Catch your crime profile

Design challenge - old “Hoe kan ik door middel van een interactieve ervaring een concrete toepassing van function creep weergeven binnen predictive policing systemen?”

Design challenge - new “Hoe kan ik door middel van een interactieve ervaring een concrete toepassing van function creep weergeven binnen predictive policing systemen?”

substructure

After analyzing and cleaning the dataset I found out that there was a problem in defining the project and what it needed to be made clear for the users. So i had to make a little change in the concept.

The dataset of historical crime data in the Netherlands between 2004 - 2015 from (CBS). The only problem is that the dataset says nothing about individual persons but only tells something about a group of people, with certain properties. So the problem with my concept is i can't just say the chance of you are committing a certain crime is x percent, that would not be reliable because It only says something about a the representation of a group of people.

New concept:

Catch your crime profile by scanning your face and see how your facial features are represented within the dutch crime data. The crime profile will be based on machine learning algorithms and historical crime data.

The machine learning part will be done through an external API service provided by Kairos (https://www.kairos.com/) which can recognize (gender, age, ethnicity) and the dataset of the CBS which contains (suspects; crime group, gender, age and migration background).

To connect the migration background with the predicted ethnicity i had to make decisions how it categorizes your face.

The API detects only the following ethnicities:

• Black

• White

• Hispanic

• Asian

• Other

The datasets contains:

• Migration background

• Dutch background

• Western migration background

• Non-western migration background

• Unknown migration background

CBS says:

Migration background

Person of whom at least one parent was born abroad.

A distinction is made between persons born abroad (first-generation) and persons born in the Netherlands (second-generation). A second distinction is possible between persons with a western migration background and persons with a non-western migration background.

(https://www.cbs.nl/en-gb/onze-diensten/methods/definitions/migration-background)

Dutch background

Person of whom both parents were born in the Netherlands.

(https://www.cbs.nl/en-gb/onze-diensten/methods/definitions/person-with-a-dutch-background)

Western migration background

Person originating from a country in Europe (excluding Turkey), North America and Oceania, or from Indonesia or Japan.

(https://www.cbs.nl/en-gb/onze-diensten/methods/definitions/person-with-a-western-migration-background)

Non-western migration background

Person originating from a country in Africa, South America or Asia (excl. Indonesia and Japan) or from Turkey.

(https://www.cbs.nl/en-gb/onze-diensten/methods/definitions/person-with-a-non-western-migration-background)

Unknown migration background

Connecting data

Connecting the API data from captured camera and (suspects; crime group, gender, age and migration background)

• API reference Black = CBS Dataset Migration background

  • Non-western migration background !

• API reference White = CBS Dataset Dutch background

• API reference Hispanic = CBS Dataset Western migration background

• API reference Asian = CBS Dataset Non-western migration background

  • Western migration background !

• API reference Other = CBS Dataset Unknown migration background

  • Migration background !

By creating this selection i got a lot of control to who will categorized in which category.