Human life has transferred online through a data deluge

“data deluge”

Rome [ENA] The growth of data has been offset by the development of technologies for storing and processing data and allowed an efficient use of vast resources. Artificial intelligence and data-analytics can exploit such vast datasets to extract information. Governments, more and more, need to improve their capability to access data relevant to defining and implementing policies, and to process such data as required.

Government’s performance across different dimensions such as detecting social issues, predicting the effect of policies through analysis and simulation, and supporting the tailored decision of individual cases could be improved by the combination of AI and Big Data. Improvement in the efficiency and effectiveness of public action can be achieved through a lawful and ethical use of data in the public sector. A “data deluge” has increased at an accelerated pace, as human life has transferred online, and most economic, administrative and social processes are computer mediated. Simultaneously the cost of data storage and processing has decreased considerably.

More data have been stored in the last few years than in all humankind’s pre-computer history. The collection/generation, curation and processing of data involves costs, as well as benefits. The costs can be avoided or reduced by reusing and repurposing the data. In particular, data collected for the purpose of implementing a policy can be used again (at no collection cost) for the purpose of evaluating the policy and possibly reforming it. The use of data must comply with law and ethics, in order to respect the rule of law, align with the collective and individual good, and contribute to fostering citizens’ trust, which in turn facilitates the collection and reuse of data. Computing techniques can be applied to (big) data for the purpose

of descriptive, predictive and prescriptive analytics. Thanks to AI (machine learning), the models used for analytics can be automatically learnt from vast datasets. AI is already helping government detect issues, predict phenomena, and simulate social dynamics. Consequently, it can contribute to improving policy-making and service delivery. AI-based prediction and simulation can increase the ability to anticipate the effects of new regulations. To make data-collection cost-effective and sustainable, data collection and creation policies need to be designed in such a way that new data are automatically produced as side effect of administrative processes.

The use of data in regulatory assessment and evaluation is in principle concerned with aggregate data, so that input data may in most cases be anonymous or at least pseudonymous, and the output data consists in aggregate, non-personal information. Thus, it seems that this processing belongs with the concept of statistical processing under the GDPR. Even when impact assessments rely on advanced statistical or AI methods, they remain highly conjectural, since they involve predicting behavioural changes and further direct and indirect effects of such changes. The uncertainty of ex-ante assessments, even when based on the best methodologies and on adequate datasets, underscores the need for interim monitoring and ex-post evaluation.

Both are needed to check the extent to which ex-ante predictions are confirmed or rather contradicted by subsequent facts, as well as the extent to which unexpected side-effects emerge. Ex-post monitoring and evaluation provide evidence of real outcomes, supports democratic dialogue with evidence, and contributes to making governments accountable toward elected assemblies, such as the European Parliament. This report has discussed the use of data in the public sector, focusing on ex-ante regulatory assessments and ex-post evaluations. A recent In depth Analysis requested by the JURI committee of the European Parliament had as result that: The public sector should catch up with the private sectors in the capacity to collect and use data.

The public sector should catch up with the private sectors in the capacity to collect and use data. Better and larger datasets should support the policy cycle, for the purpose of forecasting, foresight, policy design, assessment, and evaluation. The cost and benefits of data collection, curation, and use should be considered, and addressed by adopting cost-effective solutions, also involving the reuse and repurposing of data. The statistical processing meant to provide aggregate information should be distinguished from the processing of personal data meant to provide individualised outcomes. The opportunities offered by AI for collecting and interpreting data, extracting aggregate information through analytics, and exploring scenarios

and possible developments through simulation, should be exploited. The need to support multicriteria evaluations with appropriate data, according to methods such as the UN Sustainable Development Goals, should be considered. Quantitative metrics should be used whenever possible, consistently with the nature of the data and the goals pursued. Compliance with law and ethics should be ensured, as to uphold the rule of law, respect citizens’ rights, and foster trust. Predictions should be matched against reality through monitoring and evaluation.

An ex-post evaluation should accompany every ex-ante regulatory assessment, making it possible to adjust policies, in a process that contributes to democratic debate and makes for greater government accountability.