Please note: The algorithm descriptions in English have been automatically translated. Errors may have been introduced in this process. For the original descriptions, go to the Dutch version of the Algorithm Register.

Summarising legal objection opinions

Generative AI (artificial intelligence) created summaries of existing objection opinions. These support lawyers in their information needs and ensures faster legal assessment of new objections.

Last change on 26th of September 2024, at 11:10 (CET) | Publication Standard 1.0
Publication category
High-Risk AI-system
Impact assessment
DPIA
Status
In development

General information

Theme

Organisation and business operations

Begin date

Field not filled in.

Contact information

algoritmen@amsterdam.nl

Responsible use

Goal and impact

Summaries of existing objection opinions are created. These summaries can be used to enrich the Case Library. This can be done by displaying the core of the case first on the homepage. The Case Library is the internal database for the lawyers of the Legal Bureau (JB). The database contains objection opinions and supports the lawyers in handling new objections.

The summary contains the essence of the objection advice. With its help, lawyers can more quickly assess whether the objection advice matches the information they are looking for. The problem this solves is the long search for the right objection advice, saving time.

Considerations

This algorithm ultimately allows lawyers to consult objection opinions faster. As a result, turnaround times are shorter and review more efficient.

Human intervention

The results can be viewed internally in the objection advice database (the Case Library). Anyone searching for objection opinions will see a summary of the full opinion. The database has a feedback function to report any errors in a summary, which are then adjusted.

Risk management

Data is only viewable in a secure environment. Lawyers only use it as a search function. There is, as indicated, a feedback function to correct any errors in the data. In addition, it will be mentioned that the summary was created with generative AI.

The results of summaries will be randomly checked to avoid any incorrect output from the language model (based on the algorithm). This is done in the process of fine-tuning and will be avoided as much as possible from prompt-engineering techniques.

There are no risks to vulnerable groups, exclusion and profiling

Elaboration on impact assessments

IAMA was not implemented, as the Municipality of Amsterdam implemented the ethics leaflet.

Impact assessment

Data Protection Impact Assessment (DPIA)

Operations

Data

Unstructured textual data of objection opinions.

Technical design

The algorithm is yet to be developed. In outline, data of objection opinions will be retrieved from the dataset. Based on the guidelines we provided, the language model will create a summary using the algorithm.

OpenAI's generative AI model, gpt-3.5-turbo or gpt-4, will be used via Azure to summarise input texts according to yet-to-be-determined instruction-prompt techniques (e.g. few-shot learning). This model can process textual input and create new texts in natural language according to the instruction provided by the user.

External provider

Amsterdam municipality will create this application itself. The code will be open source.

Link to code base

https://statistiek.data.amsterdam.nl/#/projects

Similar algorithm descriptions

  • The algorithm in the software recognises and anonymises personal data and other sensitive information in documents. Governments regularly publish information related to the drafting and implementation of their policies (e.g. based on the Woo). This tool is used to render sensitive data unrecognisable in the process.

    Last change on 9th of January 2025, at 9:23 (CET) | Publication Standard 1.0
    Publication category
    Other algorithms
    Impact assessment
    DPIA
    Status
    In use
  • The algorithm detects and selects text from notarial deeds for monitoring the correct registration of the notarial deed and levying and collecting taxes.

    Last change on 25th of June 2024, at 10:22 (CET) | Publication Standard 1.0
    Publication category
    Impactful algorithms
    Impact assessment
    DPIA
    Status
    In use
  • Among other things, the algorithm identifies and anonymises (personal) data and confidential (financial) information in documents before they are published, as required by the Open Government Act.

    Last change on 16th of May 2024, at 11:52 (CET) | Publication Standard 1.0
    Publication category
    Other algorithms
    Impact assessment
    Field not filled in.
    Status
    In use
  • Among other things, the algorithm identifies and anonymises (personal) data and confidential (financial) information in documents before they are published, as required by the Open Government Act.

    Last change on 12th of September 2024, at 8:23 (CET) | Publication Standard 1.0
    Publication category
    Other algorithms
    Impact assessment
    Field not filled in.
    Status
    In use
  • The algorithm in the software recognises and anonymises personal data and other sensitive information in documents. Governments regularly publish information related to the drafting and implementation of their policies (e.g. based on the Woo). This tool is used to render sensitive data unrecognisable in the process.

    Last change on 20th of November 2024, at 14:27 (CET) | Publication Standard 1.0
    Publication category
    Other algorithms
    Impact assessment
    DPIA
    Status
    In use