The coronavirus pandemic has posed enormous challenges for governments seeking to delay, contain and mitigate its effects. Along with measures within health services, a range of disruptive public health tactics have been adopted to try and limit the spread of the virus and flatten the curve, including social distancing, self-isolation, forbidding social gatherings, limiting travel, enforced quarantining, and lockdowns. Across a number of countries these measures are being supplemented by a range of digital technologies designed to improve their efficiency and effectiveness by harnessing fine-grained, real-time big data. In general, the technologies being developed and rolled-out fall into four types: contact tracing, quarantine enforcement/movement permission, pattern and flow modelling, and symptom tracking. The Irish government is pursuing two of these – contact tracing and symptom tracking – merged into a single app ‘CovidTracker Ireland’. In this short essay, I outline what is known about the Irish approach to developing this app and assess whether it will work effectively in practice.
CovidTracker Ireland
On March 29th 2020 the Health Services Executive (HSE) announced that it hoped to launch a Covid-19 contact tracing app within a matter of days. Few details were given about the proposed app functionality or architecture, other than it would mimic other tracing apps, such as Singapore’s TraceTogether, using Bluetooth connections to record proximate devices and thus possible contacts, together with additional features for reporting well-being. The HSE made it clear that it would be an opt-in rather than compulsory initiative, that the app would respect privacy and GDPR, being produced in consultation with the Data Protection Commission, and it would be time-limited to the coronavirus response. It was not stated who would develop the app beyond it being described as a ‘cross-government’ effort.
On April 10th, the HSE revealed more details through a response to questions from Broadsheet.ie, stating that the now named CovidTracker Ireland App will:
“help the health service with its efforts in contact tracing for people who are confirmed cases;
allow a user to record how well they are feeling, or track their symptoms every day;
provide links to advice if the user has symptoms or is feeling unwell;
give the user up-to-date information about the virus in Ireland.”
Further, they reiterated that the app ‘will be designed in a way that maximises privacy as well as maximising value for public health. Privacy-by-design is a core principle underpinning the design of the CovidTracker Ireland App – which will operate on a voluntary and fully opt-in basis.’ There was no mention of the approach being taken; however the use of the HSE logo on the PEPP-PT (Pan-European Privacy-Preserving Proximity Tracing) website indicates that it has adopted that architecture, an initiative that claims seven countries are using their approach, with reportedly another 40 countries involved in discussions.
As of April 22nd the CovidTracker Ireland app is under development, with HSE stating on April 17th that it was being tested with a target of launching by early May when it is planned that some government restrictions will be lifted.
Critique and concerns
From the date it was announcement concerns have been expressed about the CovidTracker Ireland, particularly by representatives of Digital Rights Ireland and the Irish Council for Civil Liberties. A key issue has been the lack of transparency and openness in the approach being taken. An app will simply be launched for use without any published details of the approach and architecture being adopted, consultation with stakeholders, piloting by members of the public, and external feedback and assessment.
There are concerns that a centralized, rather than decentralized approach will be taken, and there is no indication that the underlying code will be open for scrutiny, if not by the public, at least by experts. It is not clear if the app is being developed in-house, or if it has been contracted out to a third-party developer and if the associated contract includes clauses concerning data ownership, re-use and sale, and intellectual property. There are no details about where data will be stored, who will have access to it, how will it be distributed, or how it be acted upon. There is unease as to whether the app will be fully compliant with GDPR and fully protect privacy, especially given that a Data Protection Impact Assessment (DPIA), which is legally required before launch, has seemingly not yet been undertaken. Such a DPIA would allow independent experts to be able to assess, validate and provide feedback and advice.
Critics are also concerned that CovidTracker Ireland merges the tasks of contact tracing and symptom tracking which have been pursued separately in other jurisdictions. Here, two sets of personal information are being tied together: proximate contacts and health measures. This poses a larger potential privacy problem if they are not adequately protected. Moreover, critics are worried that CovidTracker Ireland might become a ‘super app’, which extends its original ambition and goals. Here, the app might enable control creep, wherein it starts to be employed beyond its intended uses such as quarantine enforcement/movement permission. For example, Antoin O’Lachtnain of Digital Rights Ireland has speculated that we might eventually end up with an app to monitor covid-19 status that is “mandatory but not compulsory for people who deal with the public or work in a shared space.”
As Simon McGarr argues, the failure to adequately engage with these critiques and to be open and transparent means that “the launch of the app will inevitably be marred by immediately being the subject of questions and misinformation that could have been avoided by simply overcoming the State’s institutional impulse for secrecy.”
Internationally, there is scepticism concerning the method being used for app-based contact tracing and whether the critical conditions needed for successful deployment exist. Bluetooth does not have sufficient resolution to determine two metres or less proximity and using a timeframe to denote significant encounters potentially excludes fleeting, but meaningful contacts. There are also concerns with respect to representativeness (for example, 28% of people do not own a smartphone in Ireland), data quality, reliability, duping and spoofing, and rule-sets and parameters. The technical limitations are likely to lead to sizeable gaps and a large number of false positives that might produce an unmanageable signal-to-noise ratio, leading to unnecessary self-isolation measures and potentially overloading the testing system.
There is a concern that app-based contact tracing is being rushed to mass roll-out without it being demonstrated that it is fit-for-purpose. Moreover, the app will only be effective in practice if: there is a program of extensive testing to confirm that a person has the virus and if tracing is required; and 60% of the population participate to ensure reach across those who have been in close contact (c.80% of smartphone users). The symptom tracking relies on self-reporting, which lacks rigour and, as testing has shown, a large proportion of the population who were tested because they were experiencing symptoms returned negative. This is likely to lead to a large number of false positives and it is doubtful that these data should guide contact tracing.
At present, while Ireland is ramping up its testing capability towards 100,000 tests a week, it might need to increase that further. The Edmond J. Safra Center for Ethics at Harvard University suggest that in the United States: “We need to deliver 5 million tests per day by early June to deliver a safe social reopening. This number will need to increase over time (ideally by late July) to 20 million a day to fully remobilize the economy. We acknowledge that even this number may not be high enough to protect public health.” The equivalent rate for Ireland would be 300,000 tests per day. In Singapore, only 12% of people have registered to use the TraceTogether app, which raises doubts as to whether 60% of the population in Ireland will participate, especially since the public are primed to be sceptical given media coverage about the app have raised issues of privacy, data security and data usage.
Will CovidTracker Ireland work and what needs to happen?
There is unanimous agreement that contact tracing is a cornerstone measure for tackling pandemics. Assuming that the privacy and data protection issues can be adequately dealt with it, it would be good to think that CovidTracker Ireland will make a difference to containing the coronavirus and stopping any additional waves of infection.
However, there are reasons to doubt that app-based contact tracing and symptom tracking will make the kind of impact hoped for unless:
its technical approach is sound and civil liberties protected;
there is testing at sufficient scale that potential cases, including false ones, are dealt with quickly;
the government can persuade people to participate in large numbers.
The government might also have to supply smartphones to those that do not own them, as they did in Taiwan. Persuading people to participate will especially be a challenge since the government is not being sufficiently transparent at present in explaining the approach being taken, the app’s intended technical specification, how it will operate in practice, its procedures for oversight, and how it will protect civil liberties.
It is essential that the government follow the guidance of the European Data Protection Board that recommends that strong measures are put in place to protect privacy, data minimization is practised, the source code is published and regularly reviewed, there is clear oversight and accountability, and there is purpose limitation that stops control creep.
If implemented poorly, the app could have a profound chilling effect on public trust and public health measures that might be counterproductive. As a consequence, the Ada Lovelace Institute, a leading UK centre for artificial intelligence research, is advising governments to be cautious, ethical and transparent in their use of app-based contact tracing. Ireland might do well to heed their advice.
Our colleague from the Geography Department at Maynooth University, Alistair Fraser, has published a fascinating paper as a Progcity working paper (31) – Land grab / data grab. Focusing on the use of big data in food production he develops two useful conceptual tools ‘data grab’ and ‘data sovereignty’, using them to explore ‘precision agriculture’ and the notions that data is a ‘new cash crop’ and the ‘the new soil’.
Abstract
Developments in the area of ‘precision agriculture’ are creating new data points (about flows,
soils, pests, climate) that agricultural technology providers ‘grab,’ aggregate, compute, and/or
sell. Food producers now churn out food and, increasingly, data. ‘Land grabs’ on the horizon
in the global south are bound up with the dynamics of data production and grabbing, although
researchers have not, as yet, revealed enough about the people and projects caught up in this
new arena. Against this backdrop, this paper examines some of the key issues taking shape,
while highlighting new frontiers for research and introducing a concept of ‘data sovereignty,’
which food sovereignty practitioners (and others) need to consider.
Developments in software and digital technology have had wide ranging impacts on our leisure time, from movies on demand on our mobiles, internet on public transport and the ‘selfie’ saturated world of social media. Yet advancements in technology have also reached creative activities that are often considered far from mainstream and groups of individuals, who though they share a common interest, may pursue their leisure activity individually and in relative isolation.
One such social group is that of model railway enthusiasts. For these collectors, builders and hobbyists the developments in software have enabled fundamental changes to the way they explore and express their interests. Geographically dispersed and relatively few in number (estimated in the low hundreds in Ireland) software has offered a means of augmenting the traditional physical locations of interaction, socialising and knowledge sharing. Software and connectivity have enabled a network of online interactions that has linked individuals more closely with the commercial suppliers and the specialist manufacturers of the models they consume, extending the reach of the community beyond the traditional clubs or shows. It has facilitated efficient access to, and the sharing of, previously inaccessible or unknown historic and practical knowledge regarding even the most obscure topics such as window size and seat positions. Building upon more traditional sources of historic data such as printed media and journals, software has also enabled the capture of dispersed and divergent forms of data and facilitated their transformation, via computerised production methods, into ready-to-run models with unprecedented levels of physical detail and functionality. Continue reading →
Introduction to a series of blog posts on big data:
By Tracey P. Lauriault, Programable City Project.
My area of research on the Programmable City Project examines how digital data materially and discursively support and process cities and their citizens. This work falls within an emerging field called critical data studies. A recent working paper entitled Small Data, Data Infrastructures and Big Data, is an example of critically thinking about data, as is the forthcoming book The Data Revolution: Big Data, Open Data, Data Infrastructures and Their Consequences and my earlier work on data preservation, access, and data infrastructures. The new Big Data and Society journal also provides some theoretical orientation. Dublin and Boston, are the Programmable City’s formal study sites and I will add a little bit of Canada in that mix from time to time.
The first part of my research consists of constructing ‘data landscapes’ for Dublin and Boston, by which I mean mapping out the different organisations involved in producing, analyzing, and selling data and analytic services, of which big data are an important component. A key element is reading a number of reports, papers, and consultations on the topic, as well as conducting interviews and attending industry and academic events, I will share questions and observations here in blog posts. At first, these will mainly be descriptive and comparative. However, this reading combined with empirical case studies will lead to a more critical examination of the data assemblages of big data, open data, indicators, maps and data infrastructures. I will be guided by two theoretical frameworks, Rob Kitchin’s ‘Data Assemblages’ and a modified Ian Hacking Making Up People and Spaces framework.
The report was commissioned “to advance recommendations on measures to build up the big data and analytics talent pool through domestic graduate output, continuing professional development within industry, and, where necessary, attraction of talent from abroad including expatriate talent”. The Government’s ambition is for the Republic to become a leading country in the field of big data in Europe, and the report is designed to guide the creation of an Irish big data ecosystem based on public and private collaboration.
The research for this study was driven by the Joint Industry/Government task force on Big Data which was formed in June 2013. Big data is also considered to be a disruptive reform in the Action Plan for Jobs 2013, which is a reform considered to have the“potential to have a significant impact on job creation, to support enterprises or where Ireland can profit from a natural advantage or opportunity that presents itself in the economy” and a research priority Data Analytics, Management, Security and Privacy 2013. Forfas commissioned the research in a Request for Tender for the Provision of Research for the Study Assessing the Demand for Big Data / Data Analytics Skills in Ireland 2013-2020 (LINK TO PDF) in July 2013.
Ernst & Young (E & Y) and Oxford Economics were jointly commissioned to blend their different fields of expertise to conduct the international trend analysis, undertake the consultations with companies and key informants and to model big data and analytics skills demand scenarios. Their work was overseen by a 17 member Steering Group chaired by the EGFSN.
The Big Data Skills Report was overseen by a Joint Industry/Government task force on Big Data and a Steering Group with representation primarily from the private sector (IBM, Accenture, Glanbia, CPL, Datahug, Nanthean Technologies, Deloitte, Vidiro Analytics, Enterprise Ireland, IDA Ireland and the Quinn School of Business) and 4 government offices (Revenue Commissioners, Department of Education and Skills, Higher Education Authority and Forfás).
Provenance:
The report is the outcome of a number of fast tracked high level strategies for the Government of Ireland as follows:
The Department of Jobs, Enterprise and Innovation (DJEI) considers big data to be Priority Area B in the Research Prioritization Steering Group (Mar. 2012) report after Priority Area A which is Future Networks and Communications.
This is followed by the DJEI’s Action Plan for Jobs 2013 (Feb. 2013) that repositions big data as Disruptive Reform 1 which is to “Build on our existing enterprise strengths to make Ireland a leading country in Europe in ‘Big Data’”.
Meanwhile outside of government SAS UK and Ireland commissions the Centre for Economics and Business Research (Cebr) to assess the big data sector and releases Data equity – Ireland Unlocking the value of big data (June 2013) coining the term ‘data equity’ from the idea of brand equity. Companies will be forming data equity with the power of analytics.
Big data is then considered to be a Global Technology and Service Trends Influencing Irish ICT Skills Demand in the Forfas Expert Group on Future Skill Needs (EGFSN) report Addressing Future Demand for High-Level ICT Skills.(Nov. 2013).
As part of the actions plan, Forfas launches a Request for Tender for The Provision of Research for the Study Assessing the Demand for Big Data / Data Analytics Skills in Ireland 2013-2020 which was awarded to E & Y and Oxford Economics.
Big data is positioned again as disruptive and integrated with ICT skills, it is a “disruptive growth and innovation phase. This includes the adoption of cloud computing, the penetration of mobile devices and technologies and the Internet of things, the emergence of Big Data analytics, IT security, micro- and nanoelectronics and the adoption of social technologies in both the personal and business environment” in the EGFSN ICT Skills Action Plan (Mar. 2014) also an outcome of the Action Plan for Jobs 2013.
How are big data analytical skills categorized in the Big Data Skills Report?
The objectives of the Report were to forecast the future annual demand across the economy between 2013 – 2020, for big data & data analytics and related roles which included a mapping of qualifications, skillsets and competency requirements. Skill categorization was pre-determined in the Request for Tender as follows:
1. Deep Analytical Talent skills requiring a combination of
advanced statistical, analytical, and machine learning skills,
business skills to assess the meaning of data and to derive business insights &
communications skills to explain/persuade other executives.
These roles can originate from a range of backgrounds such as mathematicians, statisticians, actuaries, operational research analysts, economists and include the newly termed role of the ‘data scientist’.
2. Big Data Savvy; comprising of “data savvy”
managers,
CIO’s,
market research analysts,
business and functional managers.
These professionals require an understanding of the value and use of big data & data analytics to enable them to interpret and utilise the insights from the data and take appropriate decisions to advance their company strategy and drive business performance.
3. Supporting technology roles; these personnel have the skills to develop, implement and maintain the hardware and software tools required to make use of Big Data/Data Analytics software and hardware. The core technologies of those implementing Big Data solutions tend to be focused Hadoop and a growing range of SQL databases. These roles include:
programmers and software development professionals and
IT business analysts, architects & system designers.
These 3 categories were originally created by the McKinsey Global Institute (MGI) in their 2011 Big data: The next frontier for innovation, competition, and productivity report. The original MGI categorization is as follows. Note that these categories are restricted to the mention of ‘hard skills’ types of occupations. Therefore big data is considered to be primarily a technical field.
Role Categories and Occupation Types (p.38)
E & Y expanded the MGI categorization in the Big Data Skills Report as seen in the skills, competences and qualifications requirements in the table below. In this case, under the ‘Big Data and Analytics Savvy’ category, some public policy, regulatory and governance roles ‘hard skills’ such as data protection and IP Knowledge are taken into consideration, while Ethics is considered to be a ‘soft skill’ for both ‘Deep Analytical Talent’ and ‘Big Data Savvy’.
Types of Skills and Competencies Required Across Categories (p.50).
This is the last we see of these types of ‘soft’ and ‘hard’ skills in the report. Which is rather unfortunate, as leadership and innovation in any field requires good governance, and this is especially the case for big data. Furthermore, once a categorization gets amplified, institutionalized and authorized it becomes real and begins to ‘make up kinds of people’ as per Ian Hacking’s framework and as illustrated below in Lauriault’s Making up People and Places framework based on Hacking’s work. Classifications describe people and places, institutions begin to use that class, it becomes part of expert knowledge, in this case MGI, E & Y and Oxford Economics. Once set, experts begin to count, quantify, normalize and correlate things, in this case skills, academic qualifications, and employment data. Once these are reported, an institution or groups of experts take some action, in this case they develop a strategy and some action plans. Once actions are taken, the class really becomes true, scientific or real as it becomes a thing. It becomes a normal part of doing business, a part of a bureaucratic process, in this case, where public sector moneys will be directed toward increasing a set of skills to meet a demand determined by the data analytics private sector in Ireland. In terms of resistance, it may be, that with time we will see those labelled as part of the ‘Deep Analytical Talent’ pool may want to be called data scientists, the ‘Big Data Savvy’ group may want to be big data governors while the supporting Technology Group may want to be considered as big data infrastructure providers.
In this case the DJEI in its Request for Tender has firmed up these categories by making them the focus of the research, and these have now become real categories in the Republic by E & Y and Oxford Economics who structured their analysis accordingly. It is expected that these will appear in many future analyses.
Expert knowledge informing the skills demand model in the Big Data Skills Report:
The development of baseline estimates and projections for big data employment in Ireland were informed by the following reports:
In addition, the Big Data Skills Report adopts the most common definition of big data developed by Gartner in 2001, namely a field that works with datasets that are big in volume, velocity and variety, or the 3Vs. The definitions of big data have however evolved and below is a comparative table developed by Kitchin (2014). The Big Data Skills Report does not however discuss data.
huge in volume, consisting of terrabytes or petabytes of data;
high in velocity, being created in or near real-time;
diverse in variety, being structured and unstructured in nature, and often temporally and spatially referenced;
exhaustive in scope, striving to capture entire populations or systems (n=all) within a given domain such as a nation state or a platform such as Twitter users, or at least much larger sample sizes than would be employed in traditional, small data studies;
fine-grained in resolution, aiming to be as detailed as possible, and uniquely indexical in identification;
relational in nature, containing common fields that enable the conjoining of different data sets;
flexible, holding the traits of extensionality (can add new fields easily) and scalability (can expand in size rapidly).
(Boyd and Crawford 2012; Dodge and Kitchin 2005; Marz and Warren 2012; Mayer-
Schonberger and Cukier 2013).
The definition of big data, the Big Data Skills Report, skills demand models, skill categories, resulting scenarios and recommendations were all authored and created by private sector entities. Concurrently, the Steering Group overseeing this study mostly represents private sector interests. And the direction of the strategy falls under the Joint Industry/Government task force on Big Data. It can then be concluded that those who stand the most to gain from the recommendations made in the report also authored and oversaw them. In other words, there is no critical assessment of the findings, and it would seem that no one is looking out for the public interest.
A more objective approach would have included broader membership of the Steering Group and the EGFSN by including academic experts from the social sciences, information studies, communications or media studies, members of legal and financial professions, ethicists, public sector big data producers or independent think tanks. Also, what if the process was overseen by the ‘Joint Industry/Government Public Interest task force on Big Data’? Had that been the case, there may have been a focus on some of the more pernicious effects of big data & data analytics, such as privacy, procurement, security and ethics. It would have been preferable to first have a national big data research and development strategy for the Republic of Ireland with a focus on how big data can help resolve national issues (e.g., environment, housing, financial, energy, agriculture), identify where expertise in those areas lies or is lacking, and then direct resources toward the resolution of those. In addition, a national strategy would have included an inventory of all the private sector entities involved in big data on the Island and a survey of the skills sets in house and sought to develop a skills demand baseline from that, combined with an inventory of skills in the public and academic sectors. Finally, an analysis of the public policy issues related to big data, including threats to the public interest would have been assessed, and along with a big data private sector, R & D strategy, there would have been a big data governance process in place that would include a regulatory framework and some checks and balances to guide the sector toward ethical conduct.
What did the Big Data Skills Report do?
1. The drivers for a big data were defined.
Big data is a disruptive reform in the Action Plan for Jobs 2013 related to a growing industrial sector and subsequently job growth in the country, big data is also considered to be a sector from which significant value can be derived and therefore worth investing in. Interestingly, open government and open data are considered as drivers. This is unusual since most data found in open data portals are characterized as small data. It is true that that the public sector and publicly funded research produce and analyze datasets that would be characterized as big data, but these have been being produced for quite some time, are quire specialized and have been disseminated under a PSI Licence prior to the advent of open data. Open data as a driver may be attributed to the fact that the private sector can better commercialize public sector data, both big and small, by combining these with theirs or using public sector data as a baseline. All sectors stand to gain if public sector data are coordinated, centrally discoverable, well described in metadata, free and under an open licence.
The new Post Code system to be deployed in Ireland by 2015 was considered to be a “late mover adoption” in the Report. Paradoxically, this will not be an open data framework dataset available to the public as it is being produced and licenced under a cost recovery program. The “Postcode Management Licence Holder” (PMLH) contract is to Capita Business Support Services Ireland (supported by BearingPoint and Autoaddress) and the PMLH contract will be fulfilled by the Capita-owned business, Eircode. The new post code will be particularly helpful to big data companies, marketers and direct mailing companies wishing to compete with the postal service, but will not necessarily be useful for civil society organizations as it will be too costly. The average citizen may also be paying for these data multiple times:
the public paid for the creation of the dataset in the first place,
it will become a cost included in the price of products,
as part of taxation because government agencies will have to acquire these under a cost recovery regime (e.g., each local authority, all national departments, and etc.)
tuition fees as academic institutions will also have to pay for this dataset.
There are a number of other important framework datasets not mentioned in the report, most notably the small area file produced by the National Centre for Geocomputation at NUIM developed for the CSO. This allowed for the first time, in the 2011 census, fine grained analysis of neighbourhoods in Ireland and would be of benefit to the big data sector as data big and small need to be aggregated into common and small geographies. Along with open data, and framework datasets, the report did not discuss the formation of a spatial data infrastructure which is how many other nations deal with national and sub-national geospatial datasets such as road network files, small area files, post code files, demographic base files, environmental and natural resources data to name a few. SDIs are the means by which datasets of this nature, are standardized, made interoperable and accessible. This would also be of assistance to big data interests, but also for social, economic and environmental management for the country. These are more robust than open data initiatives as they are based on interoperable policies, standards and technologies that are adopted enterprise wide. The Canadian Geospatial Data Infrastructure for instance is now becoming a platform and open data will be integrated into it while departments will be geo-enabling their administrative datasets accordingly.
2. Baseline employment demand was determined
The MGI, Accenture, Cebr and SAS reports mentioned earlier informed the skills demand model. The following explains how it was done:
E & Y and Oxford Economics determined that reports consistently estimated that existing employment in this sector consisted of 1.5% to 2% of total employment, although categorizations differed.
Eurostat data were used to compare shares of employment in Ireland and the UK by sector and CSO Data from the Quarterly National Household Survey (QNHS) were used (See figure 3.2 p. 42).
Based on estimates and proportions for the US and UK, the Deep Analytical Talent employment demand for Ireland was estimated.
Also, the proportion of those in established roles was considered to be more than half of that demand while the remainder are emerging analytical roles.
The Big Data Savvy demand was determined by applying the proportion of total employment in this category in the US to the Irish employment data.
The Supporting Technology demand was derived from structured interviews with the 45 enterprises and organizations in Ireland. A ratio of 1:4 was determined of Deep Analytical Talent to Supporting Technology professionals.
To understand the sectoral distribution, the analysis in the 2011 MGI report was used. In the MGI analysis low, medium and high data intensity based on data storage capital stock per firm groups were devised. CSO data were used here to “understand the data intensity of industries as the capital stock in computer software per employee” (p.44).
Below is the table of the Results:
Estimate of Baseline Demand in Big Data and Analytics in Ireland (p.45).
The above was the extent of the methodological information provided explaining how baseline demand numbers were derived. No metadata, tabular data or dates were provided, nor details about which CSO and Eurostat data were used, nor was there any anayslis provided on how the classes were adjusted to the 3 categories of skills defined in this Big Data Skills Report. Ireland is not that big a country and a survey of all the companies doing big data analytical work could have yielded a much more robust baseline. Estimates of growth were based on the UK and the US. The assumption that the composition of the firms here differs as a result of the favourable tax regime and that Ireland is a much smaller country suggest that those estimates would be an overstatement for Ireland.
3. Structured consultations were conducted
A Structured Interview Survey about the employment levels across each of the three categories was conducted by telephone with 45 Forfás selected Irish based enterprises and organisations that employ big data & data analytics talent. This included 35 companies (foreign owned and indigenous), 7 government bodies, 2 data analytic research centres and 10 key informants. Some workshops were also held.
Demographics, vacancies, the difficulty in filling vacancies were determined. In addition future demands, predicted growth, the difficulty in acquiring and keeping talent, sources of future skills, where sources of skill could be sought within the most common disciplines, training opportunities, relationships with academia, turnover, and attrition were estimated. The results of which are presented in the Information box below (p. 59).
In lots of ways this part of the report left more questions than answers. Which companies and institutions were surveyed? How many companies do this type of work in Ireland? What were the questions asked in the survey? How were vacancies and the difficulty to fill positions determined? Did companies classify the skill sets they needed in the same way that MGI, E & Y and Oxford Economics did? If not how were differences rectified? As discussed, the big data categories and classifications did not focus on public policy issues which resulted in the social sciences and law faculties being unmentioned in the lists of university programs, and most notably absent from the report are the pools of both ‘Deep Analytical Talent’ and ‘Big Data Savvy’ skills found in geodemographics, geocomputation, geomatics, geophysics as well as the earth sciences. While humanities is mentioned in the report, the digital humanities was not.
This section, like the previous lacked a methodological explanation and the bread crumbs required to assess results. The focus is heavily oriented toward the hard skills, and the recommendations are generally quite sensible. However, of concern is the recommendation that “third level, cooperation between industry and academia is crucial to ensure that education reflects skills needs. Proposals included involving industry experts in devising curricula and teaching analytics on university programs” (p.59). This is problematic as universities could simply become instruments to serve corporate interests, and in order to cut costs research may become too instrumentally focussed into a particular sector at the expense of others, and in both cases this may thwart critical reflection in this area for fear of budget losses. Finally there is a concern that this may also lead to an overemphasis on quantitative research and undermine programming in other areas such as qualitative research, and the social sciences.
4. Big data future demand scenarios were produced
Three scenarios were offered which included expansion, replacement and skills demand projections as well as up-skilling demand. These were derived by examining the growth projections and demand baselines discussed earlier, the information collected during the consultations and E & Y and Oxford Economics domestic economy sectoral employment forecasts.
The scenarios factored in the following global drivers for big data:
Increase in the creation and availability of data
Growing recognition of economic returns from the use of big data
A few global risks were described, none of which included privacy, security, intellectual property, procurement and ethics, although the fact that along with a lack of relevant data there may also be an abundance of high quality data.
The scenarios for Ireland were based on a number of assumptions, drivers and supporting conditions as seen in the table below:
Summary of Scenario Assumptions, Drivers and Supporting Conditions (p.71)
Resulting in the following predictions as seen in the table below. Scenario 3 would Position Ireland as the leading country in Europe in big data, Scenario 2 would be the forecast if Ireland caught up to other countries such as the UK, while Scenario 1 represents low growth.
Summary of Future Demand Change Projections, 2013-2020 (p.82)
It is difficult to say if any of these projections are reasonable as there are no supporting data, data sources, cross classification mapping or algorithms provided. Furthermore, these scenarios are also based on the problematic baseline demand numbers discussed earlier. The implications of Scenario 3 are however enormous. Should Scenario 3 be adopted, significant investment in education and training in quantitative and computational areas would be required in universities, potentially at the expense of non-quantitative fields, such as those related to governance and public interest issues discussed earlier, some fields would be overlooked while inducements would be required to keep existing and new talent here. The drivers for Scenario 3 are in play, and the outcomes are entirely focussed on private sector big data R & D, and not directed at applying big data resources toward addressing issues of national importance such as environment, energy, quality of life, housing or climate change. If my observations are correct, the strategy is for public sector funding to be directed toward the alleviation of the shortage in skills in this sector, for R & D to answer questions developed by the private sector (INSIGHT and CeADAR), and for inducements to keep skilled personnel here, while also continuing to support a favourable tax regime and FDI inducements. Some Irish citizens may benefit from some of the new jobs that may come on offer, but how Irish Society benefits as a whole and what is in the general public interest in any of these scenarios is uncertain.
5. The big data skills supply was determined
The MGI 2001 and the Cebr 2013 reports informed the model used to determine the current supply of big data skills. MGI conducted an analysis of US education data in fields renowned for having advanced quantitative training. Ratios were derived of the total graduates and then applied internationally. It is only here do we find the physical sciences and social sciences listed. The Cebr report considers supply and demand but its analysis factored in a different set of academic programs where the physical and social sciences, are not factored in. The analysis in Ireland examined a series of courses and programs considered most relevant to the production of the Deep Analytical Talent cohort such as:
Dedicated big data & analytics programmes
Programmes that include significant training/elements in data analytics
Maths, Statistics and Science
Engineering programmes
Physics programmes
Data analytics programmes in Northern Ireland
Private data analytics programmes
Online education in data analytics
This part of the report was more comprehensive as the types of programs, courses and numbers were provided in the analysis and it was estimated that approximately 2000 graduates per year in courses most directly related to emerging deep analytics. Again, graduates related to the ‘Big Data Savvy’ governance and law types of programs were not focussed on, this is a missed opportunity for Ireland and a very large oversight of this report.
6. Policy measures here and abroad were examined
Policy was defined loosely to encompass a range of methods to grow the skills supply discussed throughout. A list of international measures to grow talent was discussed, such as post secondary program measures, but also collaboration between industry and higher education institutions that might entail the following:
Offering ‘real world’ work problems and large datasets to mine;
Providing data analytics software and hardware;
Providing relevant work experience opportunities;
Shaping specialisms or electives within programmes (including the actual provision of the taught modules);and
Promoting analytics as a career path for students.
These are relatively straight forward approaches, but I would add the cautionary note of ensuring that students should be able to study on multiple platforms and there should be no exclusive deals or monopolies on platforms and software; that work experience be tied with some sort of fair remuneration, that there be the allowance for critical analysis. The offering of real world data sets from the private sector would also be most exciting. Because of a lack of focus on big data governance issues, no recommendations in this area were made.
There are also many other initiatives already in place in Ireland which form part of the Action Plans for Jobs report. Significant investment in research institutes such as INSIGHT which partners with industry to address private sector problems within a number of important themes such a health, economy, journalism and humanities, energy and the environment, sport and wellness, telecom & network and media processing. There is also R & D investment into CeADAR which is another major research institute that partners with the private sector to deliver quantitative and computational skills. There has also been investment in infrastructure such as the Irish Centre for High-End Computing and the Telecommunications Software and Systems Group.
7. Recommendations in the Big Data Skills Report
Many of the recommendations in the report were previously commented on. Chapter 8 provides a number of tables with short, medium and long term suggestions. There were however a number of items scattered in this chapter that are new and notable:
“The overarching recommendation is that a consultative group comprising representatives of industry, academia and relevant agencies should be established” to oversee the implementation of recommendations over a 6 month period.
This is refreshing, and it is hoped here that more than just business school and technical fields from academia will be reflected in the composition of this consultative group and that there will be room for critical debate.
“For Ireland to become a leading country for big data a higher level of skills supply will also be required. This will involve and enhanced supply of Masters and PhD graduates along with an expansion of places on courses designed for those in advance degrees in STEM to transition into high-end analytics.”
“Introduce targeted competitive funding available for post graduate specialist analytics programmes to reduce tuition fees, incentivise participation and increase places available.”
Beyond what was previously mentioned earlier in terms the focus on ‘hard skills’, there is currently a hiring cap in universities in Ireland, a wage cap, and a devaluation of the ‘soft skills’ required to develop the governance and legal issues related to a growing big data industry in Ireland. Careful consideration here is required to ensure that the non STEM departments do not suffer losses in order to build up the ‘hard skills’ demand, and that a set of ‘soft skills’ be developed. The following are some notable examples in other jurisdictions: The Centre for Law Technology and Society, Canadian Internet Public Policy Interest Clinic, The Munk School of Global Affairs, The Citizen Lab, and The Berkman School for Internet and Society.
“Business communication skills, critical thinking and project management skills should be taught across all STEM disciplines”.
This is a welcome addition. A bonus would be up-skilling ethics, law and public policy departments toward an understanding of big data issues.
“Firms should adopt an enterprise-wide approach to managing their analytical capabilities, including the up-skilling of staff for data protection and governance”.
Beyond up-skilling staff, there needs to be a recognition that this sector needs to be regulated, and a number of public policy issues need to be addressed, which implies that resources toward academic disciplines such as political science, public policy, law, communications, contract law, media studies, and social-sciences.
“…Career guidance in schools should communicate the availability of career opportunities in analytics to students (particularly females) and their parents.”
This is also a refreshing recommendation. Women Invent Tomorrow, Digi Women, Coder Dojo’s and Women in Technology and Science (WITS) in Ireland have done great work in this area, and a number of important studies have been published on the topic. Female STEM experts often complain that they are invited to boy’s schools but not girl’s schools. The Big Data Consultative Group should include experts from these organizations and refer to existing studies to develop this recommendation.
“Industry and State Agencies should work with the CSO and Revenue Commissioners to explore the possibility of further developing official measurements of big data and analytics employment”.
This is very important, and the recommendation I would make is for the creation of an open and comprehensive business register that would include a profile of the skills in all firms, including big data, combined with an open and accessible inventory of all businesses involved in big data in Ireland. The greatest concerns about this report, is the fact that the numbers are based on very loose definitions, combined with the absence of an inventory of what is currently ongoing in Ireland, along with the imposition of categories that may not apply.
The public service recommendations are seen in the table below (p.111):
These are all welcome recommendations, and as discussed making the post code file open should also be a part of this, and this procurement process should be referred to as an example of how to acquire a service not for the public interest. There should also be serious consideration for the creation a national spatial data infrastructure. Companies can help the public sector overcome its data analytical short comings; however this needs to be accompanied by good procurement practices and agreements whereby government own the data and can look under the hood of any software or analytical output provided, and that data be interoperable with multiple analytical platforms and not locked into proprietary systems. In addition, in the science based departments, there is significant data analytical capabilities that have been overlooked here, especially in the EPA, Marine, ordinance survey, the Commissioner for Irish Lights, the CSO, population health and others. Perhaps along with an inventory of data assets an inventory of in house skills and the development of a cross departmental and jurisdictional public sector big data analytical working group should be created.
Finally, the report ends with “the economic and social benefits available from enhanced adoption of big data and analytics are potentially transformative. The spectrum of benefits on offer spans from improved health and environmental outcomes to better efficiency”. This is fantastic; however, the body of the Report did not discuss any targeted efforts at these issues. And it is this along with other issues discussed throughout this analysis that is part of the Report’s major shortcomings.
Final Observations
The report is to be lauded for attempting to address in a relatively tight time frame the criteria as set in the Request for Tender. It sets out a picture of the big data landscape in Ireland, before providing a number of recommendations as to how to nurture and expand Ireland’s role as the international site for big data research and development and commercial enterprise.
The report, however, also has some shortcomings and limitations. It is, for example, narrowly focused on developing strategies to build the human resources capacity for private sector companies in Ireland through the use of public education and research funds, with only a small focus on the private sector funding the training of its own staff. Moreover, the focus is almost exclusively on the ‘hard skills’ in areas of product and service development. In fairness, it is part of a larger Action Plan for Jobs and an outcome of this is the government’s research strategy of Data Analytics, Management, Security and Privacy 2013. Although, management, security and privacy in that strategy are primarily discussed in instrumental and technological terms and not in public policy terms, nor are these discussed in the Big Data Skills Report, most notably there is no mention of law and regulation.
Unfortunately the report was not created to inform and complement a national big data strategy. The strategy therefore is not R & D in general, or for the support of thematic areas that could benefit from big data analytics such as molecular biology, marine research, pharmaceutical and agribusiness, addressing climate change or energy production and efficiency in Ireland. Instead, the strategy is about jobs and skills for existing industry already involved in big data analytics, and to some extent the public sector, and not about a plan for the nation, nor for what might be considered the public good. For example, there is no systematic mapping out of the big data sector in Ireland; there is no inventory of the companies engaged and what they do; there is no analysis of the infrastructure required, albeit there is discussion elsewhere about cloud and high end computing; there is no understanding of the software used; nor do we know if there are clusters of excellence or knowledge where big data analytics are concentrated except for INSIGHT and CeADar which are excellent R & D instrumental programs. There was an excellent investigation on the skills supply side in Chapter 6, although digital humanities, geosciences, geocomputation, geomatics and the earth sciences were left out as major contributors to developing a pool ‘Deep Analytical Talent’.
Furthermore, the report does not include a focus on building public policy capacity related to issues such as law, privacy, data protection, regulation, copyright, procurement, patents, intellectual property (IP) and data resale, to name a few. The creation of public internet interest legal clinics, developing centres of excellence or think tanks in these areas as well as the creation of internet law and policy research chairs would be a good way to complement the demand for ‘hard skills’ focus. These ‘soft skills’, competencies and qualifications are subsumed within the Big Data Savvy category, although there is no explicit mapping of these against what is offered in academic institutions. The solution is envisaged as a hollowing out of the state with the private sector undertaking public sector work: ‘Public sector employers are less ambitious about their future employment levels, in part due to the current restrictions on recruitment. Part of their demand is likely to be addressed through outsourcing to the private sector” (p.13).
The policy implications of government outsourcing data analytical functions to the private sector are troublesome to say the least. Questions of who own the data? Can those data be shared with the public in an open fashion? What was the model used to do the analysis? What algorithms were deployed and are they propriety? There are issues of ownership and control that must be addressed. The Post Code exemplifies these problems, it is a lauded example in the Report, yet this is a significant public sector big data investment will primarily benefit the private sector and not the public. If evidence based decision making is something the Irish Government is truly invested in, and it would seem that it is with its open government and open data strategies, then those legal, policy, methodological, data life-cycle management and procurement issues need to be addressed. This is especially the case if issues of public expenditure and forecasting are based on the output of private sector companies who may stand to gain from the analysis, as is the case of the outcomes of this report. Uncritical predictive governance is also of growing concern, necessitating, along with investment in the ‘hard skills’. Investment in the social sciences and critical thinking are also required, and these skills need to be found in non STEM academic departments. Political economy, communications, media studies, political science, socio-technological studies, public policy and public affairs, should also be part of the big data ecosystem in Ireland. I think this is one of the big missed opportunities of this Report.
If Ireland truly wants to become an innovator in this field and the leading country in the field of big data in Europe then part of that leadership should include law, regulation and policy. These are conspicuously absent and perhaps this is the result of the drivers. It is overseen by a Joint Industry/Government task force on Big Data and a Steering Group with representation primarily from the private sector. The EGFSN also does not include lawyers, auditors, actuaries and public policy experts, but it does include an excellent group of subject matter experts and dedicated public servants in a number of areas. Industry needs to be tempered by solid public policy and regulation, while the public interest needs to be factored into any government strategy, industry is very well represented here, the public interest beyond jobs and FDI is not.
The Report, curiously mentions the Government’s joining of the Open Government Partnership as well as the development of its open data strategy, as these are considered as enablers of big data. This is unusual, as most of the data public administrations produce are small data, and that includes the Census and mapping agencies such as the Ordnance Survey. Making government data easier to find and under less restrictive licencing would make everyone’s job easier, and that is a good thing. Although, a spatial data infrastructure would be better in the long term. In addition, this would allow for greater commercialization of public sector data, as small data, and the big datasets produced by some departments and publicly funded research could be combined with private sector data and also be used to develop baselines. The Report does not discuss small data in a meaningful way except to mention that they exist. The Big Data Skills Report is, overly reliant on the analysis of the McKinsey Global Institute 2011 report entitled Big data: The next frontier for innovation, competition, and productivity, and it is here that open data is mentioned as a driver. This aligns with earlier work on the Programmable City that points to open data increasingly being of interest for commercialization purposes. Opening government is a good thing, as are open data in general, but these are a small part of the big data puzzle, and procurement needs to take an open by design approach, and avoid situations such as the post code and the Ordinance Survey.
The report offers an analysis of existing employment demand, produced three future demand scenarios, assesses Ireland’s skill supply, and makes some useful recommendations and these were discussed in detail earlier on. There are a number of big data strategies found in a number of government reports introduced at the beginning of this post, in addition to this one, but what is sorely missing is an overall strategy that would be of benefit to the Republic, that would tackle issues of importance to Irish society, be in the public interest, while also encouraging and supportive of private sector investment.
3. Experiences as a producer, consumer and observer of open data (Slides, Bio), By Dr Peter Mooney, Environmental Protection Agency (EPA) funded Research Fellow at the Department of Computer Science, NUIM
At the ProgCity open data event I was talking to Denis Parfenov and Flora Fleischer from Open Knowledge Foundation Ireland about some of our AIRO experiences at trying to leverage data out of government departments and to push forward an open data agenda in Ireland. I said I would share a slide I first presented a few years ago about the ostrich attitude to data and evidence-informed policy within many government departments and the arguments used against opening and sharing data and providing open access analysis tools and training.
These arguments were presented to me in one government department in a sequence as I rebutted each assertion. The last two capture perfectly the way that Ireland works politically and institutionally.
We don’t need open data – officials on the ground know what’s going on their area/domain.
Anybody who knows what they’re doing in government can access, process and interpret relevant data.
The data is too sensitive to share and might used in ways for which it wasn’t intended.
The potential gains with respect to increased understanding, and more effective and efficient government are over-stated, and it’s not financially viable.
Just because you have data it doesn’t mean it’ll get used; that’s not how Irish policy is made.
Even if it shapes policy, policy is not implemented or enforced in Ireland.
The whole line of reasoning basically leads to: what is the point of opening data when it won’t make a blind bit of difference as to how the country is run and it’s only going to create additional work and be an annoyance? Given the government has signed up to the open government partnership and to providing open data it’ll be interesting to see to what extent these attitudes persist amongst agencies and politicians (who equally dislike hard evidence as an inconvenience to gombeen politics). Today, the government announced that postcode data will not be open, which is not a great start.
It is clear from recent attempts to make freedom of information requests more difficult that, with a few exceptions within some units, the government is not so much interested in open data for the purposes of transparency, reform and to be held to account, but rather the hope that they might be leveraged economically to create apps, new data products and jobs. If that can be done in a way that does not get in the way of normal political and policy business I suspect they’d be delighted.
Hopefully, the open data movement will not get stifled by the sixth point: even if we have an open data policy it does not mean it’ll get implemented, or if it does it’ll take a long time and be limited. To go back to postcodes, that has been in the pipeline for at least ten years, with endless false starts and consultations. There is no reason why open data should take that long, but it wouldn’t surprise me if it did.
