A Conceptual Model for Context Awareness in Ethical Data Management

TL;DR

This paper addresses context-aware ethical data management by introducing Context Dimensions Tree (CDT) and Ethical Requirements Tree (ERT) to capture the contextual and ethical facets that guide data preprocessing. It surveys context-oblivious data-bias mitigation techniques and then proposes a methodology to derive a Contextual View and an Ethical View for each context, applying a transformation before data analysis or learning. A detailed promotion example demonstrates how equity, fairness, and privacy considerations can drive data replications, suppressions, or repairs to reduce discrimination. The work highlights theoretical and practical challenges in quantifying ethics, and outlines directions for future work in abstracting and mapping beliefs to technical measures.

Abstract

Ethics has become a major concern to the information management community, as both algorithms and data should satisfy ethical rules that guarantee not to generate dishonourable behaviours when they are used. However, these ethical rules may vary according to the situation-the context-in which the application programs must work. In this paper, after reviewing the basic ethical concepts and their possible influence on data management, we propose a bipartite conceptual model, composed of the Context Dimensions Tree (CDT), which describes the possible contexts, and the Ethical Requirements Tree (ERT), representing the ethical rules necessary to tailor and preprocess the datasets that should be fed to Data Analysis and Learning Systems in each possible context. We provide some examples and suggestions on how these conceptual tools can be used.

Figures (2)

• Figure 1: A simple Relational Database schema
• Figure 2: CDT (A), ERT (B) and Ethical View Production for a personnel management application