---
id: "workshop:data-series-an-introduction-to-data-and-its-reuse"
type: "InnovateUS workshop"
title: "Data Series: An Introduction to Data and Its Reuse"
url: "https://innovate-us.org/data-series-an-introduction-to-data-and-its-reuse"
date: "2023-10-24"
duration: "60 minutes"
language: "English"
instructors: ["Stefaan Verhulst"]
recording: "https://www.youtube.com/watch?v=REhFRhpl9Wk"
transcript: "automatic, from the recording"
source: "InnovateUS (innovate-us.org)"
licence: "CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/), as stated on innovate-us.org"
---

# Data Series: An Introduction to Data and Its Reuse

InnovateUS workshop · 24 October 2023 · 60 minutes

In this 60-minute virtual training geared at demystifying data and highlighting the significance of data in the digital age, participants will receive an introduction to data and its reuse, including an overview of the different forms of data, an explanation of data reuse and why it matters, and the benefits and challenges of reusing data. Participants will explore the questions that data can provide answers to and understand the scope and limitations of data-driven decision-making. The workshop will include real-world examples of data-driven solutions and outcomes, case studies showcasing the successful reuse of data across industries, and recent developments and emerging trends in the data ecosystem.

## Learning goals

- Learn about the different forms of data
- Gain an understanding of data reuse and why it matters
- Understand the benefits and challenges of reusing data
- Explore the questions that data can provide answers to
- Understand the scope and limitations of data-driven decision-making

## Speakers

- **Stefaan Verhulst**, Co-Founder of The GovLab, Co-Founder and Principal Scientific Advisor of The DataTank

Recording: https://www.youtube.com/watch?v=REhFRhpl9Wk

## Transcript

Automatic transcript of the recording: names and terms can be misheard, and the times can be a few seconds off. Each time links to that moment of the video.

**Stefaan Verhulst**

[0:00](https://youtu.be/REhFRhpl9Wk?t=0) What is already the experience of people that are joining as it relates to data, but I was hoping to use today to basically set a few or share a few observations when it comes to the current data space and how one can start thinking about data and datafication as it relates to, for instance, making policies or designing public services at the local and state level, but also, of course, at the national level as well. And so let me start by hopefully sharing my screen and not embarrassing myself and showing all kinds of other items that might be open on my laptop at the moment. But as already was mentioned, I think there are already some Zoom housekeeping things that have gone and have been announced, so I don't need to touch that. But as Hannah already was mentioned, I'm the co-founder of GovLab. And for those who don't know GovLab, we are kind of an action research center focused on how do we transform the way we make decisions. And typically what that has meant is that we have focused on kind of two assets. One is people and how do we leverage people's expertise in new ways and bring their expertise to the policy or decision or problem solving lifecycle.

[1:27](https://youtu.be/REhFRhpl9Wk?t=87) And then the other asset, of course, is data. And I guess that's why Hannah asked me to be part of this wonderful series to really reflect on what's the value of data and how can one approach data in the current environment as it relates to leveraging data to change the way we go about making decisions. And one of the key elements that we have focused on is that if we want to start using data for decision making, obviously the key challenge is having data. And so one of the areas that we have developed further is something called the Open Data Policy Lab, where we try to unlock data that is already collected, unlock that data for other purposes, such as public interest purposes as well. And I think that's also something that I will reflect on throughout those three series, which is really about what do we do with actually the data that is already collected as well. So much of the work we do around data starts from the assumption and the premise that data can help us innovate how we go about solving public problems. And it is for all of you on today's call, it's no secret that we are confronted with multiple crises at once, whether it's about homelessness, whether it's about disinformation, whether it's around migration, whether it's around climate change and the increase in environmental crises, or whether it's about economic conditions as well. And what is also clear is that all those challenges are kind of wicked problems, meaning require innovation, not only in terms of solutions, but also innovation in terms on how we develop solutions to those wicked problems. And that's where data comes in, because from our perspective, data has become kind of a new asset in the toolbox of decision making, in the toolbox of public problem solving.

[3:31](https://youtu.be/REhFRhpl9Wk?t=211) Obviously, it was already there, especially when we are talking about official statistics. But what I hope to do today is to reflect on other data sources beyond official statistics that also might be relevant and have become key assets for decision making as well. And I will try to do this, not by going into today, not by going into giving you kind of do this, do that, in order to leverage data. But today, I really want to set the stage and develop some kind of a mindset on how to think about data, and how to think about data and use it for public interest purposes as well within the current environment that we are living in, which according to some is, of course, an environment of data or the data age, as some have declared the current situation to be. And so the first observation, and I will throughout these 40 minutes that I have today, I will try to make eight observations that sets the stage and perhaps enables you to think differently about data somewhat than when you joined earlier. But I realized that some of you are already very high level data professionals. So I also hope that in the Q&A, we can have a discussion about those observations, to what extent they resonate and to what extent they actually require us to think differently about data than we've had before as well. And the first observation, and again, none of this might be new to you, but the first observation is, of course, that digitalization, the fact that most of our society, whether it's at the corporate level, but also at the public sector level, have undergone a digital transformation, and that as a result of digital transformation, we have seen a massive datafication as well, which is why

[5:32](https://youtu.be/REhFRhpl9Wk?t=332) we, of course, now live in the so-called data age, because the moment you have a digital environment, whether it's using, for instance, a digital transaction methods or whether it's about using social media, for instance, to communicate and to respond to the current environment, the moment you are in a digital environment, that automatically means that data is being collected. And so we see a massive datafication as a result of changes in the way we collect data, and especially, of course, in digital data. Also, we've seen the creation of sensors that are getting smaller and smaller, and that allows us to actually really get kind of real-time data of location-based environments, but also changes in the way data is stored, right? The fact that we have actually very limited storage concerns and storage limitations anymore is a fundamental shift than we had in the past. Of course, we can argue to what extent that has had environmental impact as well.

[6:41](https://youtu.be/REhFRhpl9Wk?t=401) But it is fair to say that the storage has also changed the way how we collect data. The fact that we now actually collect data en masse is a key differentiator than what we had in the past. We also have changes in computation, analytical capacities, and I already referred to the fact that Innovate US is also going to focus on artificial intelligence, which obviously is a new method to actually make sense of training data through machine learning or through large language models. But that in itself has also created a whole new environment. And at the same time, while we have seen a massive increase in collection, we also have seen indeed an embracing of the data in how we actually go about making decisions, whether that's at the personal level, just look at when you leave the house, you're going to check your weather forecast, or when COVID happened, everyone, of course, was checking out data before. They also were going into certain kinds of locations, for instance. And so decision making has really become more data driven and there is this thirst for actually leveraging data for decision making as well.

[7:53](https://youtu.be/REhFRhpl9Wk?t=473) So that, to a large extent, has not only led to more data. And again, we can look at all the stats that are out there, which basically triples each time when I look at it. But it also has meant that we have different types of data. And of course, big data is kind of the old fashioned way of looking at it. But it is mainly important to reflect that when we talk about big data, we're not talking just about the volume. We're also talking about different quality of the data in terms of velocity, by which I mean that in many cases, the data that we have at our possession is fundamentally different than data that we had in the past, which was quite often the recorded data and didn't reflect kind of real time data sources or didn't reflect real behavior as well. It was recorded data based upon, for instance, surveys as well. And so we have fundamentally different types of data that can provide us real time data, but also can provide us actually insight of real behavior as opposed to reported behavior as well.

[9:01](https://youtu.be/REhFRhpl9Wk?t=541) And of course, we also have seen a massive amount of variety as well. It also has meant that we have a lot more unstructured data. In the past, when we were talking about data, we were typically talking about structured data, data that was collected in a structured manner, where we had a clear column for age, we had a clear column for location, a clear column for gender. Now, that is still fundamentally important, of course, when we want to do certain kinds of analytics. But what we have seen is that as a result of digitalization, and as a result of datafication, we actually had a lot more unstructured data. And if we think about generative AI, for instance, and again, there are already courses at Innovate US that really goes into detail, but generative AI basically is all about unstructured data. It's about the fact that they have found a way to make sense of unstructured data in order to predict, for instance, answers to prompts as well. And so that's a fundamental shift in how data is being used and how data also can be used for decision making as well. The other element is, of course, that, I skipped some things here, is that we also have a lot more variety. And here I just provide a few examples on how new data sources can be used for decision making. For instance, each time when there is a certain level of consumption, for instance, you go to a grocery store and you use your loyalty card, or whether there is a financial and commercial transaction, there's quite often a digital trail. And that can now also be used for, for instance, measuring how local businesses are doing. And so just an example here is the Bureau of Economic Analysis have started using, for instance, daily card data, which is gift cards or even credit card data to measure reduction in revenue of local businesses, for instance. And so this is an example on the new kinds of data sources, which we quite often call non-traditional data sources

[11:07](https://youtu.be/REhFRhpl9Wk?t=667) that can be used for, for instance, in this case, improving or complementing official statistics as well. Similar each time when we communicate, for instance, we do a search on Google that leaves a data trail and that can be used as a proxy for actually real behavior. For instance, CDC started using, together with Google, started using search data in order to actually predict flu outbreaks. And of course, the realization from that was that when people start searching for, for instance, what do I do when I have those symptoms, that that actually might be a proxy of real behavior. And as a result, could be used for actually also predicting certain kinds of outbreaks. The key lesson from that was also that it actually has to be complemented with also additional syndromic surveillance as well, that you not can only rely on search data as well. But it is an example on how non-traditional data can be used. Same thing for, if you use, for instance, a car sharing or a car service that leaves a trail that, for instance, can be used by cities in order to understand urban mobility and to what extent urban mobility differs depending on different times, time periods, or also differs in different kind of geolocations across the city as well. Same thing is that you can actually start using, for instance, YouTube data in order to understand perceptions with regard to certain kinds of phenomenon or with certain kinds of interventions as well. When you start looking at, for instance, comments to videos, you could start leveraging that as kind of proxies of sentiment analysis. You can also start using data that emerges from producing products and producing goods in order to understand supply change and how that might impact your, for instance, policymaking as well. And similar, you can also start using, for instance, data that is being collected when managing infrastructures

[13:13](https://youtu.be/REhFRhpl9Wk?t=793) in order to understand, for instance, in this case, understand using data from waterways and the use of AI to understand the reliability of drinking water treatment systems, for instance, or even in the context of COVID that was also used for, for instance, identifying to what extent there are infections in wastewater and to what extent that can be used to actually predict COVID outbreaks as well. And so that suffice to say, and that's my observation that I wanted to make, it suffice to say that we have, as a result of datafication, a lot more non-traditional data. And this is here an example of work that we've done on non-traditional data for COVID-19, just to show you the amount of new data sources that we have at our possession in order to start leveraging that for decision making than what we had before, which was quite often very highly structured data or data that was collected for one purpose and only used for that purpose as well. And as we can see, now we have a lot more traditional, non-traditional data. And so the implications for people that are eager to think about using data for decision making are multiple. One is that you really have to start thinking beyond traditional data sources, right? Traditional data sources, by the way, remain fundamentally important because they quite often provide the ground truth of certain kinds of phenomenon, but now we have a whole range of additional data sources that can be used and that can be leveraged in an innovative way. Yet, as we will see, and that's also where I spend a lot more time, The following weeks, it quite often means that you actually have to start thinking about partnerships. Partnerships within government, because quite often those non-traditional data is being collected by other agencies, but especially partnerships

[15:14](https://youtu.be/REhFRhpl9Wk?t=914) with the private sector, because most of this non-traditional data quite often gets collected by non-traditional actors, such as startups or private sector, that has become the primary and dominant data collector, as opposed to in the past, where the public sector was quite often the dominant data holder and collector. And so that has, datafication has seen a shift in actually that kind of relationship and that kind of position, which means that if you are eager to use non-traditional data, also means that you're going to have to start looking into new kinds of partnerships. And that the first big takeaway perhaps is that A, we have a lot more data, but B, it also means that we're going to have to innovate on how we make sense of that data and innovate on how we actually have access to that data, because mostly that data will reside somewhere else. And I will come back to that. And I will spend in the third session of this series, a lot more time on how you actually can set up data collaboratives, what are the steps, and also reflect a lot more on this new function of data stewardship, which I think is fundamentally important in order to provide access to data.

[16:32](https://youtu.be/REhFRhpl9Wk?t=992) The second observation, and again, this will be obvious to some, but I think it's worthwhile reflecting on, is that data is not a thing, right? It is actually a process. And people quite often, when we start talking about data, quite often forget that actually there are multiple decisions that need to be made across the data lifecycle. And that if you want to start using data, you need to actually start thinking about what is the value chain, what's the lifecycle of data, because as I mentioned, there are multiple players along the value chain, multiple decisions that can have a positive or a negative impact on how you ultimately, at the end of the lifecycle, start using data for decision making. And so when we start thinking about data, we don't have to assume that suddenly it emerges, and now we can start using it. It actually is a result of a process where multiple decisions have been taken. And that starts, of course, with the planning stage, where you first need to have a question, and we will reflect on that also next week on what is actually the question you seek to answer, because that ultimately will determine what's the data you will collect, or what is the data you can reuse to a large extent. And so someone has to, at a certain point in time, reflect on the question, and whether that question is actually the one that matters, or whether that's the one that will inform the decision is, of course, fundamentally important, because that will determine what data is being collected. Here, there's a whole range of risks, such as, for instance, you collect data without actually, especially when it comes personal, without consent, or you collect data in some kind of a dirty fashion, so that you actually have data that is not correct because of the data collection process. Eventually, you need to start processing

[18:34](https://youtu.be/REhFRhpl9Wk?t=1114) that data, quite often clean it or prepare it for analysis. And here again, there could be bias, as we will see today as well, that gets infiltrated in how we go about preparation of the data. And then subsequently, this quite often is either shared or analyzed. And that in itself is, of course, a process that requires distinctive capacity, including quite often an understanding of the problem at hand, because what we quite often see is that data is being analyzed, but without the domain expertise, which automatically then means that you might have insight that actually doesn't make sense, given the domain. And so that, again, creates some kind of a bias, it creates some kind of a poor analysis, which ultimately leads to then presenting the data, and even the visualization of how you present data could be either positive, could be informed decision, but it could also create more noise to a large extent as well. And so the key takeaway of this observation is that you really have to think about data as a process, where there are multiple elements that comes into play, that if you want to start using data, you need to be aware of the provenance, for instance. Where does the data originate? Who decided to collect it? And for what purpose? If I now start using it for another purpose, actually doesn't make sense. But also, anyway, what happened with the data once it was collected?

[20:08](https://youtu.be/REhFRhpl9Wk?t=1208) Has this led to infiltration of new biases? Has this led to dirty data, and so on? And so it requires collaboration, which means that, again, data is in the most cases a collaborative kind of sport, it is a collaborative exercise, because you do need to know where the data originated from, and you quite often have to work with other actors across the data lifecycle as well. It also means that if you, for instance, advocate for data, but you only fund one part of it, then that quite often may mean the end of the full lifecycle. So you need to really start thinking about end-to-end resource allocation, where it's not good enough to fund for a collection, if you then have no understanding on how that will actually be processed, or you have no funding for making sense of it, let alone no funding for then subsequently visualizing the insights as well. So that's the second observation, which I think is important to keep in mind, is that there are multiple decisions, multiple opportunities for improved decisions, or multiple opportunities for actually infiltration of noise across the system that one has to be aware of. And as a result, if you start using data, it's really important to think about what was the lifecycle, what was the value chain that influences my data. The third observation, and I go quick here, because I do have eight observations that I want to go through, and most of this quite often might be already obvious, but I think it's the collection of all the observations that I hope to set the stage for the next sessions as well. And the third observation is that data in itself has actually no value. And so quite often there's a lot of discussion about what's the value of data. Well, the value of data is directly linked with, of course, the question

[22:10](https://youtu.be/REhFRhpl9Wk?t=1330) we seek to answer, or the insight that we hope to generate. But more importantly, you could say, well, all right, now we know something, right? Because now we have actually much more data, now we actually know how the city moves around, for instance. But if you then not use the data to improve decision making, again, the value of the data might be limited. And so it's important to think of the value of data as a two-step process. The first step is that you actually need to generate insight from the data, otherwise data is just data, but no insight. And then the second step is, of course, someone has to use this insight to make better decisions. And so from my perspective, if you just focus on the data without understanding who needs to make decisions, what's the kind of insight that the decision maker needs in order to make better decisions, data quite often is just an activity to generate insight, but without creating fundamental value and without actually improving decisions as well. And so it's really important to think about what's the decision one seeks to influence, what's the insight that is needed. And by the way, it's also fundamentally important to have a commitment that the insight will be used for decisions, because I've seen so many data projects where insight was generated and then victory was declared, which ultimately led to no value creation as well. The second element that I want to, because we are talking really about data for decision-making and data for policymaking, is that it's also important to think about when do you need the data at what part of the decision lifecycle, right? Because when you start thinking about decision-making, the first element quite often when it comes to policymaking is really about setting the agenda. What are we going to

[24:11](https://youtu.be/REhFRhpl9Wk?t=1451) focus on? What's the priority? That requires specific insight, specific data to inform that stage. When you then go to the following stage where you actually design the solution, where you do ideation of what should we go about and what kind of solution should we implement, then you need again different types of data, quite often data that can inform you what works, what does not work, rapid experimentation perhaps of the solution, or data that allows you to simulate how you go about designing the solution. When you go to implementation and then subsequently evaluation, again it's a different data situation where you might require new and different data in order to inform whether something worked, which is different than perhaps how you started, which was really to try to understand what was actually the problem. Having said that, you still need a baseline of the problem in order to then subsequently evaluate the impact vis-a-vis that problem as well. And so the second element with this observation is that data has no value, it depends on the insight and the decision, but also different data needs might be present depending on the stage of decision-making that you seek to inform. It's not just about any decision, it's about a decision at a particular point in time that may require a particular kind of data, a particular kind of insight, and really for those who want to start using data, it's really important to become more sophisticated in order to understand decision-makers' needs as it relates to the data.

[25:49](https://youtu.be/REhFRhpl9Wk?t=1549) And so here the implication of this observation is twofold from my point of view. The first one is what I call first mile considerations. Before you start either collecting data or soliciting access to data or thinking about a data initiative, it's really important to think about decision-makers' needs, what is the decision that you seek to influence or inform, at what stage of the decision lifecycle, and more importantly, and this is where we're going to focus a lot of time next week, what's the question that we need to answer, because the value of data is directly linked with the question, and is this the right question, is this the question that we should ask at this point in time, or is there another question, which is why I've been advocating for question literacy or question science that would complement data literacy and data science, because without focusing on the question that will inform the decision, it's very hard to know what data is actually needed as well. And subsequently, we not only have to declare victory when data and insight is being generated, we also have to really start understanding the environment in which data then subsequently will be introduced or the insight will be introduced, and that's what I call the last mile kind of consideration. First mile is really about what's the question, and the last mile is then, well, how will decisions actually be used, how will decisions be made using the data, and so here we need to understand who actually needs the data, how will it be used, and can we come up with actionable intelligence from the data in a way that actually it is user-friendly for the decision makers, but also, and we will reflect on that

[27:49](https://youtu.be/REhFRhpl9Wk?t=1669) perhaps in the third session, we also need to start thinking about actually changing the way we make decisions for data to be valuable, because quite often we still have decision cycles that actually doesn't allow for data to be integrated, and so we need to start thinking about can we also innovate at the last mile, not just the first mile of actually generating data, but the last mile on how we actually start using data for decision making as well. The fourth observation is that quite often people think is that data somehow gets born, and as a result, it's kind of perfect, but the fourth observation is that data is actually never raw, right? There's a lot of discussion about if I only would have access to the raw data, and it is also never born, it is actually created, right? Data is an artifact, right? An artifact of construction, where someone decided that we need this kind of data, and someone decided, especially when it's structured data, what kind of data points are needed to be collected, right? Do we collect gender? Do we collect age? Do we collect any other kinds of socio-demographic information as well? But even in the non-structured environment, the way how data can be shared, right? I think in the past, anyway, when you had kind of a limited amount, when you were using social media, a limited amount of to share data, well, that determined, right, what data subsequently were shared as well, and so data is a construct, right? And we need to start thinking about that and be transparent about it, because it is a construct, there will always be biases to a large extent, and we cannot really have unbiased data, because someone has decided that these are the elements that we want to collect as well, which, by the way, subsequently, of course, will have impact on how we then subsequently use this data, whether it's using algorithms, or whether it's

[29:53](https://youtu.be/REhFRhpl9Wk?t=1793) using other kinds of analysis tools, which automatically means that there will be some kind Hello, everyone, and welcome to the second part of our series on data and its reuse. Today, we're going to be talking about the role of bias in the data and in the way we go about making sense of it as well. The fact that there is no data in itself can be a challenge, right, and is ultimately a decision at a certain point in time as well. Again, I'm not going to elaborate on all the areas where there could be some kind of challenge where bias comes in, so that it's never raw, it's actually always somewhat manipulated, or always somewhat processed and prepared. But it suffices to say that we need to be honest that it's very hard to limit bias. But that in some cases, bias is there for a particular purpose to answer a particular kind of question.

[30:56](https://youtu.be/REhFRhpl9Wk?t=1856) But if we start reusing that for other purposes, then it might become complicated if, for instance, the data is not representative of a particular kind of group. If we start using, for instance, social media data, well, not everyone is on social media, so there is already kind of a selection bias there. And so we need to start thinking about what are the kinds of biases that might be part of my data set, because it is always a construct as well. But also, we also need to be honest that in some cases, no data might also be equally problematic. And so here we come into this whole area of data invisibles, which is where some groups are actually never part of the data, which means that there is, of course, bias, but it's not bias in the sense that there is biased representation, there's just no representation. And so we see this quite often as it relates to gender.

[31:56](https://youtu.be/REhFRhpl9Wk?t=1916) Many, many situations, there is no collection of entry data that is uniquely and specific to women, for instance. And for those who are interested, can check out this book, Invisible Women, where it was made clear that in many cases, the standard is always kind of for men. And as a result, you have decisions that are made based upon data, but they don't take, for instance, specific considerations into account that women might encounter. We've done a lot of work when it comes to urban mobility, for instance, in many cities, turns out that urban mobility is actually gendered. Women travel differently than men. But when decisions are made, quite often it is assumed that everyone has the same kind of mobility pattern as well. And so that's quite often because we don't disaggregate the data at a point of collection when we try to understand mobility. But also having no data might also mean no credit or might also mean that you actually have no access to services as well.

[33:00](https://youtu.be/REhFRhpl9Wk?t=1980) And so data is never wrong. There's always a point of decision to include or not include. And so we really need to become more sophisticated about those kinds of decisions and to what extent there is bias in the data. But it also requires that we need to collaborate more with the ones that might be domain experts to really understand what the bias might be and what the implications might be as well. All right. Fifth observation is that, as I already mentioned, while we have a massive amount of variety, while we have seen that, of course, we need to think about data as a process, also never raw. There's always a set of elements that are already decided that influence the quality of the data, the bias of the data. The real opportunity, from my point of view, is actually the fact that data can be reused for other purposes. And this is the result of the characteristics of digital data, which is where I started with already today's session as well.

[34:04](https://youtu.be/REhFRhpl9Wk?t=2044) And so digital data is really intangible, but also easy to copy and replicate. It's also non-rivalrous good, which means that if I actually use it or share it or transfer it, it doesn't mean that my original copy is unusable. It's a non-rivalrous good, so others can use it without loss of quality or even loss of ownership, if that's a term that can be used in the context of data. But also is that actually the value of data expands by making it more accessible and by combining it with other data. So the real potential, from my point of view, which is also where I will focus on the following sessions, is actually to reuse data that was collected for one purpose, for another purpose, because it is now digital and it can be used as well. And we've done a lot of work about the potential of reuse, especially when you open it up. But it doesn't always have to be fully open. It can also be made accessible under certain conditions.

[35:07](https://youtu.be/REhFRhpl9Wk?t=2107) But there are a whole range of potentials when you start reusing data. And so here, for instance, you can actually advance public service provision by, for instance, making data from charities more available so that actually you can understand what are the needs of charities within a particular kind of context. You can also start reusing data for, for instance, environmental justice purposes. This is an example from the city of Sainsville, where unless you started reusing data that was collected for another purpose, unless that moment, you could never make the case that there was actually some kind of redlining as it relates to provisioning of water, for instance. And of course, here in this case, by reusing data, you could actually document the unequal access to, for instance, water as well. You can also start reusing data to improve, for instance, economic conditions. This is an example that we worked on in New York City, where we provided access to data to small and medium-sized enterprises that they typically don't have access to.

[36:14](https://youtu.be/REhFRhpl9Wk?t=2174) But by reusing certain data, they now were in a better situation to actually make decisions with regard to location. And of course, COVID was a clear example. And again, if you're interested, happy to share a report on that, where if you start using data for one purpose and now starting using it for the COVID response, you obviously suddenly have actually data that was not easily available before, especially at a time of non-pharmaceutical interventions was critical to actually have access to data that informed, for instance, social distancing, but also informed whether you should wear a mask or not wear a mask and so on. So that's the key observation is that because digital data has particular kind of characteristics, it can be reused. And so it's important to publish with multiple purposes in mind, but also to embrace the FAIR principles, which is really about making data findable, accessible, interoperable, and reusable.

[37:18](https://youtu.be/REhFRhpl9Wk?t=2238) Because while data can be reused, it's not always reusable. And I think that's another key element that I want to leave behind is that while we have a lot of data, and while we can reuse the data for other purposes, it still requires a reusability kind of strategy so that actually your data or others data increases in value because now it can be used for other purposes as well. Last element I want to make with this observation is that because we are reusing data for other purposes than initially collected, that in itself could be problematic from a consent point of view, right? Because consent quite often dedicates that you can use the data, especially, of course, mainly when it's personal data, you can use the data for the purpose for which it was collected. Now, the challenge with digital data is that it can be reused for other purposes. That's actually its real value sometimes. But obviously, that means that we have to revisit consent.

[38:21](https://youtu.be/REhFRhpl9Wk?t=2301) But more importantly, that we have to start thinking about, do we actually have the social license for reusing it? As Hannah was saying, I do a lot of work with statistical agencies worldwide on how statistical agencies can reuse data for collected for one purpose for another purpose. But that actually quite often requires that statistical agencies also have, for instance, a more public engagement on whether that reuse is appropriate, under what conditions there is a certain level of social license, and to what extent this can be used for other conditions as well. And so that's my observation with regard to reuse. The sixth, and we're almost there, is that while data can be reused, right, and while we have a lot more variety of the data, and a lot more veracity of the data, it turns out that in many cases, the data you need, you don't have, but someone else has.

[39:23](https://youtu.be/REhFRhpl9Wk?t=2363) And so what we see is that as a result of datafication, we actually have new data asymmetries, new power asymmetries emerging as well, by which I mean is that the ones that need data quite often don't have access to the data. And as a society, we quite often have not figured out how to match demand and supply. And so we have the creation of new data asymmetries, whether that's within a state or within a city. There's a lot of data there. Unfortunately, quite often the data is not accessible for reuse, and as a result, we have those new asymmetries as well, which also explains why, if you think in terms of, for instance, success of data in the corporate sector versus success of data as it relates to solving public problems, that the corporate sector has been far more effective because they also have worked on overcoming some of those asymmetries. While in the public sector, quite often we have those remaining asymmetries where we don't have access to the data the public sector could use, and as a result, we also have seen a slowing of actually the impact as well.

[40:31](https://youtu.be/REhFRhpl9Wk?t=2431) Now, as I said, there's a longstanding literature on information asymmetries, data asymmetries, and to what extent that creates new power asymmetries. But it's something to think about in your own situation, right, is that what data could actually improve your mission or the way you go about your mission, and who has the data, and how, anyway, does this asymmetry actually play out as well? Which means that we need to, and that's going to be, as I said, the real focus of session three, we really need to start thinking about how do we overcome those data asymmetries so that we really can start reusing the data that has been collected in a massive variety kind of fashion for improving public service, improving policymaking as well. And that requires data collaboratives. It requires a new kind of partnership that can match demand and supply in a far more systematic, sustainable, and responsible way.

[41:35](https://youtu.be/REhFRhpl9Wk?t=2495) And it also means, and we will see this later on, is that this requires actually a distinctive capacity. I think too often we have focused on data scientists as the solution to all the data problems. I think data scientists are fundamentally important to making sense of the data. But if you have no data, or you have no access to the data, then the value of a data scientist is limited. And so we actually need data stewards in order to be able to unlock the data that resides elsewhere, or that can unlock the data they have for other purposes as well. And we need to start thinking about different kinds of data collaborative models. We also have to start thinking about how do we lower the transaction costs for accessing data that can be reused by, for instance, coming up with data sharing agreements that everyone accepts and can be reused for other purposes as well. And that leads me to my seventh observation, and I think this is my last one for today, which is that if we are starting to think about reusing data, that also means that we actually need to know something about the data before we know whether it is reusable and whether it is relevant for the question we seek to answer that can inform a decision.

[42:59](https://youtu.be/REhFRhpl9Wk?t=2579) And that means that metadata has become more important than ever, because metadata describes, of course, what the data is all about. It is really about how one can understand the quality of the data, the potential other purposes of the data, and also it understands who in the data lifecycle was responsible for the data as well. And so metadata is actually increased in importance, and we also have seen that it allows us to do data tagging, for instance. We get asked quite a lot to do data tagging in terms of sensitivity so that, for instance, when you start sharing data, you actually have an understanding of the potential risks associated to the data, potential biases that are existing in certain kind of data, because as already indicated, data is never wrong. It's always some kind of a construct, and metadata can actually provide you insights in all those aspects as well.

[44:03](https://youtu.be/REhFRhpl9Wk?t=2643) We also have been advocating that actually metadata needs to be innovated. We really need to start thinking about the traditional ways for data so that we understand, right, if we start using data and reusing data for different kinds of menus, different kinds of questions, we really need to understand what is the quality of the data and to what extent are we complicating or are we adding additional kind of biases if we start, for instance, using that data for other purposes as well. So these are the seven observations to get us started, and I hope those seven observations gives you a mindset to think about data as it relates to decision making. First of all, really start thinking beyond the narrow field, what we call data in the past, which was quite often structured data, which was quite often data collected for one purpose, quite often data that was the result of metadata. So, we have a lot of data assets, such as surveys.

[45:04](https://youtu.be/REhFRhpl9Wk?t=2704) Now, digital has really improved and increased the way we collect data. It also changed what we mean by data. And it also has created new data assets that really can provide for better insights, almost in real time, as it relates to certain phenomenon. But don't think data as any kind of a product, right? It is a process where multiple decisions are being made. And you really, if you start using data, have to be aware of the process because it determines, to a large extent, the potential user friendliness of the data. But also, it allows you to understand who needs to be contacted along the process for unlocking, for instance, data. Data has no value in itself. It's the question. It's the decision that you seek to inform that determines the data. So, it's really, when we start packing up data, I always say, don't talk about data. Talk about the question first, and then look what data you need, because that's ultimately where the value might be.

[46:07](https://youtu.be/REhFRhpl9Wk?t=2767) There's always some kind of a construct when you start using data. Data reuse is really what we should focus on more, because that's good from a perspective of, instead of creating more data and collecting more data, thinking about what data already resides there that could be used as a proxy for answering my question. I think that's really a new skill set that we need to start thinking about. But, of course, that also means that we need to have access to it. And that's the challenge, is that quite often data that can be reused resides elsewhere. And so, we need to really start thinking about meeting, matching demand and supply. And we also need to start thinking about, before we request data, what do we know about the data? And that's where metadata comes in. So, these are kind of the starting points that I wanted to share in order to open the aperture when we start thinking about data, in order to also set the stage for the following weeks.

[47:08](https://youtu.be/REhFRhpl9Wk?t=2828) And so, next week, I really want to hone in on how do we actually formulate questions? If we know that data can be reused, but that the value of the data resides and is directly linked to the question, how do we go about questioning? And what's kind of a new way of even identifying the questions that matter? And we've done a lot of work in that space, which, as I said, is quite often what I call defining question science. And then the third session will be really about how do we unlock data that could be valuable for my question? How do we actually develop data collaboratives? And how do we establish this kind of data stewardship function? How do we actually build the human infrastructure and the governance infrastructure for accessing new data sources that are now made available as a result of digitalization and other efforts? So, with that, Hanna, this was kind of the introduction I wanted to share with the participants, really as a way to broaden the aperture around those seven observations.

[48:23](https://youtu.be/REhFRhpl9Wk?t=2903) And I think many of those observations might not be new, but I think, I hope that the package of those observations really sets a new mindset to think about data as it relates to particular kind of decision-making processes as well.

**Unknown speaker**

[48:40](https://youtu.be/REhFRhpl9Wk?t=2920) Thank you so much, Stefan, and thank you for laying that out all so clearly and it sets a great foundation for our next two workshops, which I'm really looking forward to. Let me just quickly stop recording as we go into the Q&A. Sorry, just one second.

* * *

Original page: https://innovate-us.org/data-series-an-introduction-to-data-and-its-reuse · Licence: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/), as stated on innovate-us.org