Algorithm Register Template

All fields with example texts and practical tips

This template helps you correctly and completely fill in the Algorithm Register. For each field you will find an explanation, example text and points of attention.

Preview

Basic Information

Name of the algorithm

Required

A clear, recognizable name for the algorithm.

Example text:

"Fraud Risk Detection Benefits"

Avoid abbreviations or internal codes • Make the name understandable for citizens

Short description

Required

A brief explanation of what the algorithm does (max. 500 characters).

Example text:

"This algorithm analyzes applications for social benefits to identify signals that may indicate possible irregularities. The system provides a risk score used by staff when determining follow-up investigation."

Avoid jargon • Explain WHAT it does, not HOW it technically works • Be honest about the purpose

Status

Required

The current status of the algorithm.

Example text:

"Operational"

Choose from: In development / In pilot / Operational / Temporarily out of use / Phased out

Ownership

Organization

Required

The responsible government organization.

Example text:

"Municipality of Amsterdam, Department of Work and Income"

Also mention the specific department or directorate

Contact person/department

Required

Where can citizens or journalists go with questions?

Example text:

"Information Department - algorithms@amsterdam.nl"

Preferably use a functional email address • Not personal details of an employee

Application

Data & Technology

Download the complete template

Receive the complete Algorithm Register Template with all example texts directly in your inbox.

You will also receive our monthly updates on AI governance for governments.

Full content after registration:

ApplicationData & TechnologyHuman Control & Decision-makingRisk AssessmentPublication & Review

Application

Policy area

Verplicht

In which policy area is the algorithm deployed?

Voorbeeldtekst:

"Social security and employment"

💡 Use common policy areas that citizens recognize

Target group

Verplicht

Which group is assessed or affected by the algorithm?

Voorbeeldtekst:

"Residents applying for or receiving social benefits"

💡 Be specific about who is affected • Also mention potentially vulnerable groups

Purpose of deployment

Verplicht

What is the intended result of using this algorithm?

Voorbeeldtekst:

"More efficient deployment of enforcement capacity by identifying applications that may require extra attention."

💡 Be transparent about the actual purpose • Avoid euphemistic language

Data & Technology

Data used

Verplicht

Which data sources and data types are used?

Voorbeeldtekst:

"Personal data from population register, income data (UWV), debt registration (BKR), historical benefit data"

💡 Name all data sources explicitly • State whether personal data is processed • Be complete - incomplete information undermines trust

Type of algorithm

Verplicht

Is it a rule-based system or a learning system (machine learning)?

Voorbeeldtekst:

"Machine learning - classification model trained on historical files"

💡 Rule-based = fixed rules, transparent logic • Machine learning = learns from data, less transparent

Supplier

Is the system developed internally or supplied by an external party?

Voorbeeldtekst:

"Internally developed by municipal Data Science team, in collaboration with University of Amsterdam"

💡 Also mention collaboration partners • For external supplier: name them

Human Control & Decision-making

Role of algorithm in decision-making

Verplicht

How does the algorithm output relate to the final decision?

Voorbeeldtekst:

"The algorithm generates a risk score and signaling list. Staff assess each signal individually before taking follow-up steps. The algorithm does not make independent decisions."

💡 Clearly explain whether human assessment always takes place • Be honest if human control is limited

Possibility of objection

Verplicht

Can citizens object if they are affected by the algorithm?

Voorbeeldtekst:

"Citizens are informed when their file is subject to additional investigation. They can object through the regular objection procedure against decisions."

💡 Explain how citizens can challenge algorithmic decisions

Risk Assessment

Assessments conducted

Which risk assessments have been conducted?

Voorbeeldtekst:

"DPIA conducted in March 2024, IAMA conducted in June 2024, bias audit by external agency in September 2024"

💡 Name all conducted assessments with date • Link to published reports if possible

AI Act risk classification

What is the classification under the EU AI Act?

Voorbeeldtekst:

"High-risk (Annex III - access to social services)"

💡 Use the Risk Classification Decision Tree to determine this

Publication & Review

Publication date

Verplicht

When was this registration published?

Voorbeeldtekst:

"January 15, 2025"

Review cycle

How often is this registration updated?

Voorbeeldtekst:

"Annually, and upon significant changes to the system"

💡 Schedule periodic review • Update upon changes

Disclaimer: This template is based on the Algorithm Register format of the Dutch Government. Always check current requirements at algoritmes.overheid.nl.

Based on the national Algorithm Register format

Version January 2025