Artificial Intelligence and Algorithmic Sensitivity: The Case for Technology Ethics

• In Event: Artificial Intelligence and Cognition as Social Praxis

Abstract

In the age of big data, Machine Learning (ML) and Artificial Intelligence (AI) algorithms have ensured improved targeted marketing, cost effective advertising strategies and efficient business decisions. While consumers have benefitted from these technological innovations, the data driven predictability of such algorithms has inherited various human biases and raised ethical and moral questions about algorithmic cognition. In this paper, we address these questions by examining scenarios in which predictive algorithms demonstrate human biases in social and economic life, possibly leading to real harm. Drawing upon technology design theories in ML and AI, Media Studies/STS, we posit that AI predictive algorithmic cognition can better learn human bias by acquiring the human trait of sensitivity. As a design principle that will enable technologies to potentially show a deeper understanding of social situations of individuals, we conceptualize sensitivity as a counter to the marketing strategy of identifying price (in) sensitive consumer behaviour. Using discourse/media analysis of cases like the Target teen pregnancy incident, which led to the privacy invasion of a pregnant teen on the basis of her purchase history, we suggest that the algorithmic design can discriminate positively to demonstrate sensitivity to confidential aspects of consumer identity derived from personal data. We suggest the use of implicit memory and associations to be built into the word embedding models for better detection of biases.  The paper brings together concepts in ML design and theories of technology ethics, privacy and social identity in media and technology studies to connect diverse disciplinary expertise within STS.  

Authors

• Pawan Singh, Deakin University

• Nandini Seth, Indian Institute of Management, Bangalore