harvey
so we think empathy is at the top of the tree, what I'll do is I'll sort of work my way up from a data empathy perspective. So data, empathy, is a form of empathy applies to data work. And that in itself is a form of technical empathy. If you sort of go up from data work to technical work, which is under the path of the tree known as cognitive empathy. Now, the difference here is we make the split between cognitive empathy and biological empathy. Biological empathy is related to things such as mirror neurons in your brain and there are studies in from animals to plants to all of these things, this thing called empathy, which is kind of built in to the biological being cognitive empathy is a set of skills you can learn and practice and develop over time, just like any other skill. And interestingly in the journey of the book, this is comparable to some of the things that are taught to people on the autistic spectrum, for example, so they've had some great discussions with people on that end of the spectrum about how these skills are learnable can be taught can be practised, and looking at how we can grow and develop in that way. So as you see there, I've sort of split out across you know, humans and animals and plants and these things. You also get things around organisational empathy. So I think the Harvard Business Review has a book on empathy and organisational empathy in that way, and you start to move towards user experience and understand your customer
and those kinds of things as well. And if you want to go complete about it, you can start to look at what's known as effective computing and artificial empathy and those kinds of things which connects most immediately with the ideas of agents that communicate. So you could say chatbots, and virtual agents and those kinds of things. So that's the sort of big picture that we go through. But the book really focuses down into data work because data is this incredibly valuable asset that we now have available. And empathy can make you more successful in your day. To work.
Adel Nehme
And in a lot of ways as well. You mentioned here kind of artificial empathy data also powers a lot of tools that interact with different stakeholders. So also embedding that quality within the development process is highly important to create successful experiences but also safe and rich experiences. Using data products.
Phil Harvey
That's right. So data can give us new things to know and new ways of knowing it. And to do that safely with understanding and all of those things. This is what we mean when we talk about empathy will make you more successful. You need to understand your stakeholders. You need to understand the point and the value of the work that you're doing. If you're really going to connect the dots with the data work to really bring that value out. That understanding is a key part of it.
Adel Nehme
So I'd love to contextualise this conversation, you know, as you mentioned in the organisational empathy sighs specifically to data work. And you mentioned here that cognitive empathy is a learnable skill. I love if you could provide an example of data empathy and action within an organisation specifically and how this skill can be learned.
Phil Harvey
So the learning part is one of the most fascinating parts of the journey and I'll dwell on that for a little bit. Learning is about practising nobody came out of school being a great Python programmer. If you're a Python programmer and you want to do a job working our you need to learn our right you need to practice it. You don't just read a book and that's it. You need to develop the muscle you need to practice you need to spend time cognitive empathy. Is just like that. And you can practice in a load of different ways. The first way is in your work directly. And I'll talk about some specific examples there. But you can also practice with media, entertainment films, music art, all of these things, because you start to engage with it as a form of practice. So if you're going to an art gallery, if it's not something you've done recently, I'd encourage you to go and ask yourself to look at the pictures as if they relate to your data work. It may sound like a strange thing. To do, but look at a picture and go what is this show me? What does this tell me? Under what conditions was it produced? How was it produced? What can I learn about it? Open yourself up and question in that way when it comes to your work? Let's take some specific examples. of something as simple as producing a dashboard. So you're going to get a set of data. You're going to processes in some way. Suddenly, everybody who's listening has probably met at some point, and then you're going to need to show it to somebody even sit down. You can take a ticket off your work system. You can do the coding and do the work and then show it to somebody. And loads of people do that. And they get very grumpy about the feedback and they think the people who received the work don't understand them. That is not practising empathy that is not thinking about the work in the broader context. In that case, you've got to think about the audience what questions you can ask them, what opportunities you have to listen to them to understand the context that your work is going into. So that can be everything from the terminology. So let's say you work in a centralised State Department. And the dashboard is for the sales team. You may think the sales team have no technical understanding at all. But actually the sales team bring the money into the organisation. They pay for your salary they pay for your work by getting customers to engage their organisation. They'll have a language they'll have a way of learning they'll have a way of doing that job. Your dashboard will need to fit into that way of doing the job. This may sound obvious to some people I hope it does but the number of failures in these particular things. They use the wrong colours. They use the wrong terminology. They present something too technical or fashion. The lists are too long. The charts are confusing, they could simplify down to a single example of next best action as opposed to giving people 100 options, all of these kind of experiential pieces. If you practice your data, empathy, you develop the skills of listening and crushing and those things, it means that your work will be more quickly accepted. It means that you'll have a more satisfying experience delivering that product of work. And it also means that you'll get asked to do more work and more work is good because we all got to keep our jobs.
Adel Nehme
That's really great. I love the example of the dashboard. And in a lot of ways it seems like a lot of these problems and correct me if I'm wrong can be alleviated with a checklist or a systematised approach incorporating empathy within the creation of theatre products. I feel like a lot of data folks, you know, tend to be systems oriented people naturally, right. And there's a way as you mentioned here, kind of to bridge the gap and learning empathy by incorporating the systems oriented approach and developing data products. Would you agree that that's the case and have you seen that in action?
Phil Harvey
I'm slightly wary of checklists while I totally agree with what you're saying, in principle, I think I don't want the listeners, for example, to go away and say, oh, we'll create a checklist and that is empathy and we're done. The process of listening needs to be built in. So if you've got a list of work that you're doing, if you haven't included some aspect of listening, then you're not going to do as well. That's the kind of core skill to it. The systems thinking approach. What we provide in the book is a model of how cognitive empathy breaks down a model of what you're listening for the paradigm of your users. Now that can generate a really useful list a really useful set to make sure you've covered everything, and it could be how they learn their epistemology could be how they do things that methodology, the way they describe their world in their ontology all the way through to the ethical concerns and the aesthetics of how the dashboard can look. And so breaking down the system of listening, breaking down the system of the stakeholders is really key. And yes, just as you say, this can be easily written out as a worksheet or you know, a list of things that you do, but the list is not the point. The list is only the way that you make sure you've covered everything.
Adel Nehme
The most important aspect is the conversation is the listening aspect between the different stakeholders involved in the creation of the data product. Would you agree with that?
Phil Harvey
Yeah, listening is something everybody will always hear me coming back to because it's the core number one skill is the best way of practising it's the best way of getting information. In the book. We broke down a whole bunch of other methods within that. One of my favourite ones and I use this all the time is being brave enough to be wrong. People tell you so much more if you say the wrong thing, because they want to correct you than if you go in trying to be right. So while it is all about encouraging them to speak and so you can listen, actually practising the skill of being wrong to generate information is incredibly valuable. So we go into that in the book as well. But it's that it all comes down to you've got to be open to listen. And the cognitive empathy model gives you things to listen for. So it makes it easier to practice.
Adel Nehme
That's really great. So kind of on the theme of here listening, kind of creating that two way conversation. An incredibly important part of developing data empathy is recognising who is in fact uses data and how right as you mentioned in the book, you outlined kind of two common personas who interact with data, data producers and data consumers. Walk us through how you define these profiles. What are the typical roles that occupy these personas? And how do you imagine a productive conversation where empathy is displayed between these two types of personas?
Phil Harvey
So I'll switch my examples to a much more technical set of personas in this case when you're producing products in the data space. That could be the dashboard. It could be an API, it could be a data file, so a data engineer providing a data file to a data scientist in some kind of cloud storage. It could be a database database administrator, or it could be somebody from the market research side gathering data producing questionnaires and those kinds of things. At each point in that journey, there's going to be a producer, and there's going to be a consumer. The producer is providing a product or service to the consumer. And the consumer is receiving that product or service. Hopefully, that's crystal clear. I can't think of how to break it down further that you've got to identify which one of those two you are because it's easy to accidentally fall into consumer type thinking when you're a producer. And it's really easy to think as a producer, especially if you're technical when you're actually in the consumer role. So if you find yourself going things like Why have they done it like this? I do it differently. I don't like the way this API is formed or I don't like the structure this file I would have used a different file format. You're using producer thinking as a consumer. Actually, it can be much more helpful to go I'm the consumer, the producer is providing this product to me. And if I think about why they're doing it from their perspective, as the producer, you'll find much more productive ways into getting the changes that you want to happen.
Adel Nehme
Maybe walk us through an example of how you've seen that shift actually produce the desired change. Maybe for example, in an example that you've seen, or you know, a hypothetical example.
Phil Harvey
There was a particular phase of the Big Data journey around API's and accessing data so very different worlds now, but we were consuming social media API's. We were producing, essentially an email product at the end of it for downstream consumers for our stakeholders. There are two examples that those different junctions so when we're the consumer of the API's, we had testing in place, we had robust connectors. We were consuming, we were being good consumers of that API. And then the producer gave us completely arbitrary numbers. All the testing passed. Everything looked fine on our side. But the numbers were actually wrong. I can't name the company but at one of the social media providers at that point, and so what we produced downstream was wrong and obviously wrong and the customer saw it and then like all the lights were green, all the testing works. All the pipelines worked, all that kind of stuff. But what it led to was digging into the terms and conditions and the documentation to find out why they could produce the wrong numbers and not consider it a problem. Why did we get these numbers which was so obviously wrong out of that API? The reason was because in the documentation, they provided no guarantees that they promised corrections. So if we raised a support request, they would look into it and they would correct it. So they promised corrections. They didn't promise correct numbers guarantees and so that as a specific example, you go, it's easy to be cross about that, and to stamp your feet and these things, but we had to find a way in in that case, it was finding the right documentation. It was entirely non technical solution. It was a bit of reading a bit of tracing to understand how we could protect ourselves because we learn how to implement things to protect ourselves from the way that we're producing. Similarly, on the other side, on the email dashboard side, the customer the consumer of our data product told us clearly, they wrote down they made it very clear in a number of meetings that all of the data for producing the dashboard was in these files. Again, I'm being slightly general just to protect the innocent. They came in our emails, we process them in these things, but every time we sent a sample for tat checking with the end users, they tell us it was wrong. So we dug and we dug, we check the files, we double checked all the technical work, but it turns out that they were gatekeeping where the information came from. There was a key figure which was dealt with in a phone call between somebody senior and somebody on the downstream side of us, which meant the numbers were not in the data, not in the files, not in these things. And again, all the technical work was right. We wasted a lot of time checking. But what we didn't do was truly understand the context and when we found it out, we could trace back all the things that we missed all the gatekeeping the people who didn't like the fact that this work was being done, the people who were sort of protecting their own role and job and those kinds of things. And all the signs were there. We didn't listen properly. We went straight for the technical work and every time we told it's wrong, we check the technical work. We didn't check the people.
Adel Nehme
I think this is an amazing example of you know, the really the importance of developing not just empathy between the producers and the consumers as you lay it out and the technical aspect but also trying to understand the human context and the business context. And the setting around the development of a project. And I think this segues pretty well to my next question and kind of the other theme that I want to cover, which is how data empathy really can power also a data culture. One of the main themes that we talk about here on the podcast is the importance of developing a data culture and really a key hallmark of that is honest context setting, being able to understand the impact of data and how it can be used effectively by everyone within the organisation. So maybe walk us through how does data empathy intersect with data culture, and kind of the imperative for organisations to develop and drive data skills and data literacy within the organisation?
Phil Harvey
liked this question because the answer is simple. In data, there's a lot of complexity to deal with. Data. Empathy is a core skill in a data culture, because it is a skill around understanding, understanding different people understanding different roles and all of those things. So I would say that that is one of the fundamental reasons that we worked on the book was that you're not going to get cultural change unless people have the skills for listening.
Adel Nehme
That's really wonderful. And maybe walk us through how this data literacy here and the skill set to play into the data empathy component as well.
Phil Harvey
Dr. Besser say is a wonderful weapon as a term within data cultures. And one of the areas that we're exploring the book is known as sociotechnical blindness. Of course, our data work is not successful. Because marketing just have poor data literacy. Of course, the IT department don't understand us. They keep telling us to improve our data literacy, but they just don't listen to us. In that you have the words being used as a weapon by group-to-group communication, marketing is not a thing. Is this collection of people under a title? Yes, data literacy is important. Making decisions based on data you need to understand how to read and understand data. But if you find in the language, you're using the term as a weapon, much like I was hit with the term empathy. Earlier in my career, the same thing happens. So when you find yourself saying that what you have to do is to understand the people and understand before you start demanding data literacy, what that would mean there's work to be done. There's a bridge to be built there.
Adel Nehme
A really love that perspective, because oftentimes, we've seen this you know, DataCamp we work with quite a few customers on upscaling and data cultural projects. One of the key messages that we always try to say is that you know, the importance of data skills and scaling data literacy is an enabler to create a common data, language and enabler to create a common language but it's not meant to be used to punish right or to group and I find that perspective very aligned with that, because one thing that we found can be harmful as you say here is that the illiteracy can be used as a weapon to let go of the kind of onus or the responsibility of making sure that they had a product that you're developing for example, is accepted, right and is useful for the data consumer. What do you think are ways data teams or data leaders for example, can avoid this trap and are able to ensure that their data products and whatever they produce for the rest of the organisation is useful and can get continuously accepted regardless of the data skill set of the organisation?
Phil Harvey
So this comes to the more let's call it enterprise technical language when you're looking at that data culture from the roles and the team that you have. Now I talked about people practising cognitive empathy, data, empathy, technical empathy, and all of these things. It is right and correct that in some people's roles, it's not appropriate. They may be your best Python developer and actually, them spending time with people is less valuable. What that means is you need some role, which is a bridge between the non technical and technical people, yes, empathy will make everybody more successful. But in a team in an enterprise scenario, when you're looking at your culture, you can look at evolving roles to become more bridges. So one data team that I worked with, you know, we did sort of day and a half's worth of workshops. From a data empathy perspective. One of their data scientists worked with their manager to change to become a Data Architect. And the architects title as an evolution of the data scientists was a useful change to allow them to be the bridge. They were there to support, the more let's say, focused data scientists to spend more time with that code. And they were there to support the less technical people outside of the data science team to communicate effectively. They were a bridging role. And that's the kind of concrete change that you can make. And people find huge productivity gains in their day to work there because there is somebody who's tasked with that the socio technical role as it were.
Adel Nehme
I love that you know, that data architect or translator role as an important lever within the organisation that can facilitate data empathy. An additional thing while reading the book that popped out to me is that you know, outside of his role changes organisations can make is the importance of like context setting. So the ability to discover metrics, data definitions, documentation, you mentioned here the documentation example, to be able to understand the different components that go into a data product to be able to find data to work with data, how important do you find such kind of tools or components of a data strategy or data culture? And developing and facilitating data empathy?
Phil Harvey
So stick to the word data strategy, you use their documentation, standards, communication, data, dictionaries, all of these things are key parts of your data strategy, which intersect with our data culture, as we've been discussing their standards is a great example because you can find lots of problems of standards implemented incorrectly. You know, the n plus one standards example documentation of dwell on just for a second here because if you're ever wanting to do if anybody who's listening is ever wanting to do some super interesting research from a cognitive empathy perspective, go and look at the documentation of a data product API database, whichever it is, from the perspective of understanding the people who produced it, you'll find signs of boredom, you'll find signs of bad automation you'll find the language that I use the mental models for building that particular data product and all of these things. But you're doing a sort of meta reading of that documentation. You're not just going, Oh, I can't find the thing that I want. You're building an understanding the product from a different perspective because that text has put something of the data producer out there. So standards and documentation will be the two that I'd call out that you need to build into your data strategy, because they will support your data culture.
Adel Nehme
Completely agree there. That's the power of like documentation and being able to just level the playing field across the different functions. That are interacting with data is extremely important. Another thing that's really interesting in the book, especially the appendix covers a lot of different practical examples of data empathy and how they play out in action in organisations. I'd love to open the conversation here and ground the discussion a few examples to showcase read the power of data empathy. Walk us through your favourite examples from the appendix and how you've seen them play out in your career.
Phil Harvey
Thank you for raising the apprentices. We structured the book in quite a strange way that it's actually a game of two halves. The first half is the theory. The second half is the examples in the appendices. And the final appendix is a method called Data landscaping which is a soft top which be interesting. They're my current favourite example, very specifically, which we cover in the book from a data quality perspective. But I want to expand on that data quality is this really fascinating conversation at the moment from a data empathy perspective? Because we've had enough years of data scientists now for implications and ramifications of data science in the workplace to start to bubble up. And that was a survey that I mentioned that I got to discuss with the people analysing the results. And the most interesting one for me is that data quality has become a sea level issue within an organisation so CEO CXO, whoever you want to talk about, says that their organisation has a data quality issue, but that's not true. That is a ramification of the fact that they hired expensive PhD level data scientists who couldn't do the thing that they thought they were going to do. And the easiest thing to blame is that data quality because they couldn't get what they wanted. And the easiest way to express that is the data is of poor quality. Actually, no data is of quality, unless you've already done the work. So you'll find high quality data for a piece of work if the piece of work has been done or tried before. That's the only situation where everything will fit together well, because that data quality work is and still is and always will be the majority of the work of any interesting project because the interesting projects is where you're bringing things together, which weren't brought together before. And whenever you're doing that you're gonna hit producer consumer issues, product issues, all of those kinds of things. So when you work through what a data quality issue is, you find lots of examples where cognitive empathy and data empathy will really help you, because you'll identify that this field was aggregated by month because somebody who ran the database at that point, couldn't get the budget to get more storage for that database. So they had to reduce the size of a table in some way. And the way they did it was aggregate into a field. So shorten the table and dump out the actual data to a log or a file. Turns out, they thought it would persist, but it was cleaned up by an archiving process. There you go. data quality issue, loads of people involved in that.
Adel Nehme
Yeah, that's a great example. So in a lot of ways, trying to make sure I got the best insight here is the data quality example that you mentioned here. Definitely a lot of organisations have done massive investments in getting like PhD level data scientists or machine learning experts without necessarily having the foundation set up first without having necessarily paid attention. To the process of turning data into valuable assets that can be then transformed with the skill set of those PhD level data scientists. And that in a root cause is a data empathy problem. Would you say that that's an accurate summarization of the stance?
Phil Harvey
Yeah, that's a great summary. Quite often, it's not even as clearly played out, as you described there, that the data scientists will be excited to do the work. They'll pitch themselves and their experience and their skill to the organisation. The organisation will say, yeah, we've got loads of data, loads and loads of data. Yeah, we've we're really keen to get the best out of it really wants to do some machine learning. And then what you say will play out the foundation isn't there the data scientist goes, but now I have to do all the boring and hard stuff that I really didn't want to. That's the reason I left my last job because their data was a mess. And I spent all my time tidying it up and not doing machine learning. It's the same here. So the questions you asked during the interview are, how many data engineers do you have? How many will I have on my team who will be there to support me doing didn't ask any of those questions,
Adel Nehme
and that's at the root cause a lack of empathy, lack of understanding and listening conversation component. That's a really great insight. Now, you mentioned as well earlier in our conversation, how data and data empathy can really help us solve some of the biggest problems that we face as a species. You mentioned climate crisis, for example. Walk us through maybe your thoughts here and how you think data empathy solves or alleviates these types of challenges.
Phil Harvey
Thank you for bringing us on first because the climate emergency is a one top priority and should be for everybody. And this is something one of the reasons I partnered with Dr. Noelia in writing the book because she has such a fantastic perspective on these challenges. Now, data is this asset that we have is an informational resource. It's in the space of what we know sort of said data information and knowledge in the same sentence. You know, there's a lot of pieces to that puzzle to philosophically unpack, but you're looking for what data can tell us and when you're looking at systems science when you're looking at the impact that you have on any space. You can look at that from an individual perspective. You can look at that from an organisational perspective. You can look at that from a societal perspective. You can start to look at the systems and impact and processes of the planet and biodiversity of all of these things. You can go back in history into the 70s and you can find the world three model, which was system dynamics model was based on distributions developed from the data and understanding of the climate impact of hydrocarbons and society at that point, it has been proved accurate to this day. A piece of modelling from the 70s people who understand how to process data, understand how to look to the root causes of the challenges that we face. You can understand biodiversity from incredible different angles from satellite images. From testing water at mass scale, you can gather data on almost anything that you want, model it work with it and learn something on which you can take action and that's the key piece whether it's in business, whether it's personally whether it's the climate of urgency or biodiversity or any of those pieces. The data is about driving action. Just knowing is not enough. We can go to Buddhism as we want where knowledge, wisdom and action are the same thing. That's where you got to start thinking about that. How clean is the cloud that used to process the data at work? There is a way of finding out there is a way of listening to the system. Which are in the system of people, the system of the environment, the system of the world.
Adel Nehme
That's really great. And I want to go into maybe have a bit of a tangent here and kind of discuss the different dynamics of how to create a message that drives empathy. You mentioned here the climate crisis, a large portion of why there's inaction on the climate crisis is, you know, a bit of a messaging problem around the climate crisis. Do you think that there is a case to say that to drive empathy within people you know, there's the Paul Bloom book on against empathy where he describes the distinction between driving empathy with statistics and with data, versus driving empathy with stories, right, and that stories in themselves are much more personal. They drive action much more. Do you find that distinction holds when having a conversation around something like the climate crisis, and how that plays out?
Phil Harvey
Yes, I really enjoy problems book. And I think just before digging into that topic, the key call out of the spotlight effective emotional empathy is a really interesting part of Blooms theories. And it also relates to what you're saying that the stories can have that problem as well. And when you're looking at the full scope of what empathy can do, just like any other tool that we develop any other skill that we have, you can use it on a spectrum of good things to bad things, as it were. So stories are greater driving emotion. If you've ever had a film, make you cry all those things, or if you ever switch channels because the charity advert was just too hard to take in that site. Then that's the story tugging on your emotional empathy, cognitive empathy, and some people talk about rational compassion on this aspect, which you know, statistics data and information on that end, doesn't feel the same. It doesn't tug on your emotions, it doesn't create that spotlight effect. So I'd say that while it's easy to say it's a blend of each of them, if you understand that spectrum, and you understand what you're trying to do, yes, a story will help you at one point, statistics will help you at another point. And the fact that you raise those different points is incredibly important because I'm talking about cognitive empathy is a learnable skill. It won't make you a super person, right? You're not going to suddenly your head's not going to light up like a light bulb and you'll be able to fly but it gives you a tool to use just like Python, just like storytelling and all of those things. So I totally agree that storytelling is a massively important skill, but it's not the only skill.
Adel Nehme
I completely agree there. So as we wrap up, I know that we're reaching out of time, I'd be remiss not to also talk about where data empathy fits into the conversation with the rise of really robust generative models such as Chad GPT dolly to and more something you talk about in the book is that artificial intelligence has the potential to free us from drudgery and elevate our quality of life and work. And additionally, you mentioned that the road to these technology the road to that time and place where we are freed from drudgery will be turbulent as we grapple with these technologies. How do you see the next few years playing out with generative AI? How do you see anticipate data empathy or our applications of it changing and evolving with the advent of these technologies?
Phil Harvey
Huge, huge topic. So there are loads of aspects to the way that these models can be used. Let's talk about language. For a second as one of the focus pieces in that you just mentioned storytelling, and the emotional manipulation possible within language and storytelling and those things however, language as a dataset is fascinating. Because language doesn't carry meaning. Language causes meaning in the mind. And so if you use it, expecting it to carry meaning, and that to lead to emergent magic properties of AI systems, then you're going to fall foul of a whole bunch of different problems. Now, in the development of these models, we talked about the PhD level data scientist who's great at machine learning and all these things. You can track that back to the way that PhDs are trained in academic scenarios and you can find in the AI conversation, whole set of aspects where empathy will really help understanding the impact of these tools. If you take Darlie for example, there is a conversation burning at the moment around this was trained off images on the internet. If you're looking at style transfer a lot of the images that were used to generate the ability to do style transfer in that way belonged to artists who shared them under a particular consideration in that way. There are people behind all of this, right? So developing skills to understand people's paradigms and those things will play a key role in how AI develops, it will even play a key role in how AI interacts with us because our paradigm is not in the data we leave behind it's within us. So being able to model that in a different way is incredibly important. And so it's a really huge topic, if we can cover it from the perspective of how AI will change over time. But I really want to bring it back to the idea that people will be working on this. Hopefully somebody who's listening will be working on an exciting AI project and they may go Where did we get that data from? Why am I so frustrated about this thing? How are we going to deliver it to that set of stakeholders who don't understand how it works, but want to see the value out of it? That's the thing that really matters to me that we can talk in length of all the grants different topics, but it comes down to the people who are you going to listen to next, whether it's data, whether it's dashboarding, whether it's machine learning, or whether it's AI,
Adel Nehme
that's really awesome. There's definitely another episode in the making here to discuss all of these topics. Finally, Phil before we wrap up, do you have any final call to action before we wrap up today's episode?
Phil Harvey
Well, I'd obviously love it if people are interested enough to read the book. So thank you. But please understand that empathy will make you more successful. If you're not working or doing any data work to be successful, fine, carry on as you are. But if you're looking for success in that work, skills and cognitive empathy will help you be more successful. we go into great detail in the book about how that happens. you know, we don't leave it as a statement, we unpack it. The main way to do that core thing that you're going to need to learn to do is to listen and that's where I want to leave it. Everybody just starts listening a little bit more amazing.
blog
Data Demystified: Avoiding the AI Hype Trap
blog
Data Demystified: The Different Types of AI Bias
blog
Seven Tricks for Better Data Storytelling: Part I
blog
What We Learned From Teaching 1M People Data Science
podcast
Post-Deployment Data Science
podcast