Following the Italian National Guidelines ("Linee guida nazionali per la valorizzazione del patrimonio informativo pubblico"), developed within the Data & Analytics Framework project by AgID and the Digital Transformation Team, we have conducted a quality analysis of our source datasets to ensure they are in good condition and suitable for the intended use.
This analysis focuses on four primary factors for assessing data quality:

• Accuracy (syntactic and semantic): the data and its attributes correctly represent the real value of the concept or event they refer to
• Coherence: the data and its attributes do not present any contradictions with respect to other data in the context of use by the administration owner
• Completeness: the data are exhaustive for what concerns every expected value and with respect to the related entities (sources) that contribute to the definition of the procedure
• Timeliness (or promptness of updating): the data and its attributes refer to the "correct time" (up to date) with respect to the procedure they refer to

The following table showcases the quality of each of the source datasets and highlights possible flaws.

When evaluating the ethical aspects of our project's data, we drew inspiration from the Data Ethics Principles and Guidelines and the detailed Ethical Canvas from the ODI project, to ensure we thoroughly considered every ethical aspect in our data management.

Since both our source and mashup datasets exclusively feature information provided by the National Institute for Insurance against Work Accidents (INAIL), our initial focus was on examining the fairness of data collection and management by INAIL. Subsequently, we developed guidelines to address ethical concerns in processing the source datasets for our project.

Data Ethics Principles

• Human being at the center INAIL, in releasing its datasets, does not indicate anywhere on its website the policies followed to respect ethical and legal principles.
  According to information on the ISTAT website, ISTAT's policies align with ethical and legislative principles, focusing on effectively publishing statistical information and analysis results to raise awareness of Italy's conditions and enhance decision-making processes among private and public institutions.
• Equality INAIL, in releasing its open data, does not make an effort to propose the use of statistical methodologies to highlight discrimination and injustices. The data provides all the necessary information to conduct such studies, but these studies are not carried out by the institute itself.
  On the other hand, ISTAT utilizes statistical methodologies to publish data on critical equality issues across various domains.
• Transparency While ISTAT provides comprehensive documentation (SIQual) describing the methodologies for data collection, processing, and storage, INAIL does not provide any information regarding this in the production of its open data.
• Accountability ISTAT's quality policy aligns with the European framework established by Eurostat, adhering to the principles outlined in the European Statistics Code of Practice. This framework ensures and enhances accountability and governance within both the European Statistical System and the National Statistical Systems it encompasses. INAIL does not provide information in this regard, leaving doubts about its methods of accountability.
• Individual data protection ISTAT anonymizes its datasets to prioritize respondent privacy, safeguard data confidentiality, and operate transparently and independently. Collected information is protected by statistical confidentiality and adheres to personal data protection legislation. The INAIL datasets present an "event-oriented" structure, collecting individual events. This approach, despite the anonymization of the participating subjects (injured person and employer), is not compliant with the GDPR, since, especially in cases of serious injuries or deaths, identification can be obtained indirectly by checking the available information online.

Ethical concerns and their management

Despite ISTAT's compliance with ethical principles in data collection and management, the team placed significant emphasis on the ethical handling of sensitive information contained within the datasets.

Data pertaining to age, residence, religious observance, and reproductive health are particularly sensitive, and their ethical handling was carefully considered through the following steps:

• Data integrity and privacy are rigorously maintained by aggregating source dataset values and presenting them as percentage values, thereby preventing any correlation with real individuals.
• To ensure equality and protect vulnerable groups, clear boundaries were set by intentionally excluding certain data provided by ISTAT, such as distinctions based on citizenship, to mitigate potential discriminatory practices.

Anonymization

The data as collected in our final mashed up were not compliant with the General Data Protection Regulation (GDPR). In fact eight out of thirteen contained personal data.
Personal data are defined as: Personal data is any information that relates to an identified or identifiable living individual. Different pieces of information, which collected together can lead to the identification of a particular person, also constitute personal data.

Personal data is any information that relates to an identified or identifiable living individual. Different pieces of information, which collected together can lead to the identification of a particular person, also constitute personal data.

Our data analysis created the final output dataset counting INAIL datasets’ instances of events of injuries and fatalities for each year in each region. In many cases in a specific region in a given year only one event occurred. Therefore, we were producing personal data that indirectly.

While the GDPR does not apply to deceased persons (as specified in Recital 27), Italy has availed itself of the option of proposing a different approach with the harmonization decree of the Privacy Code (Legislative Decree 101/2018), which, through Article 2-terdecies, extended the provisions of the GDPR to the processing of personal data of deceased persons. Therefore we have processed a double approach to ensure that no living individual was going to be identified.
We grouped the result in different layers and at the same time we hidden the unique values signing all values between 1 and 10 as 10.

Source datasets

All source dataset have been evaluated based on the metadata model provided by AGID that classifies metadata quality on a range of 4 levels according to two factors: data-metadata bond and detail level

RDF Assertion and Metadata for the mashup datasets

All produced mashup dataset have been thoroughly described with metadata, following the specification of DCAT-AP_IT standard as recommended by AGID's public information heritage valorization guidelines.

Since all our datasets contain data of specific national interest and are derived from Istat datasets, which is an italian public research institution we decided to adopt DCAT-AP_IT (2016), the national standard.

Moreover:

• To describe in full transparency the sources and the activities underlying the creation of our mashup datasets we adopted PROV-O - the provenance ontology as strongly recommended on a european level and also allowed by DCAT-AP_IT metadata model
• To authoritatively describe the keywords and themes addressed by the datasets, we relied on the vocabularies of EuroVoc and the Publications Office of the European Union.
• We have created a Turtle RDF file to describe the collection of datasets resulting from our mashup work, and an additional Turtle RDF file to describe the documents selected from Normattiva.

FAIR

Finally, we assessed the datasets against the FAIR principles proposed by the GO FAIR Initiative. These principles, advocated by a consortium of scientists and organizations, aim to enhance the Findability, Accessibility, Interoperability, and Reusability of digital assets.