Ethical AI and Librarianship

A Resource Guide

Using Artificial Intelligence to Make Decisions: Addressing the Problem of Algorithmic Bias

Field Description
Title Using Artificial Intelligence to Make Decisions: Addressing the Problem of Algorithmic Bias
Type Reports
Creator Australian Human Rights Commission (AHRC)
Link
Creation Date 11/24/2020
Last Updated Date --
Summary This technical paper, a collaborative effort by the Australian Human Rights Commission, Gradient Institute, Consumer Policy Research Centre, CHOICE, and CSIRO's Data61, investigates how algorithmic bias arises in AI-based decision making, focusing on its potential to cause unfairness or unlawful discrimination. The report used a simulated case study of an electricity retailer deciding which customers receive market-competitive contracts. The report identifies five scenarios where algorithmic bias can arise: 1) Data reflecting societal inequality. 2) Outdated data reflects past discrimination. 3) Inaccurate labels in training data due to subjective judgment. 4) Underrepresentation of certain groups. 5) Inflexible models failing to account for demographic differences. This report introduces a toolkit of mitigation strategies, including 1) Acquiring more appropriate data to improve representation, though it can be costly with privacy trade-offs. 2) Preprocessing to remove or mask sensitive attributes, though proxy variables may still encode indirect bias. 3) Using more complex models to identify nuanced differences between groups, with higher computational costs. 4) Adjusting AI system outputs directly to improve outcomes for a disadvantaged group, but with accuracy tradeoff. 5) Reassess whether the variable being predicted (e.g., profitability) is appropriate and fair across different groups. Further resources, including a Discussion Paper, Issues Paper, White Paper, and Final Report, are available through a blog article (see the Link field).
Topic Ethical AI
Source and Link Australian Human Rights Commission. https://humanrights.gov.au/
Access Open
Accessibility --
Audience General. Scholars and Students.
Platform or Format Document (.pdf)
Length 73 pages
Geography USA/AUS
Language eng
Description Date 06/23/2025

Ethical AI and Librarianship: A Resource Guide