The report can be found here.

Summary

Framework within risk profiles can be assessed on discriminatory practices. Highly relevant for (semi-)automated decision-making since many algorithmic techniques are based on profiling methods.

How to assess whether a risk profile leads to discrimination based on etnicity or nationality?

1. Does a risk profile differentiate based on race or nationality? If no, no discrimination on the basis of etnicity or nationality occurs.

If yes, does unequal treatment occur in comparison to others in a similar situation?

If yes, is unequal treatment based on the basis of etnicity (e.g., skin color, origin, national origin) or nationality?

If yes, does direct differentiation on the basis of etnicity occur?

If yes, the Dutch directive for Equal Treatment applies: Closed system of justification.

If no, open system of justification.

2. Does an objective justification exists for differentiation?

Is etnicity or nationality the only selection criterium for the risk profile?

If yes, prohibited discrimination.

If no, is the risk profile targeted to only one specific etnicity or nationality?

If yes, prohibited discrimination.

If no, does the risk profile contain a selection criterium through which direct differentiation on the basis of etnicity or nationality occurs?

If yes, only in very limited circrumstances justification exists.

If no, indirect differentiation occurs. Does the risk profile pursues a legitimate aim?

If no, prohibited discrimination.

If yes, is this specific risk profile well-suited for the pursued aim? (selection criteria are relevant and objective)

If no, prohibited discrimination.

If yes, is this specific risk profile necessary and proportional? (Reasonable balance between the involbed interests, does not exceed necessary involvement, no reasonable alternatives)

If no, prohibited discrimination.

If yes, no conflict with prohibition of discrimination.

Nieuwsbrief

Blijf op de hoogte van ons werk door je in te schrijven voor onze nieuwsbrief

Nieuwsbrief

Blijf op de hoogte van ons werk door je in te schrijven voor onze nieuwsbrief

Publieke kennisopbouw voor ethische algoritmes