Ethics: responsibility, explainability, bias and privacy

Introduction

When contemplating on a computational solution to a problem, one must pause and reflect upon both the obvious and the less obvious impacts the proposal may have in the society, the economy, and the environment, whether immediately or after a significant delay. Aspects to focus on include responsible use of technology (fairness, benefit, risk) as well as explainability (the reasoning behind the decisions), possible presence of bias in data sets or assumptions or the ways in which design decisions are taken, as well as possible breaches to the privacy of individuals (either in the building or the employment of the proposed solution).

There is no new assessment in this module. The tools, techniques and concepts of this module contribute to successful completion of upcoming assessments and can also be leveraged in previous assessments that have not yet been turned in.

Learning outcomes

This module will help you do the following:

• Discuss ethical concerns arising from proposed research or the use of its results
• Identify ways in which technology can be developed and used responsibly
• Identify potential sources of bias and plan how to counter them
• Analyze possible implications to the privacy of individuals arising from proposed research or the use of its results

Readings

• Qinghua Lu, Jon Whittle, Xiwei Xu, Liming Zhu: Responsible AI: Best Practices for Creating Trustworthy AI Systems, 2023. https://mcgill.on.worldcat.org/oclc/1412007594 (optional, eBook)
• Joseph Migga Kizza: Ethics in computing : a concise module, 2016. https://mcgill.on.worldcat.org/oclc/949923519 (optional, eBook)
• Avinash Manure , Shaleen Bengani , Saravanan S: Introduction to Responsible AI, 2023. https://mcgill.on.worldcat.org/oclc/1410562127 (optional, eBook)

Tools

Ethics in research involves various principles such as responsibility, explainability, bias, and privacy. Several tools can aid researchers in upholding ethical standards in these areas:

• Data Management Plans (DMPs): DMPs outline how research data will be collected, stored, shared, and preserved. This ensures responsible data handling and future access for verification.
• Peer Review: The traditional method of having colleagues scrutinize research proposals and publications helps identify flaws and ensures research quality and ethical considerations.
• Randomization: Randomly assigning participants to groups helps control for pre-existing differences and ensures a fair comparison between groups.
• Informed Consent: Participants must be fully informed about the research, its potential risks and benefits, and their right to withdraw before agreeing to take part. Consent forms should be clear and easy to understand.
• De-identification: Removing personal identifiers from data helps protect participant privacy while allowing for valuable research to be conducted.