PwC to introduce ‘inclusive AI prompting’ to combat bias in Gen AI outputs
The Australian business of Big Four firm PwC will introduce a new set of initiatives aimed at better ensuring more inclusive AI and combating entrenched gendered biases.
The new initiatives come in response to the ‘AI pledge’ campaign led by the Champions of Change Coalition and Women’s Leadership Institute of Australia, which seeks to address how artificial intelligence is designed, accessed, and used.
In addition to signing the pledge and promoting firm-wide engagement, PwC has added the Champions of Change ‘Inclusive AI Prompt’ to its internal generative AI ‘ChatPwC’ platform, which works to challenge biased or outdated assumptions.
“Artificial intelligence draws on historical data, and without the right checks and human judgement, there’s a risk that prejudiced AI outcomes are repeated at scale across operations, decision-making, systems, and services.” commented PwC Australia chief Kevin Burrowes. “These frameworks ensure inclusivity is built into AI as a standard from the start, not as an exception.”
PwC was one of the early-movers in Australia to implement Open AI’s ChatGPT technology into its internal systems to support staff, but wider questions were already being raised back then and even earlier regarding the potential of AI to only further reinforce biases due to the ongoing gendered imbalance of STEM workers, based on the ‘garbage in, garbage out’ or GIGO principle.
As a simple and practical tool to help drive awareness and greater equality, the ‘Inclusive AI Prompt’ can be added to workflows to catch any existing potential biases early on, such as by typing ‘apply a gender equality lens to this task’, which can in effect force chatbots to further ‘reflect’ on their own outputs and highlight more diverse perspectives in providing recommendations.
Burrowes said that he has already been using the prompt over the past month, and found the way AI re-evaluates itself and adjusts to be fascinating; “I’m super impressed with how it explains the changes it makes based on key criteria. The choices we make now about how AI systems are created and deployed will determine whether they expand equality or add to existing inequalities.”
Burrowes continued: “Without clear guardrails, AI could replicate negative norms and stereotypes, but with inclusion at the core of how AI is built, scaled and governed, we can accelerate broader uptake of AI and responsible use practices. This is an important moment for leaders to act decisively and set clear expectations, although you don’t need to be a CEO to make a difference
