TY - JOUR
T1 - Values and inductive risk in machine learning modelling
T2 - the case of binary classification models
AU - Karaca, Koray
N1 - Funding Information:
I would like to thank Owen King, Christin Seifert and the two anonymous?referees of this journal for helpful comments and discussions. Earlier versions of this paper have been presented at conferences and colloquia at Concordia University, the University of Stuttgart and the University of Twente. I thank the audiences at these events for their feedback.
Publisher Copyright:
© 2021, The Author(s).
PY - 2021/12
Y1 - 2021/12
N2 - I examine the construction and evaluation of machine learning (ML) binary classification models. These models are increasingly used for societal applications such as classifying patients into two categories according to the presence or absence of a certain disease like cancer and heart disease. I argue that the construction of ML (binary) classification models involves an optimisation process aiming at the minimization of the inductive risk associated with the intended uses of these models. I also argue that the construction of these models is underdetermined by the available data, and that this makes it necessary for ML modellers to make social value judgments in determining the error costs (associated with misclassifications) used in ML optimization. I thus suggest that the assessment of the inductive risk with respect to the social values of the intended users is an integral part of the construction and evaluation of ML classification models. I also discuss the implications of this conclusion for the philosophical debate concerning inductive risk.
AB - I examine the construction and evaluation of machine learning (ML) binary classification models. These models are increasingly used for societal applications such as classifying patients into two categories according to the presence or absence of a certain disease like cancer and heart disease. I argue that the construction of ML (binary) classification models involves an optimisation process aiming at the minimization of the inductive risk associated with the intended uses of these models. I also argue that the construction of these models is underdetermined by the available data, and that this makes it necessary for ML modellers to make social value judgments in determining the error costs (associated with misclassifications) used in ML optimization. I thus suggest that the assessment of the inductive risk with respect to the social values of the intended users is an integral part of the construction and evaluation of ML classification models. I also discuss the implications of this conclusion for the philosophical debate concerning inductive risk.
KW - Inductive risk
KW - Machine learning
KW - Social values
KW - Underdetermination of model construction
KW - UT-Hybrid-D
UR - http://www.scopus.com/inward/record.url?scp=85118207795&partnerID=8YFLogxK
U2 - 10.1007/s13194-021-00405-1
DO - 10.1007/s13194-021-00405-1
M3 - Article
AN - SCOPUS:85118207795
SN - 1879-4912
VL - 11
JO - European journal for philosophy of science
JF - European journal for philosophy of science
IS - 4
M1 - 102
ER -