Skip to main content
Search form
  • 00:00

    [MUSIC PLAYING][Delivering Data Science Solutions for Charities &Social Enterprises--DataKind]

  • 00:19

    EMMA PREST: I'm Emma Prest.I'm the executive director at DataKind UK.We are a UK charity, and we're part of the wider DataKindnetwork.I don't have any particular data background.I did a politics and geography degree,and I was working for a variety of NGOs,like Amnesty International.And as part of that work, I was quite

  • 00:40

    EMMA PREST [continued]: interested in freedom of informationand how activists and campaigners gethold of government information, government datato hold those in power to account.And while working on that, I got into the idea of open data,and that governments are increasinglyopening up these data sets.But the problem was that civil society doesn't havethe skills to analyze them.

  • 01:00

    EMMA PREST [continued]: So this felt slightly kind of a wasted opportunity,and so what I suddenly, thought, God,you know, we need to find a way to bringthese kind of analytical skill sets into the charity sector.And I found DataKind, and it felt like a perfect match.[MUSIC PLAYING]So over the past 10 years, a lot of these data science tools

  • 01:22

    EMMA PREST [continued]: and techniques have actually become fairly mainstreamin the private sector.I include artificial intelligencewhen I talk about data science.Most companies now have really big business analytics teamsand now data science teams, and thathasn't happened in the same way in the charity sector,in the social enterprise space.We've worked on 65 data science projects,

  • 01:43

    EMMA PREST [continued]: we have a community of 1,800 data scientists,we have three staff.But then we have 25 volunteers who would scope our projects,they're interviewing onboard new volunteers,they do impact assessment and our continuous improvement.So the organization's really run by its community, whichcreates just this lovely space of peopleare doing it because they're passionate about itand they're not paid to do it.It's not their day job.

  • 02:04

    EMMA PREST [continued]: They're doing it in their spare time.So I think kind of the excitement and enthusiasmthat we experience on a daily basisis probably not what most people experienceon a daily basis in their jobs.We work with a charity or a social enterpriseto identify how data science or AI might

  • 02:24

    EMMA PREST [continued]: be able to benefit them.We have a whole kind of rigorous scoping process nowwhere we look at what the problem is, do wehave enough data to actually answer that problem,what would the impact be on the organization,and what's the organization's data maturity?So there's no point in us building a really sophisticatedalgorithm if they don't have the capacity to then embedthat in their work.And so we go through, often, quitea few months of scoping out to see whether it's a good fit.

  • 02:46

    EMMA PREST [continued]: And it may be that it's not.It may be that they need some really fundamental, basic kindof analysis and that would suffice, and that's fine.We don't think data science is the solution to everything.But in some instances, it really can drive change and reallyhelp an organization have a greater impact.We do a lot of what we call data dive weekends.We'll bring together 60, 70, 80 data scientists into one room

  • 03:11

    EMMA PREST [continued]: with three or four charities.It's like a hackathon-style event,and they will work throughout the weekend findinginsights, building prototypes for the nonprofits.So it's quite an amazing resourcefor these charities who would never normallybe able to afford data scientists, whohave very high pay rates and are hugely in demandin the private sector.So what we do is create a way for these charities

  • 03:32

    EMMA PREST [continued]: to access that skill set.[MUSIC PLAYING]So we just started working with the Business and Human RightsResource Center, who are an international non-profit.They happen, fortunately, to have officesaround the corner from us and theyare a non-profit that looks at human rights abuses,

  • 03:55

    EMMA PREST [continued]: often overseas, that are committedby companies and corporations around the world.And they try to track those allegations,and then track the companies' response to those allegations.[MUSIC PLAYING]

  • 04:12

    JOE BARDWELL: A lot of our work is aroundthe impacts of agriculture and extractive companies,and that's more about land issues.Say making sure communities have the right sort of informationbefore a big project comes into their area and disrupts them.One example is recently we contacteda HQ of a company that had unwittingly

  • 04:33

    JOE BARDWELL [continued]: blocked a road for a community and for over 10 years,they hadn't been able to get proper accessto hospitals and to schools.And just by taking the concerns of the communityand articulating them properly to the international HQ,we were able to get them to build a through-road, whichhas really changed their lives.

  • 04:55

    JOE BARDWELL [continued]: So sometimes it can be that the very small thingsmake a big difference.The information we collect is news reports, civil societyreports, company statements, government legislation.That all helps people understand what's going on in the worldand how businesses are impacting human rights,but also how they can help hold companies accountable

  • 05:18

    JOE BARDWELL [continued]: and enhance human rights in business.We have researchers in 18 countries across the world,and we work in eight languages.And on a daily basis, we track the environmental and socialimpacts of companies around the world,and bring that information into our international platform

  • 05:39

    JOE BARDWELL [continued]: through our website.And we've been doing that for about 15 years,and through that time, we've grownthe largest database on business and human rights issues.But we've never done any large-scale analysisof the trends in our data, and the sort of relationshipsbetween, for example, a sector and a type of human rightsissue.

  • 06:02

    JOE BARDWELL [continued]: We've worked with DataKind to come up with a waythat we can analyze our data at large-scaleand pull out some of those insights.We know from talking to investors, local civil societygroups where businesses operate, that they need accessto information quickly about the track records of companies.

  • 06:22

    JOE BARDWELL [continued]: The data part of it comes in when our researchers collecta piece of information, they tag it with the company name,the location, the sector, and that allows us to pull outbroad trends from our data.We've been in touch with DataKind for about two years,discussing how we might be able to put together

  • 06:43

    JOE BARDWELL [continued]: a project that lets us start exploring the datathat we've now collected.Until recently, we weren't able to actually gethold of that data in any sort of large way to allow analysis.But having built an API recently,that's really opened up doors for usto bring in data scientists and look at how we might

  • 07:04

    JOE BARDWELL [continued]: examine the trends in our data.And ultimately, look at how we visualize the information wehave so that people coming to our website and the peoplethat we meet on the ground can get insights veryquickly from our information.[MUSIC PLAYING]

  • 07:22

    EMMA PREST: What we're looking at doingfor them is how they can better use this data to work outwhere are risky places for companies to expand into,to invest into.So are there certain kinds of human rightsabuses, certain kinds of wrongdoingthat are more likely to happen in certain countries,in certain industries?So we're doing a first take analysison this really quite meaty data set

  • 07:44

    EMMA PREST [continued]: all about corporate wrongdoing.[MUSIC PLAYING]We'll do a data dive with them just using exploratory analysisto understand more about what's possible to do with this data,and then we might be looking to do a longer termproject with them to build a data science model.But we obviously would need to scope that out more.We're just in the early stages, it justkicked off a few weeks ago.

  • 08:04

    EMMA PREST [continued]: So right now, there are a team of data ambassadorswho are leading them through the whole prep phase,are going to be kind of getting their heads around the data,working out what prep they need to do to get the data in shapefor the data dive weekend.

  • 08:17

    ALICE JACQUES: We've done lots of interpretinggraphs trying to pick out the story from the data.Because at this point of the day,you've usually got like loads and loads of insights,a million graphs, and you've got to try and pick outthe ones that tell the story, that the charity can take backto their investors and present.Everybody's been working in Python or R.All sorts of data cleaning, taking out null values,

  • 08:39

    ALICE JACQUES [continued]: that sort of thing.And then it ranged from the very, very simpleproducing percentages and bar graphsthrough kind of naive based simple classification models.Some people have done some incredible stuffwith tableau and beautiful PowerPoint presentation typestuff.And someone's even built a little neural networkin a day and a half.[MUSIC PLAYING]

  • 09:03

    JOE BARDWELL: Working with DataKindhas been fantastic because not only do theyunderstand data science, but they understandsome of the limitations and challengesthat small charities face when they're trying to takeon these sorts of projects.We've been working with three data ambassadors whoare all volunteer.And they're data scientists in their day job

  • 09:24

    JOE BARDWELL [continued]: and they volunteer to help us.And we've been looking at the questionsthat we want to answer from our data,and really digging down into what thatmeans at a technical level.And they've also been doing the workof flattening out our data, which is a very complex dataset.Which these are simply things that we would nothave been able to do ourselves.So it's been extraordinarily helpful.

  • 09:49

    EMMA PREST: We have a fairly steady stream of organizationslooking for help, but I think it's fair to say a lot of themneed help with some fairly simple kind of spreadsheetanalysis.So we now run a whole stream of workwhich we call data therapy.So we have a monthly office hourswhere charities can drop in, get some advice with data experts,sit down for an hour and just look at their spreadsheets.

  • 10:11

    EMMA PREST [continued]: We have something called the social data society.So all the data scientists and data analystwe found in UK charities, we've kind of brought them togetherinto a peer-to-peer group where they meet every six weeks,and they're supporting each otherwith some of their problems.So we've been quite kind of creativein how we're trying to think about levels of support,since we know that not every charity needs

  • 10:32

    EMMA PREST [continued]: a full-on predictable model in data science solution.There's a huge array of challengesthat we've come across in the social change sector.So if the model is wrong and if the model isn't updated,and actually it's producing results--which often you can--the person on the ground would be able to tell that looks off.

  • 10:55

    EMMA PREST [continued]: You stop using it immediately.You rewind.But also the point here is you neverwant the model to make the decisions.You want the model to be advising a human being.So the idea of a lot of these modelsis that it's crunching historic data,it's able provide fresh insights to someone making a decisionwhere they wouldn't be able to know all thousands of cases

  • 11:16

    EMMA PREST [continued]: in the past.It in some way summarizes that and says,this person is more likely to x.But the human being standing in front of that personuses their own judgment and doesn't blindlyfollow what the model says.So that's the kind of thing we'retrying to get people to think about,especially our nonprofit partners, when they'reusing these kinds of tools.

  • 11:37

    EMMA PREST [continued]: I think this idea that because something is a black boxalgorithm and is automated, it is somehow a machine,it is therefore impartial, is not true.These algorithms are built by humans, and in many cases,they're built by a group of often middle class, white men

  • 11:58

    EMMA PREST [continued]: in California sitting around a table with a certain ideaof what the world is.And that's a problem, because youdon't have thinking from different perspectives.Data is generally been historically collected,and societies' biases are deeply embedded in it.And if you blindly put that into a data science model,

  • 12:20

    EMMA PREST [continued]: you will be just carrying those biases through to the results.And if anything, making them worse because in theory,you're scaling up the whole process.And so what we've been doing at DataKindis trying to think sensibly and create the spaceto reflect what we're doing in these projects,and think about the possible negative consequences,

  • 12:41

    EMMA PREST [continued]: and trying to get our charity partners to think criticallyabout what they're doing and why and what the data is.It's a huge topic and a massively growing fieldright now, as to understand what does fairness meanin artificial intelligence because we're all humanand we all think different things are fair or unfair.

  • 13:03

    EMMA PREST [continued]: So it's really hard.It's a really hard, I would say, philosophical topic, as well asactual technical problem to deal with.And so what we're trying to do with DataKindis just talk out loud and learn out loudas we go about this stuff.There's this whole new trend, this wave right nowof how do we make sure we're building

  • 13:23

    EMMA PREST [continued]: cross-disciplinary teams that can really think throughabout what we're building?And then beyond that, if we are building black box algorithms,how do we make them explainable?To a certain extent, they aren't explainable.The data scientists themselves don't always know quitehow the algorithms work.But we still need to be able to communicate,

  • 13:44

    EMMA PREST [continued]: I think, at a high level, at a layman's levelwhat is going on, what were the variables used,what are the results that came out of it?

  • 13:52

    JOE BARDWELL: The role of data science for the Businessand Human Rights Resource Center in the future is pivotal.We have a strategy around using datato gain insight and make sure that there is informationaccessible on how companies are impacting human rights.And I think it's an indication of how important datais in broader society.

  • 14:14

    JOE BARDWELL [continued]: That is what we're doing.I mean, one of the reasons why we'redoing this is because the demand for our data and insightsfrom our data is greater.We're also looking at ways that wecan use other technologies, like machine learning,to strengthen our data set and make it future proof.

  • 14:33

    EMMA PREST: There's a huge amountthat my job is rewarding.DataKind is built off of warm and fuzzies.So if you think that we are a community of peoplewith a niche expertise who just are hungry to do somethingbeneficial for society.

  • 14:51

    JOE BARDWELL: The most exciting thingabout this job for me is working with people with a completelydifferent skill set to look at how we can bring ideastogether and learn from each otherto really scale up the work at the Business and Human RightsResource Center, and ultimately improve people'slives across the world.

Video Info

Publisher: SAGE Publications Ltd

Publication Year: 2019

Video Type:In Practice

Methods: Data science, Data linkage, Data cleaning

Keywords: accountability (business); challenges, issues, and controversies; charities; data analysis; data classification; data processing; data visualisation; human rights; human rights abuses; international business; peer-to-peer; Social enterprises; Social impact; technology ... Show More

Segment Info

Segment Num.: 1

Persons Discussed:

Events Discussed:



Emma Prest, Executive Director at DataKind UK, and Joe Bardwell, Senior Officer at the Business and Human Rights Resource Centre, discuss how working together has resulted in opportunities for in-depth analysis of a large volume of historical and contemporary data using previously unavailable methodologies and technology.

Looks like you do not have access to this content.

Delivering Data Science Solutions for Charities & Social Enterprises: DataKind

Emma Prest, Executive Director at DataKind UK, and Joe Bardwell, Senior Officer at the Business and Human Rights Resource Centre, discuss how working together has resulted in opportunities for in-depth analysis of a large volume of historical and contemporary data using previously unavailable methodologies and technology.

Copy and paste the following HTML into your website