Artificial intelligence at Lånekassen

When Lånekassen uses the term artificial intelligence, we refer to computer systems that have been developed to work in a way that enables the system to make certain types of assessments, recommendations and assessments autonomously.

Artificial intelligence can provide major advantages to promote our social mission, but the use of such technology can also entail an increased risk of discrimination and abuse of authority. Such use may also entail an increased risk to the privacy of our customers. Lånekassen is therefore committed to ensuring that we use artificial intelligence in a manner that is lawful, fair and responsible.

Using the machine learning model in residence checks 

When Lånekassen selects who needs to document where they have lived during the 2024-2025 support period, this may be done using a machine learning model.  
The purpose of using machine learning in residence checks is to carry out the checks in a more effective and accurate manner. Using machine learning, Lånekassen will be able to ask significantly fewer customers to document their residential status. Nevertheless, we will still be able to identify the same number of customers who are not living away from home as we would through a major check without the use of machine learning. 
The machine learning model can make recommendations for who we should or should not include in a check. Only a small proportion of those checked are unable to document having lived away from home. Being selected for a check therefore does not mean that we think incorrect information has been provided. Nevertheless, a machine learning model can help us narrow the selection so that we can avoid checking large customer groups that are unlikely not to be living away from home.  
Developing the model is a process in which we examine the information that enables the model to develop the ability to provide accurate recommendations.  
Modellen blir trent på resultat og opplysningar frå tidlegare bukontrollar  som Lånekassen har gjennomført.

The model is trained using results and information from previous Residence controls carried out by the Norwegian State Educational Loan Fund.
The personal data we use includes credits, age, place of study and place of residence. The model examines the factors that increase the likelihood of customers not meeting the criteria, as well as the factors that decrease this likelihood. Lånekassen has also used a machine learning model in previous residence checks. The experiences we gained showed that the key factors for the model include the distance between the parental home and the place of study, the student’s place of residence and how far into their studies the student had progressed. 
For the 2024-2025 academic year, we will use the machine learning model as a support tool for selecting those customers who need to document having lived away from home. The model will provide a recommendation for who to check. We will review the recommendation manually before we determine who needs to submit documentation. We will examine whether the model may have made recommendations that could be discriminatory or unfair to our customers. 
In addition to the group recommended by the model, we will also select a random control group. This means that we will subsequently be able to compare the model’s prioritisation and the control group, thereby gaining information about the quality of the model.  
The actual checking of documentation submitted to Lånekassen in connection with residence checks will be carried out by our case officers.   
Our use of this model is considered profiling pursuant to the General Data Protection Regulation. You can read more about your rights in our privacy statement.