TVL Risk model

Netherlands Enterprise Agency

The algorithm is used to make an automated risk assessment for all Applications for Fixed Chargesgemoetkoming, prior to automated or manual granting and payment of the advance.

Last change on 30th of May 2024, at 12:57 (CET) | Publication Standard 1.0

• Publication category: High-Risk AI-system
• Impact assessment: Field not filled in.
• Status: In use

General information

Theme

• Economy
• Social Security

Begin date

2020-12

Contact information

Voor contact over deze registratie neem contact op met RVO via het contactformulier; https://www.rvo.nl/onderwerpen/over-ons/contact/formulier onder vermelding van Bedrijf/instelling, Covid-19: TVL/TOGS/ATE, of Overheidsorganisatie, Covid-19 regelingen.

Responsible use

Goal and impact

With the deployment of the algorithm, decision making is accelerated so that as many entrepreneurs as possible can receive liquidity support on time. The algorithm speeds up decision-making for granting relief and it provides risk interpretation for manual flow.

Considerations

In the TVL, the risk model (algorithm) is used to automatically check all applications against the scheme criteria and make a risk classification on improper use. Prior to the advance payment and before the final determination of the grant. This speeds up the decision-making chain so that entrepreneurs can receive liquidity support as soon as possible. The deployment of the risk model enables RFO to make positive decisions on large numbers of subsidy applications in a very short time, so that entrepreneurs who need fixed expense support quickly during corona lockdowns can get it. Furthermore, the risk model enables RFO to process only positive decisions without the intervention of an employee, allowing employees to assess the remaining decisions more quickly. A similar outcome without the deployment of the risk model was assessed as not feasible during the preparation of the scheme: it would require a large-scale scaling-up of specialised staff, which cannot be organised at short notice. Even if it had succeeded, the turnaround times for entrepreneurs would have been much longer without the risk model.

Human intervention

With each new TVL round (application phase as well as determination phase), the risk model is adjusted and resubmitted for approval to an interdepartmental advisory committee (rolling review), which advises the general director of RFO on the application of the risk model. Once it has been put into use, the risk model determines the qualitative and quantitative risk of each application; these factors together form the overall risk category. Category 1 is awarded automatically, from Category 2 10% falls out to manual and the rest is awarded automatically. In case the algorithm identifies an application as high risk, Category 3, it falls out to a manual check by an employee. For example, because the amount is high, the application is incomplete, or there are indications of abuse or improper use. The application is then readied for human review, and the algorithm provides advice to the clerk on what to look out for, such as data that is incomplete or contradictory. This can be useful: the advice helps the clerk get a quick picture of the situation. The clerk then reviews the entire application. The clerk can then approve or reject applications, possibly after a further request for information. A negative decision (no or lower grant) therefore does not follow from the risk model but always from an employee's assessment. In case of deviation from the customer request and for subsidy amounts above a certain limit, this is always done via the 4-eye principle.

Risk management

Field not filled in.

Legal basis

Legal basis TVL scheme: https://wetten.overheid.nl/BWBR0044808/2022-10-19 The House of Representatives was informed about the largely automated implementation through the following progress report, see: https://www.rijksoverheid.nl/documenten/kamerstukken/2021/03/22/voortgangsrapportage-tvl

Operations

Data

Field not filled in.