This submission outlines our recommendations on how we ensure automated decision-making (ADM) in government is fit for the AI age, and that rules and frameworks stay relevant despite rapid technological development. In particular we will address these questions raised in the consultation paper:
- How should the need for transparency about the use of ADM be balanced with the need to protect sensitive information about the business process and systems?
- What transparency rules would be appropriate to build into the framework?
- Should safeguards be different depending on the risks associated with the use of ADM for a particular decision or administrative action?
We recommend that:
- Regulation around ADM clearly accounts for artificial intelligence (AI) vs. simple algorithms that don’t have learning or iterative capabilities, and have an increased risk threshold for AI activity
- That a core tenet of ADM includes interpretability – the ability for any end user to clearly interpret or explain the processes, logic and eventual result which the algorithms have developed
- Core privacy principles are adopted to apply to ADM, thereby protecting individuals’ data privacy and rights as part of ADM systems
