WEBVTT 00:00:04.460 --> 00:00:09.985 Hi, my name is Stian Westlake and I'd like to talk today about 00:00:09.985 --> 00:00:15.510 the idea of a data moon shot for better and more prolific data 00:00:15.510 --> 00:00:19.335 collaboration. I've got three particular interests in this. 00:00:19.335 --> 00:00:24.435 Not long ago I wrote a book about the intangible economy and what 00:00:24.435 --> 00:00:29.110 needs to make it work. I recently worked as a government 00:00:29.110 --> 00:00:32.935 adviser on science innovation policy, but most importantly I 00:00:32.935 --> 00:00:36.335 recently joined the Royal Statistical Society as their 00:00:36.335 --> 00:00:40.470 chief executive. I'm going to first of all kick off by talking 00:00:40.470 --> 00:00:44.322 a little bit about the economics of data and how that affects 00:00:44.322 --> 00:00:48.495 what we might do. If you really don't like economics, this is a 00:00:48.495 --> 00:00:52.026 great time - this being a video conference to pop out and make 00:00:52.026 --> 00:00:55.878 yourself a cup of tea. I will probably only take about 2 mins 00:00:55.878 --> 00:01:00.051 so it will probably be a weak cup of tea, but you're forewarned 00:01:00.051 --> 00:01:04.420 The story that I tell in 'Capitalism without capital' 00:01:04.420 --> 00:01:08.932 is a story of how the world economy once-upon-a-time 00:01:08.932 --> 00:01:13.444 the capital in that economy, the things we used to invest in 00:01:13.444 --> 00:01:17.956 used to be physical things you could see and touch like machines 00:01:17.956 --> 00:01:22.468 or buildings or plants. That used to represent about 15% of world 00:01:22.468 --> 00:01:23.596 GDP per year. 00:01:23.630 --> 00:01:28.796 It was the sinues of the economy 40 years ago. Over the last 40 00:01:28.796 --> 00:01:33.224 years that's changed and what now makes the world economy work is 00:01:33.224 --> 00:01:36.914 intangible assets. Things you can't see or touch like R&D, 00:01:36.914 --> 00:01:39.866 software, brands, supply chains and of course, data. 00:01:40.520 --> 00:01:44.194 Data and other intangible assets are kind of funky from an 00:01:44.194 --> 00:01:47.200 economic point of view. They behave differently from other 00:01:47.200 --> 00:01:50.540 types of investments for two reasons that are going to be 00:01:50.540 --> 00:01:54.214 important to us. One is that they have spillovers, so if 00:01:54.214 --> 00:01:58.556 you're a business and you invest in some data, and that data is 00:01:58.556 --> 00:02:02.849 widely published. Other people may benefit from it as well as 00:02:02.849 --> 00:02:05.828 you. You can't... your competitors may benefit from it 00:02:05.828 --> 00:02:10.131 There is a kind of challenge there if you're relying on self 00:02:10.131 --> 00:02:13.441 interested businesses to do your investment, that won't get you 00:02:13.441 --> 00:02:16.420 all that way. The other thing about intangible assets 00:02:16.420 --> 00:02:19.730 generally in data in particular is they have what economists 00:02:19.730 --> 00:02:23.040 like to call synergies. They're really good when you combine 00:02:23.040 --> 00:02:26.350 them together, which is exactly why data collaboration is so 00:02:26.350 --> 00:02:29.660 important. If you have a little bit of information - information 00:02:29.660 --> 00:02:32.639 about a small number of customers or a 00:02:32.639 --> 00:02:33.963 small amount of epidemiological 00:02:33.963 --> 00:02:37.394 information it's not hugely useful. If you can combine that 00:02:37.394 --> 00:02:40.980 with lots of similar data. Or lots of different data. 00:02:40.980 --> 00:02:44.240 Suddenly those things can become much more valuable, and that's a 00:02:44.240 --> 00:02:46.848 general property of these intangible assets. So we're 00:02:46.848 --> 00:02:50.434 moving from an economy where you could just build your factory 00:02:50.434 --> 00:02:54.346 and get on with things to an economy where the capital that 00:02:54.346 --> 00:02:57.932 we all depend on has these characteristics that it will be 00:02:57.932 --> 00:03:01.844 under produced if you just rely on businesses to do their thing, 00:03:01.844 --> 00:03:05.104 and it really matters how you combine it, those synergies. 00:03:05.190 --> 00:03:08.680 Into this kind of problem, typically steps government and 00:03:08.680 --> 00:03:12.868 when there are investments that have a lot of 00:03:12.868 --> 00:03:15.660 these so called spillovers, you typically expect 00:03:15.660 --> 00:03:19.848 government to step up to the plate and invest in some of them 00:03:19.848 --> 00:03:23.640 And that's what's traditionally happened with, for example R&D 00:03:23.640 --> 00:03:27.432 The British government funds about £10 billion worth of R&D 00:03:27.432 --> 00:03:29.271 a year, business backs 20 billion. 00:03:29.271 --> 00:03:33.110 If government didn't do that, funding the UK economy would be 00:03:33.110 --> 00:03:35.204 doing even less well than it 00:03:35.204 --> 00:03:38.866 currently is. If we think about something like education and 00:03:38.866 --> 00:03:41.365 training, another really important intangible, in Britain 00:03:41.365 --> 00:03:45.649 the government spends £90 billion a year on that. So in some 00:03:45.649 --> 00:03:49.219 areas our politicians, our political culture, have got the 00:03:49.219 --> 00:03:52.432 idea that investing in these intangibles is something really 00:03:52.432 --> 00:03:55.645 important to do. Something that is worth spending taxpayers money 00:03:55.645 --> 00:03:58.144 on if we want a thriving economy 00:03:58.144 --> 00:04:02.782 and society. But that kind of message hasn't quite got through 00:04:02.782 --> 00:04:06.362 that much on data. Now, obviously there is a tradition 00:04:06.362 --> 00:04:09.584 of governments publishing and providing data. Things like the 00:04:09.584 --> 00:04:12.090 Ordinance Survey and the tradition of National Statistics 00:04:12.090 --> 00:04:16.386 are really old. In some cases centuries old, but the scale at 00:04:16.386 --> 00:04:20.682 which that goes on the scale at which government sees it being 00:04:20.682 --> 00:04:24.262 worth publishing data is still pretty small compared to the 00:04:24.262 --> 00:04:27.842 funding of General Research or let alone funding education. And 00:04:27.842 --> 00:04:29.632 it's something that isn't quite 00:04:29.690 --> 00:04:33.722 in the mindset how we talk about politics. So when people talk 00:04:33.722 --> 00:04:38.090 when we talk about the purpose of the money that we spend on 00:04:38.090 --> 00:04:40.778 say the Office for National Statistics or providing health 00:04:40.778 --> 00:04:44.474 data, we see it from a functional point of view. But 00:04:44.474 --> 00:04:48.170 often we lack the bigger picture about how real investment in 00:04:48.170 --> 00:04:51.866 data and really making that data sing can have a transformative 00:04:51.866 --> 00:04:56.570 effect on society. So I guess my pitch when I talk to people in 00:04:56.570 --> 00:05:00.602 politics or to anyone who is a citizen and who cares about 00:05:00.602 --> 00:05:05.768 these things is to say that if you kind of want to follow the 00:05:05.768 --> 00:05:09.408 current UK's government advice that we need to build, build, 00:05:09.408 --> 00:05:13.776 build, then data is a really important place to start. And if 00:05:13.776 --> 00:05:17.416 we're thinking of moonshots, if we're thinking of big, ambitious 00:05:17.416 --> 00:05:21.056 technical projects that can improve the world, data is a 00:05:21.056 --> 00:05:25.424 neglected area. But it's an area that deserves a lot more love 00:05:25.424 --> 00:05:29.428 and a lot more credit, and probably a lot more public 00:05:29.428 --> 00:05:30.520 funding and government 00:05:30.520 --> 00:05:36.650 attention. One of the places where the RSS has 00:05:36.650 --> 00:05:40.500 been vocal about this before is through our data manifesto. 00:05:40.500 --> 00:05:44.350 Our kind of manifesto for what should happen in the data world 00:05:44.350 --> 00:05:48.200 We published the most recent version last year, and this was 00:05:48.200 --> 00:05:52.050 kind of a product of the input of our many members who are 00:05:52.050 --> 00:05:53.800 statisticians and who work with 00:05:53.800 --> 00:05:58.045 data. And it made, in my opinion, some really important 00:05:58.045 --> 00:06:02.170 arguments that I think should form part of what you 00:06:02.170 --> 00:06:04.795 might call a a data moonshot. 00:06:05.560 --> 00:06:09.688 The first and kind of most obvious thing is this idea that 00:06:09.688 --> 00:06:13.816 we should be investing much more in our data infrastructure as a 00:06:13.816 --> 00:06:17.600 country in the same. We're very happy at the moment. Or 00:06:17.600 --> 00:06:20.696 governments are very happy to spend billions of pounds 00:06:20.696 --> 00:06:23.792 producing academic research papers, some of which are really 00:06:23.792 --> 00:06:27.920 important, some of which perhaps maybe only get read a few dozen 00:06:27.920 --> 00:06:31.360 times and are sealed away in an academic journal somewhere. 00:06:31.410 --> 00:06:36.935 We should be thinking much more about how we invest 00:06:36.935 --> 00:06:38.210 in data in the same way. 00:06:38.210 --> 00:06:42.060 There's a really political message here. If you look at something 00:06:42.060 --> 00:06:45.210 like Anna Powell-Smith's missing numbers project or her thoughts 00:06:45.210 --> 00:06:48.360 about the government data graveyard, there are loads of 00:06:48.360 --> 00:06:51.914 things that Governments - the UK government - should be 00:06:51.914 --> 00:06:54.866 informing citizens about should be holding itself to 00:06:54.866 --> 00:06:57.449 accountable to account for which simply aren't 00:06:57.449 --> 00:07:01.139 published. So I think the first thing we 00:07:01.139 --> 00:07:02.615 would call for is. 00:07:04.180 --> 00:07:07.022 Whenever government does something, whenever the law is 00:07:07.022 --> 00:07:10.676 changed. Whenever something important is going on that we 00:07:10.676 --> 00:07:15.954 should be pushing for data to be made available. Data to be made 00:07:15.954 --> 00:07:17.172 open about that. 00:07:18.280 --> 00:07:23.376 But I think this is not just a kind of government 00:07:23.376 --> 00:07:25.924 proberty in Government accountability issue. This is 00:07:25.924 --> 00:07:29.200 also about producing bigger datasets that other people can 00:07:29.200 --> 00:07:32.112 use and make valuable research contributions to. 00:07:32.112 --> 00:07:36.480 If I think about things which you know I would call 00:07:37.010 --> 00:07:39.374 moonshots in a sense. Really important infrastructural data 00:07:39.374 --> 00:07:43.314 projects that create lots of new opportunities. I think of 00:07:43.314 --> 00:07:48.042 something like if I were in Leeds for this event today down 00:07:48.042 --> 00:07:52.376 the road in Bradford the Act Early North Project which brings 00:07:52.376 --> 00:07:55.922 together an absolutely unique set of data on early year's 00:07:55.922 --> 00:07:59.468 performance in Bradford, the youngest city in the UK. 00:07:59.520 --> 00:08:04.380 Which is going to be an absolute treasure trove for public health 00:08:04.380 --> 00:08:07.620 researchers that anyone who cares about tackling deprivation 00:08:07.620 --> 00:08:11.265 and child poverty, and more generally about improving the 00:08:11.265 --> 00:08:15.720 world. That was, I'm told by the people who run it 00:08:16.520 --> 00:08:20.502 incredibly hard to get funding for because it didn't look like 00:08:20.502 --> 00:08:24.484 a real academic project, but it wasn't a public health project. 00:08:24.484 --> 00:08:28.828 Thanks to some farsighted work by the council and local NHS. 00:08:28.828 --> 00:08:32.810 It did get funding, it cobbled it together and it created 00:08:32.810 --> 00:08:36.792 something really impressive. But we have a kind of system where 00:08:36.792 --> 00:08:40.412 funding academic papers, as Ben Goldacre one of the earlier 00:08:40.412 --> 00:08:43.308 speakers today said funding academic papers is easier, 00:08:43.308 --> 00:08:46.566 funding data tools is harder, so I think pushing 00:08:46.610 --> 00:08:50.510 on those kind of big projects is really important. Another 00:08:50.510 --> 00:08:54.410 example from my time in government. I was involved in the 00:08:54.410 --> 00:08:57.920 design of an R&D project called the Industrial Strategy 00:08:57.920 --> 00:09:02.600 Challenge Fund and the idea of that was to put together pretty 00:09:02.600 --> 00:09:04.940 big collaborations between business and researchers, 00:09:04.940 --> 00:09:08.840 typically in the kind of hundreds of millions of pounds 00:09:08.840 --> 00:09:11.816 range. The idea was, they would improve society. They would 00:09:11.816 --> 00:09:15.358 generate economic growth and would be a good thing for the 00:09:15.358 --> 00:09:20.776 government to do. Lots of the bids for those projects were 00:09:20.776 --> 00:09:23.940 from kind of industrial companies, people who 00:09:23.940 --> 00:09:28.008 manufacture things which are all really valuable. There were few 00:09:28.008 --> 00:09:32.528 to none that were about building really valuable data resources 00:09:32.528 --> 00:09:36.596 whether that's digital twins, public health data sets. 00:09:36.630 --> 00:09:42.077 And I think this is an area where we really want to see 00:09:42.077 --> 00:09:45.848 governments using more, being more inspired and thinking 00:09:45.848 --> 00:09:50.038 about how they can invest and work with businesses, Local 00:09:50.038 --> 00:09:53.809 government to generate really valuable data resources 00:09:53.809 --> 00:09:55.485 and to be more ambitious. 00:09:56.800 --> 00:10:01.966 Those points all kind of come back to this economic idea 00:10:01.966 --> 00:10:05.526 of spillovers. The reason why we want government spending money on 00:10:05.526 --> 00:10:09.442 this is because the private sector will invest to a certain extent, 00:10:09.442 --> 00:10:13.002 but there are public benefits 00:10:13.002 --> 00:10:14.782 these data assets are public goods. 00:10:14.850 --> 00:10:17.618 But I also mentioned that intangibles and data 00:10:17.618 --> 00:10:20.732 specifically has this kind of idea of synergies. They're 00:10:20.732 --> 00:10:24.192 especially valuable when you combine it, and I guess there 00:10:24.192 --> 00:10:28.344 the call to action for governments is to say, well, how 00:10:28.344 --> 00:10:31.804 do we make those those ideal combinations happen? 00:10:31.804 --> 00:10:34.918 How do we create a world where collaboration works well, 00:10:34.918 --> 00:10:38.724 happens a lot and is positive? I guess there's kind 00:10:38.724 --> 00:10:42.876 of a few really interesting things that I look at that I 00:10:42.876 --> 00:10:45.298 think we ought to be doing more 00:10:45.298 --> 00:10:49.494 of. One is the affective and privacy respectful merging of 00:10:49.494 --> 00:10:53.554 datasets. So the work that the Oxford evidence based medicine 00:10:53.554 --> 00:10:58.426 data lab have been doing on the open safely platform, which has 00:10:58.426 --> 00:11:02.080 seemed to be providing some really valuable insights into 00:11:02.080 --> 00:11:06.546 the progress of the covid epidemic would be one example of 00:11:06.546 --> 00:11:08.982 that. How do you bring together 00:11:10.560 --> 00:11:14.648 extremely confidential medical information about people in a 00:11:14.648 --> 00:11:19.758 way that you can analyze it with respect for confidentiality. 00:11:19.760 --> 00:11:23.764 The idea that this is a really valuable investment that our 00:11:23.764 --> 00:11:26.676 research funders, that our government, should be backing 00:11:26.676 --> 00:11:30.316 seems really important if we want to make those synergies 00:11:30.316 --> 00:11:33.228 work, that's absolutely essential. I guess 00:11:33.228 --> 00:11:37.960 the second thing, and you know ODI Leeds is a great place to 00:11:37.960 --> 00:11:41.964 be saying this is more hubs and programs for building networks 00:11:41.964 --> 00:11:45.968 to share data. I am a longstanding fan of the work 00:11:45.968 --> 00:11:49.972 that ODI Leeds has been pioneering in this area and I 00:11:49.972 --> 00:11:54.663 was thrilled that they did some work with me when I was in 00:11:54.663 --> 00:11:58.620 government. I think ODI Leeds have really blazed a trail here. 00:11:58.620 --> 00:12:02.220 I think other projects like data kind UK that brings together 00:12:02.220 --> 00:12:05.460 data scientists and charities who often have under utilized 00:12:05.460 --> 00:12:09.780 data sources or don't know how to collect the right data is 00:12:09.780 --> 00:12:10.860 another great example. 00:12:11.590 --> 00:12:16.348 So the idea of backing hubs of creating more of them 00:12:16.348 --> 00:12:20.374 seems like something that maybe it looks very touchy feely, but 00:12:20.374 --> 00:12:24.400 it's hugely important if we want more collaboration in this area 00:12:24.400 --> 00:12:29.158 and then I guess the issue which which we've heard a bit about 00:12:29.158 --> 00:12:33.184 already today is how data ethics and standards interact with all 00:12:33.184 --> 00:12:35.746 of this and I think if I'm 00:12:36.380 --> 00:12:39.684 pitching this message at government. The most valuable 00:12:39.684 --> 00:12:44.640 thing that they should be looking at is how do we create 00:12:44.640 --> 00:12:48.357 an further and codify the results of a national 00:12:48.357 --> 00:12:52.074 conversation on data ethics. Thinking with my kind of 00:12:52.074 --> 00:12:54.139 General Science policy hat on? 00:12:54.720 --> 00:12:59.436 This is something that there are success stories in the past in 00:12:59.436 --> 00:13:03.366 other fields, so if we think about embryology for example, 00:13:03.366 --> 00:13:08.082 which has been and in many places the world still is an 00:13:08.082 --> 00:13:11.619 incredibly controversial field, the UK government in the 1990s 00:13:11.619 --> 00:13:14.763 was actually pretty forward thinking there and really 00:13:14.763 --> 00:13:17.907 fostered a national conversation. Got ethicists involved in some 00:13:17.907 --> 00:13:21.051 cases funded research, but also brought together discussions 00:13:21.051 --> 00:13:24.981 with the result that it created a relatively stable consensual 00:13:25.050 --> 00:13:28.757 ethical framework for how that ran, which was 00:13:28.757 --> 00:13:33.138 kind of to the advantage of the UK, not just in preventing bad 00:13:33.138 --> 00:13:36.508 things, but actually in giving stability so that good things 00:13:36.508 --> 00:13:37.856 could be done. 00:13:37.870 --> 00:13:42.381 Clearly there's been a lot of talk and a lot of investment in 00:13:42.381 --> 00:13:46.198 organizations that look at greater ethics. But this is, I 00:13:46.198 --> 00:13:50.362 think, something that we just need to keep on pushing on and 00:13:50.362 --> 00:13:51.403 keep on supporting. 00:13:52.170 --> 00:13:57.020 The other side of this is perhaps the international side, 00:13:57.020 --> 00:14:01.385 and you know this conference, the collaboration between the 00:14:01.385 --> 00:14:05.750 Netherlands and the UK is a really great example. 00:14:05.760 --> 00:14:09.912 Great to see the Netherlands reaching out to us in this way. 00:14:09.912 --> 00:14:13.372 I would hope that the UK government could start reaching 00:14:13.372 --> 00:14:17.178 out as well and playing a more active role in international 00:14:17.178 --> 00:14:21.676 efforts to set standards. I think as a British person that 00:14:21.676 --> 00:14:24.790 something where a relatively small investment could yeild big 00:14:24.790 --> 00:14:30.596 benefits both for the world as a whole and for Britain specifically. 00:14:30.596 --> 00:14:32.730 And then I guess. 00:14:33.750 --> 00:14:38.154 A further dimension to this I was in the briefing for this 00:14:38.154 --> 00:14:42.558 session I was asked how we could make data work for the 00:14:42.558 --> 00:14:47.696 many, not for the few, but when I think about I think perhaps we 00:14:47.696 --> 00:14:52.834 should turn that on its head and say how can we make data work 00:14:52.834 --> 00:14:56.871 for minorities wherever they may be in society rather than just 00:14:56.871 --> 00:15:00.908 data that serves the interest of majority, and particularly at a 00:15:00.908 --> 00:15:04.211 time when the Black Lives Matter protest have made 00:15:04.280 --> 00:15:10.256 systematic discrimination so salient in society the need for 00:15:10.256 --> 00:15:16.232 granular specific data that cast light on discrimination and 00:15:16.232 --> 00:15:20.880 highlights injustices seems more urgent than ever. 00:15:20.890 --> 00:15:23.179 I think this is an area where 00:15:24.050 --> 00:15:28.430 there've been historical, huge oversights in the part of public 00:15:28.430 --> 00:15:32.810 data that we've been collecting, and obviously there is a 00:15:32.810 --> 00:15:36.314 separate issue about the potential for discrimination by algorithms 00:15:36.314 --> 00:15:41.132 that we use. This strikes me as a huge opportunity where the 00:15:41.132 --> 00:15:45.950 right kind of intervention could make a huge difference. 00:15:45.950 --> 00:15:51.073 So I guess where I would... it seems to me really 00:15:51.073 --> 00:15:55.726 clear that there is a moon shot opportunity on data, an 00:15:55.726 --> 00:16:03.225 opportunity for us as a society to invest more, to build more of 00:16:03.535 --> 00:16:08.106 a collaborative around this, and to build norms 00:16:08.106 --> 00:16:11.813 where we use data well and come together effectively around it. 00:16:11.813 --> 00:16:15.857 I've talked about this to some extent for a public policy point 00:16:15.857 --> 00:16:19.901 of view, but the public policy is not going to write itself. 00:16:19.901 --> 00:16:23.945 The politicians are not going to do this themselves for all the 00:16:23.945 --> 00:16:27.652 fine words that occasionally we hear about how they really like 00:16:27.652 --> 00:16:30.685 Bayesian statistics and Monte Carlo simulations. To quote one 00:16:30.685 --> 00:16:34.392 Minister from last weekend, this needs a movement and it will 00:16:34.392 --> 00:16:37.088 essentially have to be a movement comprised of 00:16:37.180 --> 00:16:41.886 people like you who work with data who know data and who know 00:16:41.886 --> 00:16:46.230 the contribution that it can make to a good society. From that 00:16:46.230 --> 00:16:50.212 point of view, it is really fortunate that data analysis and 00:16:50.212 --> 00:16:54.194 data analysts are cool at the moment, are in demand, whether 00:16:54.194 --> 00:16:57.814 that's in business or in government or in wider society. 00:16:57.814 --> 00:17:02.158 Because it gives you and gives us a platform to make those 00:17:02.158 --> 00:17:05.416 demands to point out how investment is really important, 00:17:05.416 --> 00:17:09.760 how this needs to be done in the right collaborative and the 00:17:09.760 --> 00:17:10.846 right ethical framework. 00:17:10.860 --> 00:17:18.462 And the time to do this is very much now, so this is a 00:17:18.462 --> 00:17:21.177 chance to make your voices heard. 00:17:21.177 --> 00:17:24.048 To really drive 00:17:24.950 --> 00:17:29.390 these moonshot projects that can change the world and to push 00:17:29.390 --> 00:17:30.870 for better understanding and 00:17:30.870 --> 00:17:34.659 more collaboration. So I think this is exciting to be 00:17:34.659 --> 00:17:38.182 here and I look forward to seeing what we can do next.