Spark's Artificial Intelligence Principles

When using AI technologies, we will put people at the heart of our thinking – considering people’s needs and respecting the human rights, autonomy and diversity of individuals.

All Spark employees involved in the creation and design of AI technologies, including technologies that leverage third-party software, must

• consider the end-to-end ethical impacts of the technology in use and the intended purpose; and
• adhere to Spark’s Code of Ethics, which includes requirements for Spark people to conduct themselves with integrity in an open, honest and transparent way.
  

AI technologies rely on data sets to detect patterns and train machine learning algorithms to predict outcomes. Data sets and algorithms that contain intentional and unintentional biases can lead to outcomes that impact people unfairly. We will implement processes to help identify and minimise bias in the data sets, AI technologies and algorithms we use.  This includes overlaying a human lens to review output for potential bias before use.

Deployers of AI technologies are accountable for ensuring that the AI technology is fit for purpose and operating safely, responsibly, reliably and effectively, at implementation and throughout the AI technology lifecycle.

Users of AI technologies are accountable for ensuring that they:

• use each AI technology as intended by the deployers and within guidance aligned to Spark’s AI Principles;
• monitor whether the AI technology is producing valid results and outcomes for the particular use (and adapt or stop usage to ensure appropriate performance); and
• report any unanticipated outcomes or non-performance to the deployer.
  

When designing AI and automated solutions we will adhere to Spark’s Privacy Policy and our Privacy Values. See the Your Privacy at Spark section on our website for more information.

As for all decisions at Spark, when our AI technologies relate to sensitive or impactful topics an appropriately skilled human must be accountable for informed and responsible decision making. In this situation the limits of our solution must be clearly understood and articulated to the user. The data used, the automated decisioning process, and the automated logic behind the recommendation should be clear and transparent to the individual responsible for the human decision and action.

Our AI technologies and algorithms must be explicable by humans and transparent. We must be able to clearly articulate and explain how the solution works and how any automated decisions are made. Spark AI technologies must have a means of logging or capturing data sets used, automated decisions and actions which can be used by humans to review and understand the automated process and actions that led to the outcome.

We will inform customers in advance if they are communicating directly with an AI technology (rather than a real person), and enable customers to easily request contact with a human if needed.

We will inform customers if we have used AI to generate content sent to them without human review.