,
84 tweets,
40 min read
Read on Twitter
THREAD 1/ I'll be tweeting from the launch of @ODIHQ #DataTrusts's report. theodi.org/article/odi-da… 
1a/ btw I should say, in this thread I'm going to be focusing on areas that look at the governance of data/ #DataTrusts and public engagement in the collection, analysis, use and sharing of data
4/ Aim, says @sanakb, is to reduce friction in sharing data to help ensure the benefits are gained. > however, I suspect some frictions are useful in slowing work down allowing more time for trust to be built. #DataTrusts
5/ Focus is on unlocking benefits from data. However, #DataTrusts, if they are going to build trust are going to also have to grapple with and reduce the risks and harms that arise from the collection, use, etc of data
6/ The link to the live stream of #DataTrusts launch is here theodi.org/event/could-da… ht @EinsteinsAttic
7/ @sanakb highlighting the importance of diversity in the #AI industry. < As with other areas of government, this will also increase the connection to the wider social issues raised by the tech, lowering the need for formal public engagement and making it easier #DataTrusts
8/ [ME] This point seems critical to me. Formal public engagement is so critical right now because government, officials and representatives are undiverse and unrepresentative of the populations they serve #DataTrusts
9/ @sanakb - If we can bring government, industry and public together around a shared vision then the prizes are huge > this notion of a shared vision is critical and we can only do it by engaging widely #DataTrusts
10/ next up is @peterkwells, he's going to be giving background. Suspect you'll be better off reading the report rather than me live tweeting this theodi.org/article/odi-da… #DataTrusts
11/ Key take away for me from @peterkwells for my agenda (though all really interesting in general) is that #DataTrusts need to be thought about as part of the infrastructure for delivering the benefits, but the development of infrastructure itself is itself a political act
12/ Next up we have people representing the three pilots into 1. City Data, 2. Food Waste, 3. Illegal Wildlife Trade on a panel to look at the specifics of the questions #DataTrusts
13/ @EdParkes on the City data pilot who worked with Greenwich Council and the LGA to look at #DataTrusts in the context of existing data sharing processes and the development of smart cities. @involveUK and @commcham worked with @ODIHQ on this.
14/ focus of pilot was to promote electric vehicle use, and how to use data to increase this. Part of it is to provide charging points and use sensors to make sure that only vehicles charging are using those spaces. Want to pull this data onto London City data store #DataTrusts
15/ other aspect was smarter heating of homes, collecting data to make buildings more efficient, to predict demand. In the end it will come down to trust of resident, citizen to allow sensors to be put into homes so that data can be collected #DataTrusts
16/ one learning was that these questions may be more pertinent further down the line, rather than right now. #DataTrusts
17/ because we didn't work on it, and largely to save your timeline, I'm not going to tweet the background on the other two pilots on wildlife trade and food waste, again, see the ODI #DataTrusts report for this.
18/ Here are the three pilot backgrounds - cities theodi.org/?post_type=art… - food waste theodi.org/article/data-t… - wildlife trading - theodi.org/article/data-t… #DataTrusts
19/ If my tweet thread sparks questions you can ask them here app2.sli.do/event/a7zi4qwo… #DataTrusts
20/ We're now to questions, again, I'll report on questions relevant to the governance and public engagement points #DataTrusts
21/ questions quite sceptical that #DataTrusts are that new, indeed the trusts (not sure if all) aren't going to be developed on a legal basis, so what about what we know about data collaboratives, for eg.
22/ and scepticism that the purpose of #DataTrusts slipping, moving from protecting individuals' rights to building trust to help make it possible to share data
23/ next up to talk about decision-making processes. Up now are @buntms and @involveUK's @suzlansdell #DataTrusts
25/ for a more traditional form of data sharing the decision-making processes are relatively simple, for many reasons, perhaps because there are many different data providers, varied types of data etc and not all known at start of setting up sharing agreement #DataTrusts
26/ #DataTrusts will need to balance needs of data users and providers (in way that more simple arrangements will only need to worry about providers)
27/ given the ambiguity of who will be involved, the types of data etc, clear need for trust which will be founded on a clear, shared purpose and values to guide the decisions of #DataTrusts
28/ and all of this needs time to get right, because ambiguity about purpose and the promise of the role #DataTrusts will play will lead to problems later.
29/ now it's @suzlansdell talking about the role of deliberative public engagement in the decision-making of #DataTrusts
30/ this kind of engagement critical at the establishment of #DataTrusts. in scoping its role and designing its structure, as in the end the public will be affect by the decisions of many of these kinds of trusts
31/ she's highlighting the results from the focus groups we ran as part of the project. This backed up conclusion that purpose is critical and must be clear, benefits must be obvious and also make sure that citizens can see that this is happening #DataTrusts
32/ anxiety about data being shared for commercial reasons alone. Also had high technical expectations so that data trusts don't become a weak point in the data system. Also very clear about need for #transparency in data use, collection etc. #DataTrusts
34/ we're going to hear about the legal structure of #DataTrusts. Again, I'll not live tweet, but will pull out any key issues that resonate with me
35/ Critically, and backing up @buntms and @suzlansdell's conclusions that time needs to be taken in setting #DataTrusts up, because trust law a bad legal basis and each data trust will need a bespoke structure and set of rules.
36/ key question that people may not think about when setting up #DataTrusts is that they will cost money to run (and set-up). So where does this income come from. > again comes back to incentives, will model deliver enough value for actors that they'll pay for it, but.....
37/ ... still give value to users, and ultimate beneficiaries while building and maintaining their trust as well. > a big challenge I think #DataTrusts
38/ We are rattling through. Last up is @jhardinges who's going to talk through findings and recommendations. Deep breath. Nearly there. #DataTrusts
39/ The work has identified a range of benefits from #DataTrusts: balancing conflicting views and incentives; reducing costs; making decisions more transparent and participatory; ensuring benefits are distributed more widely, equitably; new opportunities for developing new tech
40/ he notes that #DataTrusts won't bring these by right, depends on many factors, not least motives of those setting up the trusts
41/ Lots in terms of set-up, but key for me (esp as need to build in public perspectives) the scoping and co-design of #DataTrusts must be deliberate and will take time and significant commitment from data holders > raises again questions of their incentives to do this
42/ some recommendations for government/ regulators: put tools into people's hands to help them assess their trustworthiness; think about where and when to mandate that they are used #DataTrusts
43/ it's tea time, back in 30 mins... #DataTrusts
44/ Can you bear it, we're back. For a big panel of people inc @peterkwells Moritz Godel, @Adrian_Weller, @NoraNiLoideain, chaired by @Anna_D_Scott #DataTrusts
45/ are we right to be thinking about data ownership asks @adrian_weller, better to think about data rights, he says channelling @martintisne. He highlights the trade-offs between various aspects of this, which means different contexts needing different structures #DataTrusts
46/ point being reiterated that setting up the governance of data use will require a multidisciplinary team> and this in itself is going to need more upfront time as they try to understand each other #DataTrusts
47/ @peterwells noting that we need engagement even if the data isn't personal data, because we will have a stake in the outcome of the #DataTrusts, eg the environmental benefits from food waste, >and also the harms will not be evenly distributed and need to be engaged with fully
48/ getting deeper into complexity here, type and volume of data, number and type of users, cost structure of the data, the way benefits will accrue, and how to benchmark will all influence the structure and establishment of #DataTrusts
49/ @Anna_D_Scott asking questions about the incentives of different actors to set-up #DataTrusts (see tweets passim). @peterkwells noting that different actors from different sectors coming to ODI to ask qus about protecting data, making best use of it etc, not trusts directly
50/ I asked about the incentives that might operate on actors to come together to form #DataTrusts (or other governance structures). @adrian_weller noted that sometimes incentives might be large, others much smaller. It'll be context specific, and ...
51/ ... it's worth noting that information asymmetries may mean that making data open/ sharing may allow new entrants to a market, or entrench an existing one. Understanding context is going to be critical #DataTrusts
52/ Now we're hearing from @ns_ahmed on open data in Canadian example, the Toronto waterfront refurbishment, a controversial project that generated lack of trust about use of data and lack of meaningful engagement #DataTrusts
53/ and it sounds like the politics of the controversial development bled over into the debate about setting up a trust making it much more controversial (or brought the controversy into sharp relief possibly). #DataTrusts
54/ next up @SylvieDelacroix from University of Birmingham and Alan Turing Institute talking about 'bottom-up' #DataTrusts
55/ she highlights that there are noble aims from #DataTrusts, one not mentioned today is benefits of data for medical research, eg solving big social problems. however, the leaking of data reveals more about us as individuals than we would tell closest friends,
56/ there are many people who associate data ownership with control over data, but it is a very different type of property as it provides non-exclusive rights. Think more ownership of a river flowing at bottom of garden, rather than ownership of a house #DataTrusts
57/ and #DataTrusts will struggle to reverse the power imbalance inherent in data controller - subject relationship
58/ solution is to focus on rights of data subjects and way to protect these is by having data trustees who have a fiduciary obligation and undivided loyalty, ie not benefiting for selves and trying to enhance benefits for data subjects #DataTrusts
59/ her proposal is that people (data subjects) should be able to switch #DataTrusts as they think differently about what benefit they want to gain > this suffers from the same problem as personal data stores though, will people have enough information and time to do this?
60/ She notes that there are a number of implementation challenges (she answers too, but I'll only note challenges): 1. trust law based in common law systems, how apply to non-common law systems? #DataTrusts
61/ 2. low level of data literacy [point I was making earlier]; 3/ data trusts may inadvertently increase deanoymisation; 4. how deal with data that is shared, what happens when they want to leave the trust; #DataTrusts
62/ #DataTrusts and 5. accountability of data trustees >and how enforce if appointed?
63/ Now we have a panel made up of @RTaylorOpenData @LDN_CDO Erika Lewis (DCMS) @IancthomasUk chaired by @JeniT. It's the final panel of the day you'll be pleased to hear!! #DataTrusts
64/ @ldn_cdo notes that the Mayor has a target of 80% of Londoners to be walking, cycling or public transporting (that's a verb right?) within 25 years, and this is a massive data question. Raising all sorts of trust type questions #DataTrusts
65/ Erika Lewis reflecting on DCMS focus, need to look beyond question of benefits to the costs as well. > big +1 from me if this moves beyond the costs to invasion of privacy and data security to wider harms, exacerbating inequality for eg #DataTrusts
66/ also a focus on what the key problems are that need to be solved, and thinking about the rules that need to be in place #DataTrusts
67/ @IancthomasUk noting that there is a huge overhead of getting data collection, use sharing wrong and so need to invest in thinking this through #DataTrusts
68/ @RTaylorOpenData being asked about ethics washing and risk of same problem facing #DataTrusts. He notes that the structure and governance is critical, and for it to work power and control must be handed over.
69/ and this is the reason why the market hasn't led to #DataTrusts being set-up. Which does suggest that someone needs to change the incentives for businesses and other organisations.
70/ question from the internet - how do trustees gain trust? > personally I think this takes us back to Onora O'Neil again - you can't get trust, you can only make yourself trustworthy, sorry for sharing so regularly but it's key ted.com/talks/onora_o_… and so... #DataTrusts
71/ ... #DataTrusts need to have a structure that is open, transparent, participatory and genuinely shifts power. Anything else will fail. And @RTaylorOpenData highlight's the findings from the @involveUK engagement process about the high bar the public likely to set
72/ @LDN_CDO 'organisational design should be seen as part of democracy' rather than engaging to avoid being judicially reviewed or something tacked on at the end ... #DataTrusts
73/ ... > massive thumbs up from me on this, if the benefits are big enough, the downsides equally so public must be involved #DataTrusts
74/ Worry from questionner is that #DataTrusts are being seen as a panacea for multiple problems, might get a better idea of what they are by exploring what they are not
75/ So #DataTrusts are not.... Erika Lewis - not separate organisations, can voluntary organisations extend their operations for example; @IancthomasUk - not one size fits all and not scalable;
76/ @RTaylorOpenData - need clearer definition of terms, when acting on behalf of individ vs community rights?; @LDN_CDO not something that can solve all the problems around data #DataTrusts
77/ Qu from audience what sort of external oversight will #DataTrusts need? >this worries me as a question because it suggests lack of confidence in the structure @LDN_CDO seems to share my concern...
78/ #DataTrusts ...Getting this right means building governance structures that the public find trustworthy... @RTaylorOpenData highlighting big ethical questions need to be taken democratically> In the end we need a robust, open, wide public debate about the use of data
79/ @jeniT asks for one takeaway for us. @ldn_cdo - watch London, going to start a new venture with digitally leading boroughs to scale innovation in the city #DataTrusts
80/ so this is the right time to talk about algorithmic transparency and openness; @RTaylorOpenData we're moving rapidly to a data driven world, models of governance not fit for purpose so need to start redesigning, but it's complicated and expensive... #DataTrusts
81/ ... the public's expectations are high though; @IancthomasUk we need to take personal accountability for this, if it seems creepy it probably is, stop doing it, stop your organisation doing it, start small (not one thing, but good stuff tbf) #DataTrusts
82/ Erika Lewis, not spoken about the corporations that hold a lot of our data, government will need help talking to them #DataTrusts
83/ Now wrapping up. This was an epic thread. If you weren't interested in #DataTrusts, apologies... but as the last few tweets show, this stuff is real and affects us all. We need to engage widely on what data is used for and how it is governed if we want to keep control of it.
