De grootste kennisbank van het HBO

Inspiratie op jouw vakgebied

Vrij toegankelijk

Terug naar zoekresultatenDeel deze publicatie

Operationalization of a glass box through visualization: applied to a data driven profiling approach

Open access

Operationalization of a glass box through visualization: applied to a data driven profiling approach

Open access

Samenvatting

The profiles from data-driven profiling applications are a model of the reality. The interpretability of these profiles for end users, e.g. policymakers, is often far from trivial. How and why these models are obtained by the applications are often regarded as a black box. In recent years several profiling applications used by public organizations have led to wrong interpretations of the obtained models and impacted individuals and society adversely. Hence, the research focus has increasingly shifted towards dealing with the trust and interpretability issues of the models. In support of a more careful and proper interpretation of these models, several scholars have advocated a glass box approach that aims at making these models more transparent to end users. In this paper, we operationalize the glass box approach for a Genetic Algorithm (GA) based profiling application. To enhance the interpretability of the models provided by the application, we aim at facilitating the interaction of domain experts with the models. Hereby domain experts can gain insight to the evolvement of the profiles and what happens to the profiles if we change or add a new pieces of information. Adding such an interactive visualization provides more transparency about the derived models, making them more understandable for end users and policymakers. As a result, they can better assess and explain the consequences of those models when they apply to practice.

Toon meer
OrganisatieHogeschool Rotterdam
LectoraatKenniscentrum Creating 010
Gepubliceerd inDennehy D., Griva A., Pouloudi N., Dwivedi Y.K., Pappas I., Mäntymäki M. (eds) Responsible AI and Analytics for an Ethical and Inclusive Digitized Society. I3E 2021. Lecture Notes in Computer Science, vol 12896 Springer, Cham
Datum2021-08-25
TypeConferentiebijdrage
DOI10.1007/978-3-030-85447-8_26
TaalEngels

Op de HBO Kennisbank vind je publicaties van 26 hogescholen

De grootste kennisbank van het HBO

Inspiratie op jouw vakgebied

Vrij toegankelijk