Start Date

16-8-2018 12:00 AM

Description

The growth in AI-capabilities and proliferation of AI-enabled artifacts raises questions about unintended consequences of such technologies including the agency problems between intelligent agents and their human principals. This essay demonstrates how the agency theory and the actor-network theory (ANT) offer different, yet complementary views of the issue. Whereas the agency theory is best applied to the mitigation of the agency problem, ANT can be inform our understanding of the heterogeneous goals and interests of IT artifacts. Using an ethnographic mini-case study involving the application of machine learning algorithms to image classification, the essay traces interests inscribed in AI artifacts. The example highlights how interests of sources of training data are inscribed in AI models, and how such interests become apparent when the model is adopted by a user. Implication for future research and practice are discussed.

Share

COinS
 
Aug 16th, 12:00 AM

Interests and Agency in AI: The case of image with Inception 3 model

The growth in AI-capabilities and proliferation of AI-enabled artifacts raises questions about unintended consequences of such technologies including the agency problems between intelligent agents and their human principals. This essay demonstrates how the agency theory and the actor-network theory (ANT) offer different, yet complementary views of the issue. Whereas the agency theory is best applied to the mitigation of the agency problem, ANT can be inform our understanding of the heterogeneous goals and interests of IT artifacts. Using an ethnographic mini-case study involving the application of machine learning algorithms to image classification, the essay traces interests inscribed in AI artifacts. The example highlights how interests of sources of training data are inscribed in AI models, and how such interests become apparent when the model is adopted by a user. Implication for future research and practice are discussed.