Novel algorithmic assessment toolkit to guide the development of AI systems that focus on human wellbeing
Author Kathy Nickels
Date 16 May 2024
A team led by Prof Paul Henman from the ARC Centre of Excellence for Automated Decision-Making and Society at the University of Queensland, have developed an internationally novel algorithmic assessment toolkit to guide the development and deployment of AI systems with human wellbeing front of mind.
This toolkit contributes to the important and urgent work of building human wellbeing in a world with AI. It moves beyond the focus on digital harms to one that is positively framed in being trauma aware.
“AI and automated decision making have enormous potential to improve society, but as Robodebt and many other examples show, they can also do great harm,” said Professor Henman.
“This Toolkit was specifically designed for AI developers and deployers to guide them to harness AI and automation to enhance wellbeing.”
The toolkit and its methods of development are outlined in the new report: Building a Trauma-Informed Algorithmic Assessment Toolkit.
Researchers co-designed the toolkit with social service professionals drawing together two fields of research and practice – trauma informed approaches and ethical and accountable AI/algorithms.
The toolkit aims to assist organisations to think through, document and review algorithmic supported services. It includes 100 prompt questions for design consideration across categories of Empowerment and Choice; Collaboration; Trust and Transparency; Safety; and Intersectionality.
Prompt questions range from “are service users aware that an algorithmic system is used?” to “can the service user choose to interact with a human?”
While of particular use for social service organisations working with people who may have experienced past trauma, the tool will also be beneficial for any organisation wanting to ensure safe, responsible and ethical use of automation and AI.
Researchers have successfully piloted the toolkit across case studies including Robodebt, Allegheny County Family Screening Tool and chatbot Tessa with plans to work with partner organisations to apply the toolkit to real live case studies.
Researchers
- Paul Henman, Professor of Digital Sociology and Social Policy at the University of Queensland
- Suvradip Maitra, a practising lawyer and researcher in ethics of AI, data and algorithms and senior research assistant at the ADM+S Centre
- Dr Lyndal Sleep, Senior Lecturer in the Queensland Centre for Domestic and Family Violence Research at Central Queensland University, and Affiliate at the ADM+S
- Suzanna Fay, Associate Professor of Criminology at the University of Queensland
This research was supported by The University of Notre Dame-IBM Tech Ethics Lab 2022-23 Auditing AI funding (Award # 262812UQ) with additional support provided by the Australian Research Council’s Centre of Excellence for Automated Decision Making and Society (CE200100005).
We acknowledge the contributions of Philip Gillingham to this project in its initial stages. We thank all our participants and their organisations for their time and invaluable insights into the formation of the project’s resulting Trauma Informed Algorithmic Assessment Tool.
