Digital Governance: Automated Decision-Making, Algorithms, and Artificial Intelligence

Automated Decision-Making, Algorithms, and Artificial Intelligence

Governments are increasingly using automated decision-making (ADM) to assess eligibility for government benefits, detect fraud, and allocate resources. ADM can potentially make governments fairer, more efficient, and more effective. To do so, however, it needs appropriate oversight and safeguards. Without safeguards, ADM can reproduce or amplify existing biases in denying welfare benefits, misidentifying innocent people, or misdiagnosing illness. In the context of private sector activity, the use of ADM systems and algorithms also benefits from safeguards that balance innovation with protection of rights. Improving transparency, participation, and accountability can help maximize the benefits and minimize the harms.

Key Terms The Evidence Reform
Guidance Active OGP
Partners Benchmarking
Data Guidance
& Standards

Open Gov Challenge: Digital Governance

With OGP’s 2023-2028 Strategy, OGP members are set to work toward a number of aspirational thematic reforms through the Open Gov Challenge. This section of the Open Gov Guide addresses Digital Governance.

Challenge prompt: Strengthen transparency and public oversight of AI, automated decision-making, and data protection frameworks.

Actions and reforms that fit within the scope of this challenge area are:

Making underlying data publicly available (e.g. through transparency registers for algorithms).
Embedding human rights impact assessments.
Creating public grievance mechanisms.
Creating or strengthening independent oversight agencies.
Creating specific public participation opportunities.
Creating specific mechanisms to promote digital inclusion on AI, automated-decision making, and data protection frameworks and prevent online discrimination and disinformation based on their use.
Reforms that protect personal data and privacy frameworks through establishing transparency, accountability, and consent mechanisms and requirements.

Key Terms

Definitions for key terms such as algorithmic transparency and automated decision-making (ADM).

Algorithmic transparency: Algorithmic transparency is the ability of internal or external actors to obtain information, monitor, test, critique, or evaluate the logic, procedures, and performance of an algorithmic system.
Automated decision-making (ADM): ADM is the processing of personal data using digital means, either without human involvement or in hybrid systems where some portion of decision-making is carried out by algorithms.
- ADM includes algorithmic decision-making, which uses artificial intelligence (AI) algorithms to process data and conduct statistical analyses to automate or support a decision-making process.
Data Protection Authority (DPA): In many jurisdictions, the DPA oversees and regulates algorithmic decision-making and personal data processing. Powers may include issuing regulations and guidelines, hearing complaints, and investigating issues of concern.

SHOW KEY TERMS

HIDE SHOW KEY TERMS

The Evidence

Because it is a relatively new area of work, more research is needed to understand the most effective open government approaches to govern ADM. Early research, however, shows that increasing transparency, civic participation, and public accountability have a positive effect.

Public scrutiny of algorithms has been shown to help identify bias according to social status, gender, and other characteristics.
Explanations of how ADM systems work can build trust with the public and create avenues of participation, with recent participatory initiatives showing that members of the public can investigate highly technical topics.
ADM has been used effectively to support open government practices, such as detecting fraud or corruption. For example, Valencia, Spain’s SALER system allows officials and members of the public to flag conflicts of interest in the public procurement process.

Reform Guidance

The recommendations below represent reforms that national and local governments, representatives of civil society organizations, and others can consider for their action plans and the Open Gov Challenge. The reforms are categorized according to OGP’s principal values: transparency, civic participation, and public accountability. Reforms should be adapted to fit the domestic context, and involve and coordinate with other levels and branches of government.

Reforms across policy areas are also tagged by the estimated degree of difficulty in implementation. Though progress is often not linear, the recommendations have been categorized using these labels to give the reader a sense of how different reforms can work together to raise the ambition of open government approaches.

Recommended Reforms Key

OGP Values

Transparency: Transparency empowers citizens to exercise their rights, hold the government accountable, and participate in decision-making processes. Examples of relevant activities include the proactive or reactive publication of government-held information, legal or institutional frameworks to strengthen the right to access information, and disclosing information using open data standards.

Civic Participation: When people are engaged, governments and public institutions are more responsive, innovative, and effective. Examples of relevant initiatives include new or improved processes and mechanisms for the public to contribute to decisions, participatory mechanisms to involve underrepresented groups in policy making, and a legal environment that guarantees civil and political rights.
Public Accountability: Public accountability occurs when public institutions must justify their actions, act upon requirements and criticisms, and take responsibility for failure to perform according to laws or commitments. Importantly, public accountability means that members of the public can also access and trigger accountability mechanisms. Examples of relevant activities include citizen audits of performance, new or improved mechanisms or institutions that respond to citizen-initiated appeals processes, and improved access to justice.
Inclusion: Inclusion is fundamental to achieving more equitable, representative, and accountable policies that truly serve all people. This includes increasing the voice, agency, and influence of historically discriminated or underrepresented groups. Depending on the context, traditionally underrepresented groups may experience discrimination based on gender, sexual identity, race, ethnicity, age, geography, differing ability, legal, or socioeconomic status.

HIDE OGP VALUES

Degree of Difficulty

Foundational: This tag is used for reforms that are the essential building blocks of a policy area. “Foundational” does not mean low ambition or low impact. These recommendations often establish basic legal frameworks and institutional structures.
Intermediate: This tag is used for reforms that are complex and often involve coordination and outreach between branches, institutions, and levels of government, with the public or between countries.
Advanced: This tag is used for reforms that close important loopholes to make existing work more effective and impactful. Specifically, “Advanced” reforms are particularly ambitious, innovative or close important loopholes to make existing work more effective, impactful or sustainable. They are often applied in mature environments where they seek to institutionalize a good practice that has already shown results.

HIDE DEGREE OF DIFFICULTY

Branch of Government

Executive: The executive branch of government is responsible for designing, implementing, and enforcing laws, policies, and initiatives. It is typically led by the head of state or government, such as a president or prime minister, along with their appointed cabinet members. The executive branch’s functions also include overseeing the day-to-day operations of the government, managing foreign affairs, and directing the country’s armed forces. In democratic systems, the executive branch is accountable to the legislature and the electorate, with its powers and limitations outlined in the constitution or legal framework of the respective country.
Legislative: The legislative branch of government is responsible for making laws and regulations and overseeing the functioning of the government. It typically consists of a body of elected representatives, such as a parliament, congress, or assembly, which is tasked with proposing, debating, amending, and ultimately passing legislation. The legislative branch plays a crucial role in representing the interests of the people, as its members are elected to office by the public. In addition to law-making, this branch often holds the power to levy taxes, allocate funds, and conduct certain investigations into matters of public concern. The structure and powers of the legislative branch are usually outlined in a country’s constitution or legal framework, and it serves as a check on the executive and judicial branches to ensure a system of checks and balances within a state.

HIDE BRANCH OF GOVERNMENT

Recommended Reforms

Policy Area	Recommendation	OGP Value	Degree of Difficulty	Government Branch
Foundations	Create a legal framework for data protection Establish basic framework laws on data protection, including the establishment or enhancement of a DPA. Policy Area Foundations OGP Value Public Accountability Transparency Degree of Difficulty Foundational Government Branch Legislative	Public Accountability Transparency	Foundational	Legislative
Foundations	Establish agency-level guidance Establish basic guidance and shared terminology on technology design and adoption for agencies to follow in terms of procurement to ensure that community needs are met, including fairness, transparency, privacy, and human rights. Such standards should be publicly available, reported on, and developed in consultation with community input. Policy Area Foundations OGP Value Public Accountability Transparency Civic Participation Degree of Difficulty Foundational Government Branch Executive	Public Accountability Transparency Civic Participation	Foundational	Executive
Foundations	Create a public interface for DPAs and other regulators Ensure that the DPA and any other relevant regulators are capable of receiving, investigating, and prosecuting (or referring) cases of harm. This may also include expansion of subpoena and inspection powers by the DPA. Policy Area Foundations OGP Value Public Accountability Transparency Civic Participation Degree of Difficulty Foundational Government Branch Legislative	Public Accountability Transparency Civic Participation	Foundational	Legislative
Foundations	Report regularly on high-risk activities Authorize the DPA to identify major high-risk activities inside and outside of government and to make referrals to other agencies for regulatory or other action. This includes submitting regular reports to the legislature on data protection and other topics under the DPA’s mandate. Policy Area Foundations OGP Value Transparency Degree of Difficulty Foundational Government Branch Legislative	Transparency	Foundational	Legislative
Foundations	Mandate the use of impact assessments Authorize the DPA to mandate the use of impact assessments for agencies under its purview. This can identify potential bias or discrimination in an ADM system before it is put into use, and can include mechanisms for the public to give input on and access such assessments. Canada’s Directive on Automated Decision-Making provides an example of what this looks like in practice. Policy Area Foundations OGP Value Transparency Civic Participation Degree of Difficulty Intermediate Government Branch Legislative	Transparency Civic Participation	Intermediate	Legislative
Foundations	Establish guidelines to purchase AI systems Establish and publish basic guidelines for government procurement of AI systems, such as assessing risks and creating accountability mechanisms before purchase. Such guidelines can include requirements for the formation of a multi-stakeholder oversight group to advise and report on AI procurement. Policy Area Foundations OGP Value Transparency Degree of Difficulty Intermediate Government Branch Executive	Transparency	Intermediate	Executive
Foundations	Train public officials on ADM Provide training for public officials on how ADM systems work, including the general capacity building needed to understand the basics of ADM technology as well as knowledge of the specific systems used in a given agency. Policy Area Foundations OGP Value Civic Participation Degree of Difficulty Intermediate Government Branch Executive	Civic Participation	Intermediate	Executive
Transparency	Require notifications of data breaches Policy Area Transparency OGP Value Transparency Degree of Difficulty Foundational Government Branch Executive	Transparency	Foundational	Executive
Transparency	Require notifications for the use of ADM Require notifications to those who are affected by decisions made by an ADM system. This should include mechanisms for the public to challenge automated decisions or to request exclusion from ADM systems, such as in the Republic of Korea. Policy Area Transparency OGP Value Transparency Degree of Difficulty Foundational Government Branch Executive	Transparency	Foundational	Executive
Transparency	Require notice and consent for processing Require disclosure and consent to data processing activities when using online tools, including data collection, storage, use, and modification. Policy Area Transparency OGP Value Degree of Difficulty Government Branch
Transparency	Create an algorithm register Establish or enhance a data processing register. Such registers may focus on government (and contractor) processing or cover non-governmental activities. A register may support the proactive publication of algorithmic impact assessments. Algorithmic transparency may include data on many elements of an ADM system beyond the source code, including the system’s existence, purpose, governing rules and structure, real-world impacts, training data, inputs, and outputs. Policy Area Transparency OGP Value Transparency Degree of Difficulty Intermediate Government Branch Legislative	Transparency	Intermediate	Legislative
Transparency	Require data processing reports Require all ADM systems to disclose information that establishes and regularly renews a clear legal and ethical basis for how they collect, store, share, use, and modify data. This might include establishing individual consent or establishing social license through participatory processes. Policy Area Transparency OGP Value Transparency Degree of Difficulty Intermediate Government Branch Executive	Transparency	Intermediate	Executive
Transparency	Standardize impact reporting Adopt or enhance procedures to standardize privacy and human rights impact assessments. Ensure that the public can scrutinize and give inputs on such reports and raise flags about irregularities. Policy Area Transparency OGP Value Transparency Civic Participation Degree of Difficulty Advanced Government Branch Legislative	Transparency Civic Participation	Advanced	Legislative
Civic Participation	Conduct consultations with the public Conduct consultations where relevant members of the public can inform the design and pilot stages of technology that affects them (such as algorithms used in ADM systems). Examples of this include deliberative models of participation at national and local levels. Ensure that such consultations include targeted outreach to vulnerable communities, who tend to experience negative impacts from data collection and technology use. Policy Area Civic Participation OGP Value Civic Participation Inclusion Degree of Difficulty Intermediate Government Branch Executive Legislative	Civic Participation Inclusion	Intermediate	Executive Legislative
Public Accountability	Create public complaint mechanisms Create public complaint mechanisms (including judicial, quasi-judicial, and ombudsman structures) to report violations of the law. This is necessary to ensure the public has a means to communicate whether and how the social acceptance of a company’s or industry’s practices have changed, which can be used to identify gaps in current legal frameworks. Policy Area Public Accountability OGP Value Public Accountability Degree of Difficulty Intermediate Government Branch Legislative	Public Accountability	Intermediate	Legislative
Public Accountability	Conduct regular audits Conduct regular audits of ADM systems through independent oversight bodies and civic monitors (such as a social audit). Policy Area Public Accountability OGP Value Public Accountability Degree of Difficulty Intermediate Government Branch Executive Legislative	Public Accountability	Intermediate	Executive Legislative
Public Accountability	Publish enforcement and compliance history Make enforcement and compliance history available online, which can be used to support future enforcement actions. Policy Area Public Accountability OGP Value Public Accountability Transparency Degree of Difficulty Advanced Government Branch Executive	Public Accountability Transparency	Advanced	Executive

Examples of Reforms from OGP and Beyond

The following examples are commitments previously made within or beyond OGP that demonstrate elements of the recommendations made above. This is an emerging area of focus for OGP members, with recent commitments made on ADM.

OGP Reforms

CANADA Algorithmic Impact Assessments in ADM: Requires agencies using ADM to conduct and publish impact assessments before the deployment of any ADM system and to update the system when there is a change in its functionality or scope.
KENYA Ethical and Inclusive Digital Transformation: Committed to closing the digital divide to ensure that everyone, especially vulnerable groups, can access government services and public participation opportunities as these activities move online. Also aims to create a framework for the inclusive, safe, and responsible governance of emerging technology, especially artificial intelligence.
NETHERLANDS Guidance for Making Algorithms Accessible: Committed to establishing guidance and a decision tree for agencies, including guidance and tools for making algorithms openly available.
NEW ZEALAND Plans to Implement the Algorithm Charter: Committed to implementing its Algorithm Charter, such as consulting with the agencies that signed the charter to prioritize recommendations related to understanding and addressing the risk of using algorithms in their work.
SCOTLAND, UNITED KINGDOM Transparency Measures for Government Decision-Making: Committed to increasing the accessibility of government data, such as opening data related to government decision-making and developing a public register of AI algorithms.
URUGUAY Multi-Stakeholder Oversight Body for AI Use: Committed to creating a multi-stakeholder oversight body on the government’s use of AI.

Beyond OGP Action Plans

EUROPEAN UNION Requiring Human Intervention in AI Systems: Explicitly requires human intervention in high-risk AI systems, such as facial recognition software and systems to evaluate the eligibility of natural persons for public benefits and creditworthiness.
NIGERIA DPA Process to Increase Auditing Effectiveness: Established a process to license and register authorized auditors, conduct training and provide consulting on data protection compliance.
SOUTH AFRICA DPA Enforcement to Protect Personal Data: Mandated to take steps to limit or stop processing personal data, based on a specified time frame where the controller of data must comply with the DPA’s decision.

The Role of Local Governments

Local governments are frequently the procurers or developers of ADM systems. They play a special role in ensuring that such systems meet their own objectives, including effectiveness, efficiency, fairness, and explainability. These objectives must be made transparent to the public and explained to intermediaries, such as journalists.

Indeed, several cities (such as Amsterdam, New York, and Nantes) have been much quicker than their national counterparts in publicizing their use of ADM, such as through algorithm registers. In the absence of national legislation, provincial and state governments have often led legal reform and innovation in the area of transparency.

Of course, many local governments or agencies may not have the capacity to evaluate certain ADM systems. In this case, they may have a special role to act together to evaluate the merits of different systems and vendors.

Who is working on this topic?

A

Australia

B

Bulgaria

C

Canada

E

Estonia

F

Finland

France

G

Germany

I

Italy

K

Kenya

N

Netherlands

New Zealand

Norway

O

Ontario, Canada

Q

Québec, Canada

S

Scotland, United Kingdom

Spain

U

Ukraine

United Kingdom

Uruguay

This list reflects members with commitments in the “Automated Decision-Making” policy area of the Data Dashboard.

Active OGP Partners

The following organizations have recently worked on this issue in the context of OGP at the national or international level. They may have additional insights on the topic. Please note that this list is not exhaustive. If you are interested in national-level initiatives, please contact research@opengovpartnership.org.

Benchmarking Data

The OGP 2023-2028 Strategy sets out the Open Gov Challenge and aims to provide clear benchmarks for performance through reliable data.

While benchmarks for individual countries and Open Gov Guide recommendations are not yet integrated, for this chapter, interested individuals may rely on the following data sets:

Data for Development (D4D) will release the Global Index on Responsible AI on the quality of AI oversight, including adherence to OGP values of transparency, civic participation, and public accountability.
Data Protection Africa, managed by ALT Advisory, maps data protection laws in 55 countries on the continent.
OGP commitments on this topic can be found on the Data Dashboard.

Guidance & Standards

While the list below is not exhaustive, it aims to provide a range of recommendations, standards, and analysis to guide reform in this policy area.

Ada Lovelace Institute, AI Now Institute, and the OGP Support Unit collaborated on a publication that draws lessons from public sector policy implementation examples related to algorithmic accountability. The resource includes an analysis of the limits of legal and policy mechanisms in ensuring safe and accountable algorithmic systems.
ALT Advisory (South Africa) and the OGP Support Unit published Data Protection in Africa: A Survey of Member Progress, which aims to analyze the context and major barriers to effective data protection among OGP members in Africa. It also makes recommendations that strengthen data protection in the region.
The University of Adolfo Ibáñez and the OGP Support Unit conducted a state-of-the-evidence review of algorithmic transparency and accountability standards and recommendations, which informed the “Algorithmic Transparency” chapter of the 2022 edition of the Skeptic’s Guide to Open Government.
The United Nations General Assembly published a new resolution in March 2024 on seizing opportunities for safe, secure, and trustworthy AI systems for sustainable development.
Connected by Data produced a report on how to secure meaningful commitments on data and AI governance, following a design workshop at the 2023 OGP Summit in Estonia with civil society representatives, government officials, and academics.produced a report on how to secure meaningful commitments on data and AI governance, following a design workshop at the 2023 OGP Summit in Estonia with civil society representatives, government officials, and academics.
The OGP Support Unit‘s Open Algorithms blog series brings together recommendations and examples of good practices from government reformers and civil society members working on algorithmic accountability.

Data for Development (D4D) will release the Global Index on Responsible AI on the quality of AI oversight, including adherence to OGP values of transparency, civic participation, and public accountability.
Many countries have begun considering guidelines to address AI and data protection concerns, with the United Kingdom‘s Data Ethics Framework providing an especially useful example of more formalized guidance.
There are several international and regional agreements and statements related to this topic, including:
- The United Nations General Assembly resolution on seizing opportunities for safe, secure, and trustworthy AI systems for sustainable development (March 2024)
- The Council of Europe Convention on artificial iIntelligence, human rights, democracy, and the rule of law (March 2024)
- The African Union draft policy for its Artificial Intelligence Continental Strategy for Africa (February 2024)
- The draft guiding principles for the UN AI Advisory Body (December 2023)
- The Bletchley Declaration made by countries attending the United Kingdom AI Safety Summit (November 2023)
- A joint statement by Mozilla and civil society on AI safety and openness (October 2023)
- The Santiago Declaration, an agreement among Latin American and Caribbean countries to promote ethical artificial intelligence in the region (October 2023)
- The United Nations Educational, Scientific and Cultural Organization (UNESCO) guidelines for the governance of digital platforms (2023) and recommendations related to the ethics of AI (November 2021)
- The G7 Hiroshima Process
- The Organisation for Economic Co-operation and Development (OECD) AI Principles (2019)
- The G20 principles for responsible stewardship of trustworthy AI (May 2019)

Featured Resources

Broken Links

How can open data help shine a light on political corruption and make political systems fairer and more inclusive? How can we further link people and data to create a chain of accountability? Explore our new report.

13th September 2022

Global Report

The promise of democracy is often defined by the ballot box, where citizens determine who will represent their interests in government. That promise, however, too often fails to translate to the reality of people’s daily lives...

14th May 2019

The Skeptic’s Guide to Open Government – 2022 Edition

Evidence continues to show that open government affects people’s lives. But there are still skeptics who are not aware of all the benefits associated with this approach. Use this guide to convince them to take an open government approach when implementing reforms.

16th May 2022

turquoise-ish25- Justice offices in ogp paper

Justice Policy Series

The OGP Justice Policy Series is a three-part series on Access to Justice, Open Justice, and Accountability for Democratic Renewal. It aims to highlight the important synergies between justice and open government and the ways in which countries can use OGP to make accountable, credible improvements to their justice systems.

14th December 2020

Data Dashboard

Access data on OGP commitments. Filter commitments by location, policy area, and year. Search by keyword.

21st January 2023

A Guide to Open Government and the Coronavirus

This guide is a one-stop shop for the best current resources on how open government projects and approaches can support tackling the pandemic.

25th August 2020

Digital Governance