Open Algorithms (NL0031)

• Open Algorithms: Experiences from France, the Netherlands and New Zealand (June 1, 2020)

• 2020 Annual ReportAlgorithmic Transparency in France, Netherlands and New Zealand (March 23, 2021)

Open Algorithms
1 July 2018 - 30 June 2020
Main action owner (organisation) Rijkswaterstaat Centrale Informatievoorziening (Central Information Services)
Description of the action point
Which social issue does the action point seek to address? More and more of the government's management and policy decisions are based on data and algorithms. Currently, decisions based on algorithms are not transparent to private citizens and companies, whereas the consequences of applying these algorithms can be of significant importance for society.
What is the action point? Drafting and mapping frameworks and guidelines for government organisations as a tool for making algorithms openly available. A decision tree will be drafted as a result of these frameworks and guidelines. The aim is to apply these frameworks, guidelines and the decision tree in a pilot while publishing some algorithms.
How will the action point contribute to remedying the social issue? The knowledge of and experience with making algorithms openly available is still limited. However the increase in the use of algorithms, and the importance of algorithms in policy and management decisions, makes it important that knowledge, experience and tools about, and for this, are developed and shared. The action point shows which legal, technical, policy and organisational considerations come into play when deciding whether or not to make such algorithms openly available. These considerations result in a decision tree. Based on this, the ambition is that some algorithms should actually be made openly available.
A task force will be set up with representatives from different government organisations, possibly also from international organisations.
Making algorithms openly available sheds a light on the substantiation and operation of algorithms and can thus make a contribution to both the accountability of government policy and government decisions and their quality.
Why is this action point relevant to OGP values? The action point is relevant to transparency. The action point contributes to the publishing of government information, i.e. algorithms.
The action point is relevant to public accountability. The action point gives the public and community organisations access to management and policy decisions taken by the government.

Additional information There is limited availability of, but also a limited need for, budget for the action point.
It is expected that the lion's share of the work will be the efforts by civil servants which will not be claimed.
An indicative annual amount of €25,000 to €50,000 will be needed, e.g. for legal advice or for the technical realisation of the decision tree.

There are intersections with other government and non-government programmes, such as at the Ministry of the Interior and Kingdom Relations, the Ministry of Justice and Security, and the Dutch National Archives.

Milestone with a verifiable result (please note: SMART) Start date: End date:
Task force prepared. Two meetings organised.
Action plan drafted. 1 July 2018 31 December 2018
Workshop focussing on sharing knowledge.
Completion of the report: mapping and analysis of the playing field; legal, technical aspects; analysis of actors 01 January 2019 30 June 2019
Completion of a draft report on frameworks and guidelines on open algorithms; draft decision tree (graphic); one algorithm open (pilot) 01 July 2019 31 December 2019
Completion of a final report on frameworks and guidelines including decision tree ; one to three algorithms open. 01 January 2020 30 June 2020
3 – 5 workshops during the term of the project aimed at sharing knowledge 01 July 2018 30 June 2020
Contact information
Name of the responsible person representing the main action owner Eric Blaakman
Position, organisational unit Data Management Center Manager, Rijkswaterstaat CIV
Email and phone number Eric.blaakman@rws.nl; +31(0)6-46344475Other actors involved Authorities involved Ministry of the Interior and Kingdom Relations (DIO)

To be determined.

Other organisations or bodies (such as community organisations or the private sector) To be determined.

6. Open Algorithms

Language of the commitment as it appears in the action plan:

Drafting and mapping frameworks and guidelines for government organisations as a tool for making algorithms openly available. A decision tree will be drafted as a result of these frameworks and guidelines. The aim is to apply these frameworks, guidelines and the decision tree in a pilot while publishing some algorithms. [22]

Milestones

6.1. Task force prepared. Two meetings organised. Action plan drafted.

6.2. Workshop focusing on sharing knowledge. Completion of the report: mapping and analysis of the playing field; legal, technical aspects; analysis of actors.

6.3. Completion of a draft report on frameworks and guidelines on open algorithms; draft decision tree (graphic); one algorithm open (pilot).

6.4. Completion of a final report on frameworks and guidelines, including decision tree; one to three algorithms open.

6.5. 3-5 workshops during the term of the project aimed at sharing knowledge

Start Date: July 2018     

End Date: June 2020

Context and Objectives

In 2014, the Rathenau Institute [23] published a report stating that, at that point, government, regulators, businesses, and society at large were insufficiently equipped to deal with many new digital challenges. The report argued that transparency over algorithmic decision-making is increasingly important in order to prevent their possible manipulation. [24] Two years later, the Netherlands Scientific Council for Government Policy (WRR) published a report providing recommendations to the Dutch government on how to deal with the increasing role of big data, artificial intelligence, and algorithmic decision-making vis-à-vis privacy, security, and transparency. [25] The Dutch government concurred with most recommendations, including that algorithms used for big data-analyses should be ‘appropriate’ and meet certain criteria, and are preferably scientifically validated. The government also agrees that algorithms need to be transparent for reasons of oversight and legal supervision. [26] The extent to which algorithms are used, however, appears to be unknown. In October 2017, following debate in parliament where concerns were expressed over possible bias in algorithms and whether specific regulation should be put in place, the government committed to stocktaking and mapping the use of algorithms in government practice, including eventual challenges and dilemmas. [27]

Against this backdrop, the third action plan features a commitment on open algorithms and has been inspired by an introspective view by civil servants themselves. The commitment text mentions how big data is increasingly impacting decision-making in the public sector and how that currently is not a transparent process. Such reflections, and the ensuing self-appointed internal investigation on how technological solutions that support public decision-making can be more transparent, are relevant for the government on the use of such technologies going forward. It is therefore relevant to the OGP values of access to information and technology and innovation for transparency and accountability.

Through this work, involved organizations intend to draft guidelines for government agencies on how to make algorithms openly available and develop a decision tree to assist in such a process. The planned activities are specific and verifiable, however only generally. Planned activities, such as the organizing of workshops and mapping the playing field (6.2), as well as drafting a report on guidelines and a pilot open algorithm (6.3, 6.4), do not explain in detail the commitment’s broader objective, or what criteria will drive the referenced decision tree for selecting the pilot algorithm. In addition, from the text it is unclear if linkages exist to the broader policy context and the studies and discussions mentioned above. While the commitment, as written, is unlikely to significantly change thinking around how algorithms are used in central government decision-making, the activities could provide an important first step towards greater algorithm transparency. The potential impact is therefore considered moderate.

Next steps

While recognizing that algorithmic transparency is a highly complex matter, both from a technical and political point of view, the IRM researcher recommends the following steps:

• Involved stakeholders could draw more on existing bodies of domestic work in this area, and where possible join forces so that duplication of efforts can be prevented, and valuable lessons learnt can directly feed into broader policy discussions at the national political level;
• Consider refining and sharpening the objectives of the work towards pioneering algorithmic accountability and transparency, including a mapping (or drawing on the mapping exercise commissioned by the government) of the most impactful or important algorithms influencing government decision-making (high-value datasets) and specifically target such algorithms in future work, as well as explicitly seek to audit such algorithms for undesirable results or bias;
• Following the 2014 Rathenau report, the IRM researcher suggests exploring what it would take for citizens to be made ‘future-proof’ and directly involve citizens and CSOs in this work. That way, we can better assess what skills or knowledge citizens are missing that inhibits their developments of ‘technological citizenship’. [28]

[22] The complete text of this commitment, https://www.opengovpartnership.org/wp-content/uploads/2018/11/Netherlands_Action-Plan_2018-2020_EN.pdf

[23] The Rathenau Institute is an independent Dutch research institute managed under the auspices of the Royal Netherlands Academy of Arts and Sciences, funded mainly by the government. https://www.rathenau.nl/en/about-us/who-we-are

[24] https://www.rathenau.nl/nl/digitale-samenleving/maatschappij-niet-klaar-voor-digitale-samenleving

[25] https://english.wrr.nl/publications/policy-briefs/2017/01/31/big-data-and-security-policies-serving-security-protecting-freedom

[26] https://www.tweedekamer.nl/kamerstukken/detail?id=2016D43893&did=2016D43893

[27] https://www.tweedekamer.nl/kamerstukken/detail?id=2018D16670&did=2018D16670

[28] https://www.rathenau.nl/nl/digitale-samenleving/technologisch-burgerschap-de-democratische-uitdaging-van-de-eenentwintigste

Commitment 6. Open Algorithms

Limited

A taskforce on open algorithms was established and held two meetings during the action plan period. [12] The taskforce planned to consult with a broad range of stakeholders from academia, CSOs and others. However, varying priorities and limited availability among involved staff made it difficult for the work to materialize. There is no evidence available to determine that the other planned activities under this commitment were completed, such as a draft report on frameworks and guidelines on open algorithms, draft decision tree, and opening one to three algorithms. Informal contacts are still in place and the theme of algorithms has been carried over to the fourth action plan (2020-2022).