AI — The ethical dilemma
One of the standout sessions of our 39th National Conference centred around the ethical dilemmas posed by artificial intelligence (AI).
The expert panel discussing these dilemmas included CSIRO National Artificial Intelligence Centre Director, Stela Solar and Oz Minerals Robotics Technologies Lead, Sue Keay.
The session was chaired by LexisNexis Managing Director Pacific, Greg Dickason.
According to the panellists, as AI use by organisations increases, governance professionals must factor in the ethical implications this technology could have on people and planet.
Greg takes a look at the key themes that emerged during the session.
We’re living through an interesting time in history for many reasons – most of them technologically driven. One that’s unfolding right before our eyes is the real-time creation of our digital future. It’s rife with ethical uncertainties, and – like many things in our lives – data is at the core.
Every business is now a data driven business. We’re constantly gathering and measuring more than ever before. And we’re generating more efficiency as a result by applying AI models on top of that data, which enables us to understand our businesses, customers, and supply chains better.
The magic of AI beguiled us for a while: when we thought about AI, we thought about how it could help cancer diagnoses or beat the world’s best GO players. It became magical, and we all rushed to see how we could apply AI in our businesses and how we could get that magic internally.
But as we’re starting to see the implications of that, we’re also starting to realise that there are ethical dilemmas as a result of AI and using AR models, especially if they’re not properly tested, or if we’re not understanding the biases inherent in how these models were created.
At the very core of the ethics issue is data (no surprises there!). Many AI systems are built with legacy data – that is, data from the past – a week, a month, yesterday. The biases we hold in society will be baked into this data – and there are lots of documented cases of this. So are we comfortable with using that legacy data to train the models that will drive decision making in the future?
We find ourselves navigating an ethical quandary – how do we actually build and guide systems with visions of a better future, rather than replicating the biases of the past? There are no guidelines on how to do this. It becomes about each of us reflecting internally on what the errors in the data could be, what is ethically right, and then steering it forward.
And when we look at issues of AI-related bias in the organisational context, it becomes apparent that businesses should tread carefully when implementing new models. While there tends to be a lot of focus on the unconscious biases of data scientists and engineers who build these models, biases can also exist within the very problem AI is trying to remedy.
A case in point is a legal firm who wanted to adopt AI to identify with greater accuracy which cases would get most quickly through the courts, and be the most profitable. The contractors who began to build the AI discovered that if the company were to implement the algorithms they developed, they would exclusively only take on cases for men. The data had highlighted a bias within the judicial system – that cases for men tend to be resolved faster, with higher payouts.
So when it comes to the way organisations implement AI technology, accountability to govern the responsible adoption, innovation and use of AI really sits with the business itself.
Right now, we’re navigating the ethical issues to do with AI, but that doesn’t mean its governance sits purely in the realm of the hypothetical. Legislation like the Privacy Act, Anti-Harassment and Discrimination Acts have already begun to codify the prevailing ethical norms as they relate to AI. But more work is still needed to create a normative ethical framework for us to continue navigating the unknown, while the system catches up to some of those more defined paths.
LexisNexis is a Gold Sponsor of Governance Institute of Australia’s 39th National Conference.
LexisNexis is part of RELX Group, a world-leading provider of information and analytics for professional and business customers across industries. LexisNexis helps customers to achieve their goals in more than 175 countries, across six continents, with over 10,000 employees.