The Algorithm Charter

Humans, rather than computers, review and decide on almost all significant decisions made by government agencies. As agencies continue to develop new algorithms, it’s important to preserve appropriate human oversight and ensure that the views of key stakeholders, notably the people who will receive or participate in services, are given appropriate consideration.

The Algorithm Charter is a commitment by government agencies to manage their use of algorithms in a fair, ethical and transparent way.

For more information about the Algorithm Charter, see:

Algorithm Charter for Aotearoa New Zealand 2020(external link)

What do we mean by 'algorithm'?

The Algorithm Charter doesn’t commit to a specific technical definition of the term ‘algorithm’, as a wide range of advanced analytical tools fall under that term. These range from less advanced techniques, such as regression models and decision trees, which primarily support predictions and streamline business processes, through to more complex systems, such as neural networks and Bayesian models, which can take on properties of machine learning as they make advanced calculations and predictions.

A good discussion of the different types of predictive algorithms and the challenges of defining these is contained in the following report published in 2019 by the New Zealand Law Foundation and Otago University:

Government Use of Artificial Intelligence in New Zealand(external link)

Algorithm use in the Ministry of Justice

Te Tāhū o te Ture - the Ministry of Justice is committed to transparency and accountability in our use of operational algorithms as set out in the Algorithm Charter.

Our Ministry currently uses three operational algorithms within the Collection Services Business Unit1:

None of these algorithms incorporate machine learning capability.

This unit is responsible for the collections of court-imposed fines and reparation, infringements issued by third parties such as New Zealand Police and local authorities, legal aid debt and civil enforcement.

Back to top

Task Generator algorithm

The Ministry receives thousands of fines, infringements, and other debt files to collect daily and therefore needs to prioritise incoming work. The Task Generator (TG) is a series of systems that capture, prioritise and distribute tasks for Collections Registry Officers2 to action.

Why do we use the Task Generator algorithm?

We use this algorithm to automate a repetitive manual process. This enables a larger volume of work to be captured, prioritised and distributed.

What information does the Task Generator algorithm use?

Tasks are prioritised using a points system based on information about the case, including:

  • type and value of penalty
  • availability of information about a participant (for example, address or employer)
  • Ministry priorities (for example, reparation would be prioritised above other collection activity).

What does the Task Generator algorithm do with the supplied information?

It prioritises and distributes tasks to Collections Registry Officers and Bailiffs based on available participant information and Ministry priorities.

Who uses the resulting information provided by the Task Generator algorithm?

Collections Registry Officers and Bailiffs use the information to make decisions and take appropriate action.

What is the Ministry doing differently because of the information provided by the Task Generator algorithm?

We’ve been able to increase our volume of automated prioritisation of task management and allocation for Collections Registry Officers and Bailiffs. 

What is the Task Generator algorithm's assessed risk?

Using the Algorithm Charter Risk Matrix, we’ve assessed the TG algorithm’s risk as follows:

  • Risk Rating: Low
  • Impact: Low (An error would result in tasks being incorrectly prioritised, but is unlikely to negatively affect participant outcomes)
  • Likelihood: Occasional (The Collections Registry Officer regularly reviews the priority list, which reduces the likelihood of errors occurring)

A Collections Registry Officer is responsible for helping participants resolve their fines and civil matters. 

Back to top

Automatic Attachment Orders algorithm

An Automatic Attachment Order (AAO) instructs an employer or Work and Income to transfer money from the debtor’s wages or income support benefit to the Ministry to offset monies owed by the debtor for outstanding fines, reparation or infringements. A list of cases eligible for AAOs is automatically generated each day based on criteria using case information and participant history. Our Ministry has been using this algorithm since 2016.

Why does our Ministry use the Automatic Attachment Orders algorithm?

We use it to automate a repetitive manual process, enabling us to process a higher volume of tasks and improve our efficiency.

What information does the Automatic Attachment Orders algorithm use?

It uses information such as:

  • personal information, including location
  • employment or benefit details
  • fine status and any current repayment arrangements.

What does the Automatic Attachment Orders algorithm do with the supplied information?

It generates a list of cases eligible for AAOs based on case information and participant history.

Who uses the resulting information provided by the Automatic Attachment Orders algorithm?

Collections Registry Officers use the information to quality assure their decision lists.

What is the Ministry doing differently because of the information provided by the Automatic Attachment Orders algorithm?

Our Collections Registry Officers can more efficiently issue AAOs.

What is the Automatic Attachment Orders algorithm's assessed risk?

Using the Algorithm Charter Risk Matrix, we’ve assessed the AAOs algorithm as follows:

  • Risk Rating: Low
  • Impact: Low (Unlikely to negatively affect participant outcomes)
  • Likelihood – Occasional (The algorithm generates lists of suitable cases that would otherwise need to be produced manually. Quality review by the Registry Officer will ensure the appropriate order is attached)

Back to top

Algorithm to identify new contact details for debtors

Te Tāhū o te Ture manages a large volume of outstanding fines, infringements and reparation arrangements. To do this successfully, it’s essential to have up-to-date contact details for debtors. If the Ministry is unable to contact debtors, it will data-match contact details with the Ministry of Social Development (MSD) Inland Revenue (IR) and a credit reporting agency. A list of debtors is electronically submitted to these agencies, where further data-matching algorithms look for additional contact information that may help our Ministry locate the debtor. 

Why does our Ministry use the algorithm to identify new contact details for debtors?

We use it to get more-up-to-date contact details for debtors.

What information does the algorithm to identify new contact details for debtors use?

It uses information such as:

  • personal information, such as first name and date of birth
  • fine status
  • status of any arrangements
  • whether there are any enforcement orders against the debtor.

What does the algorithm to identify new contact details for debtors do with the supplied information?

It generates contact details for the debtor.

Who uses the resulting information provided by the algorithm to identify new contact details for debtors?

Collections Registry Officers use the information to contact debtors. In some cases, automated processes may be activated using the new data, such as sending a letter to the debtor.

What is the Ministry doing differently because of the information provided by the algorithm to identify new contact details for debtors?

Successfully contacting debtors increases our Ministry’s ability to collect monies for outstanding fines, infringements and reparation.

What is the algorithm to identify new contact details for debtors' assessed risk?

Using the Algorithm Charter Risk Matrix, we’ve assessed the algorithm to identify new contact details for debtors as follows:

  • Risk Rating: Low
  • Impact: Low (Unlikely to negatively affect participant outcomes)
  • Likelihood – Occasional (The algorithm generates lists of suitable cases that would otherwise need to be produced manually. Quality review by the Registry Officer will ensure the appropriate people are contacted regarding payment.)

For further information, email: justicestats@justice.govt.nz

Back to top