Open Data and AI Policy (FI0033)

The commitment and measures on open data contribute in particular to the OGP initiatives of transparency and technology & innovation. The measures are in line with the Act on Data Management in Public Administration, data policies and implementation thereof.

Open data:
The public sector promotes government openness by opening public interfaces if there are no specific reasons to keep them restricted. The process will prioritise the most influential data resources. Easy-to-use, developer-friendly interfaces that follow the outlines of standard architecture will be developed to access public administration data resources. The data available on the interfaces will be recorded using standard procedures into a machine-readable and -interpretable format in order to make it easier to benefit from. The guidance needed in order to open the data and create the interfaces will be provided. A successful user experience of the data and its easy accessibility will lower the threshold of using the data and increase its usage. As a result, data producers will be more motivated to improve the quality and usability of data and data resources. The objective of this is to achieve a virtuous circle in which the quality of data will be improved and it will provide benefits, and the benefits of open data will spread far and wide inside the society, also when companies join to become producers of open data. These measures will acknowledge the outlines and development measures of European Union within this theme, including the Open Data Directive (EU/2019/1024), also known as the PSI directive.

The requirements concerning data protection and the data confidentiality regulation will be taken into account when plans are made for opening up data.

The project will be completed in a cross-administrative effort of cooperation, so that the cumulation of data will not put the data security of public administration at risk.
Quality criteria: Quality criteria intended to facilitate the utilisation of data will be prepared.
Quality improvement measures following the quality criteria will be primarily applied to the most significant data resources in the data opening process. Setting quality criteria and the extent of their validity will be planned to make up part of putting the quality criteria into practice.

Ethical guidelines: A general set of guidelines on the ethical use of artificial intelligence will be prepared in order to ensure that the artificial intelligence will not utilise directly or indirectly discriminatory operational models in the AI systems.
The measures to open public sector data will promote a data and AI policy that is ethically, financially and socially sustainable. Meta data that contributes to data resources management of high quality will also contribute to the creation of unified information resources required by machine-learning and AI in our linguistic area and, subsequently, the realisation of linguistic rights in an indirect manner.

Special groups will be consulted and the standards laid down in international human rights conventions and UN recommendations on the ethics of AI as well as data security questions will be acknowledged as a part of the preparatory process.

4. Open Data

Main Objective

“The commitment and measures on open data contribute in particular to the OGP initiatives of transparency and technology & innovation. The measures are in line with the Act on Data Management in Public Administration, data policies and implementation thereof.”

Milestones

• Open data: The public sector promotes government openness by opening public interfaces if there are no specific reasons to keep them restricted. The process will prioritize the most influential data resources. Easy-to-use, developer-friendly interfaces that follow the outlines of standard architecture will be developed to access public administration data resources. The data available on the interfaces will be recorded using standard procedures into a machine-readable and -interpretable format in order to make it easier to benefit from. The guidance needed in order to open the data and create the interfaces will be provided.
• Quality criteria: Quality criteria intended to facilitate the utilisation of data will be prepared. Quality improvement measures following the quality criteria will be primarily applied to the most significant data resources in the data opening process. Setting quality criteria and the extent of their validity will be planned to make up part of putting the quality criteria into practice.
• Ethical guidelines: A general set of guidelines on the ethical use of artificial intelligence will be prepared in order to ensure that the artificial intelligence will not utilise directly or indirectly discriminatory operational models in the AI systems. The measures to open public sector data will promote a data and AI policy that is ethically, financially and socially sustainable. Metadata that contributes to data resources management of high quality will also contribute to the creation of unified information resources required by machine-learning and AI in our linguistic area and, subsequently, the realisation of linguistic rights in an indirect manner. Special groups will be consulted and the standards laid down in international human rights conventions and UN recommendations on the ethics of AI as well as data security questions will be acknowledged as a part of the preparatory process.

Editorial Note: For the complete text of this commitment, please see Finland’s action plan at https://avoinhallinto.fi/assets/files/2019/09/ENGLANTI_Avoin-hallinto_IV_toimintaohjelma_FINAL_240919-1.pdf.

Commitment Analysis

This commitment seeks to enhance public access to information by improving the quality and usability of open data. It also aims to foster the ethical use of artificial intelligence (AI) systems and prevent the use of discriminatory operational models in AI systems.

Open data has been an ongoing priority in Finland’s OGP action plans. During previous action plans, Finland launched a national open data portal (avoindata.fi) and developed guidelines for open data publication and use, as well as open data capabilities in new IT systems. As part of the current action plan, the government is starting a new open data project (2020-2022), which involves, among other goals, formulating strategic objectives for opening up and using public data, developing quality criteria for open data, and establishing an interoperability platform along with tools to support the semantic interoperability of open data. [57] As the commitment focuses on releasing more and better-quality public data, it is relevant to the OGP value of access to information.

Finland generally performs well on open data. The Act on the Openness of Government Activities [58] lays out the basic legal framework for publishing government data and further measures have been taken through government open data programs. During the third action plan (2017-2019), the publication of high-value open datasets, such as public procurement data, improved government openness in a major way. [59] The third action plan also sought to expand the access to information principle to publicly funded enterprises, but the implementation was delayed due to legal obstacles. [60] However, the Ministry of Justice published a report outlining possible models for amending the law [61] and the government program adopted in December 2019 envisages continuing to broaden the scope of the law. The new EU Open Data Directive, which Finland needs to transpose to its national law by 2021, also encourages EU member states to extend open data publication requirements to public undertakings and private companies providing services of general interest. [62]

Finland’s national open data portal currently has more than 1700 datasets from 793 publishers [63] and 6700 monthly visits. [64] Stakeholders say the portal works very well, [65] although some organizations prefer to publish data on their own repositories. [66] The European Data Portal’s (EDP) 2019 Open Data Maturity survey ranks Finland 9^th in Europe. [67] However, in the 2019 OECD’s OURdata index, Finland’s scores decreased in data availability and government support to reuse. According to the OECD, this decrease was due to a reduction in stakeholder engagement and lack of government’s attention to fostering data reuse outside the public sector. [68] Teemu Ropponen from Open Knowledge Finland (OKFI) notes that one of the key problems is the lack of continuous dialogue mechanisms between open data providers and users, which would help data providers better understand the value of their data and adapt their data publication efforts to users’ needs. [69]

Finland also has room for improvement regarding accessibility of key datasets. A 2018 report by Transparency International Latvia and OKFI found that only five out of 10 key anti-corruption datasets could be accessed through the central open data portal. [70] In addition to lobbying data, which will be addressed in Commitment 3, the public currently lacks access to beneficial ownership data. [71] Furthermore, the government has not proactively opened all relevant data on the COVID-19 pandemic. OKFI raised the issue and filed several Freedom of Information requests asking the National Institute for Health and Welfare (THL) to publish the source codes of the models the government used for forecasting the progress of the pandemic in Finland. [72] THL eventually only published the models after the Chancellor of Justice [73] began investigating the issue [74] and the Prime Minister called for respecting the principle of openness in publishing data related to government decisions. [75] However, THL rejected the FoI requests to publish the source code of the models. [76] According to Teemu Ropponen (OKFI), the crisis has revealed the need to increase government officials’ awareness of the benefits of government openness. Due to the adoption of the General Data Protection Regulation in 2018, public officials have generally become more cautious about opening data for fears of breaching privacy regulations. [77]

If implemented as planned, this commitment could improve the quality and accessibility of datasets with high economic and societal value. The government plans a comprehensive package of measures, from agreeing on common strategic objectives and quality criteria to providing guidelines and tools for enabling data interoperability. The new EU Open Data Directive will likely strengthen the impact of these activities, as it involves defining high-value datasets at the EU level by 2021, which all member states need to publish as open data. The data categories defined as having high value also include information on company ownership. This may give an impetus to the Finnish Government to also open beneficial ownership data. According to Riitta Autere from the Ministry of Finance, the government will engage stakeholders in defining additional priority datasets at the national level to promote the use of public data in decision making, business, research, and civic engagement. [78] However, the current plans do not include any activities to promote the actual use of the published data or a dialogue between data providers and users. Since stakeholders see the lack of two-way communication as a major gap, the overall impact of this commitment on access to information will likely remain minor. To fill this gap, the government could consider implementing the Helsinki region’s “Helsinki Loves Developers” (Hel<3Dev) model at the central government level. Hel<3Dev is an initiative providing an open platform for discussion and co-creation between data providers and users, comprising a dedicated website, Facebook discussion group, and regular meetups. [79] OKFI highlights this as a successful case of bringing data holders and users together around shared goals. [80]

Regarding the responsible use of AI, the Ministry of Finance initially planned to prepare guidelines to support the ethical use of AI solutions. [81] Although existing law protects equality and non-discrimination, [82] the government’s first AI Program [83] in 2017 produced a report which called for the adoption of ethical guidelines for developing algorithms and architectures to avoid biases and adverse effects on human dignity and equality. [84] Due to the COVID-19 crisis, the ministry has postponed development of the ethical guidelines for the time being. [85]

The Finnish Center for Artificial Intelligence (FCAI) Society, an interdisciplinary group of experts on AI, considers the adoption of ethical guidelines as useful but not sufficient to ensure the ethical development of AI. [86] They see a need for adopting new methods of data collection and use, provision of more high-quality open government data, revision of regulations, funding for long-term interdisciplinary research on the risks of AI, better AI awareness among government employees, and involvement of the public in discussions around AI. According to the FCAI Society, it is important to involve critical perspectives from civil society in the regulatory design. To enable this, the government should provide CSOs with the necessary resources to participate. Moreover, the FCAI Society believes a more balanced view of AI is needed: while AI entails risks, it could also be an enabler of open government goals such as understandability, participation, and transparency, and could help detect discriminatory practices. The Society, therefore, recommends the government to provide CSOs access to high-quality open government data and tools to use the data for developing AI solutions. They also recommend the government investigates how to provide access to background data behind algorithm-based decisions. In further work on AI ethics, the government could consult the ethical guidelines of the EU High-Level Expert Group on Artificial Intelligence [87] and follow a sandboxing approach to test ideas before going forward. [88]

1. describing an administrative issue or tool? (E.g., “Misallocation of welfare funds” is more helpful than “lacking a website.”)
2. Status quo: What is the status quo of the policy issue at the beginning of an action plan? (E.g., “26 percent of judicial corruption complaints are not processed currently.”)
3. Change: Rather than stating intermediary outputs, what is the targeted behavior change that is expected from the commitment’s implementation? (E.g., “Doubling response rates to information requests” is a stronger goal than “publishing a protocol for response.”)

Starred commitments

One measure, the “starred commitment” (✪), deserves further explanation due to its interest to readers and usefulness for encouraging a race to the top among OGP-participating countries/entities. Starred commitments are considered exemplary OGP commitments. To receive a star, a commitment must meet several criteria.

• Potential star: the commitment’s design should be verifiable, relevant to OGP values, and have transformative potential impact.
• The government must make significant progress on this commitment during the action plan implementation period, receiving an assessment of substantial or complete

These variables are assessed at the end of the action plan cycle in the country’s IRM Implementation Report.