Artificial intelligence (AI) is inescapable in modern life, but businesses and governments are pushing hard to ensure that, whatever purpose it is used for, the technology is used in a responsible manner.
This requires ensuring safety, transparency and accountability all through the process. Given how quickly AI is evolving, responsible governance is vital to ensure users don’t make grievous errors.
“Responsible AI is not simply a property of a model,” says Hippolyte Lefebvre, assistant professor in management information systems at UCD. “It addresses how an organisation designs, deploys, uses and governs an entire AI-enabled system.”
That means the technical brilliance of an AI platform isn’t the sole basis on which it is judged; the full system needs to be taken into account. This covers its purpose, deployment and oversight as much as the data itself.
RM Block
“In practical terms, organisations must ask not only, ‘Can we use AI?’ but also, ‘Should we, for whose benefit, and under what conditions’?” says Lefebvre.
There’s a line uttered by Jeff Goldblum’s character in Jurassic Park that rings true in this situation: “Your scientists were so preoccupied with whether or not they could, they didn’t stop to think if they should.”
“Responsible AI is about ensuring AI is used in a way that people can trust,” says Bronagh Riordan, AI and data partner at EY Ireland.
Trust should really be the goal with any tech deployment, not just AI. An AI model must be able to engender confidence from the business using it, that company’s customers and the regulators that oversee it.
“Responsible AI should be viewed as a business discipline rather than an ethical add-on,” says Keith Power, data and AI partner, PwC Ireland.
That’s the core of the challenge. Knowing the fundamental principles is the easy part. The real work comes in putting these principles at the centre of everyday decision making.
“The challenge is not defining these principles, but embedding them into how AI is designed, deployed and operated every day,” says Power.
Posting a responsible AI policy on a company website is a good start but it’s more important to ensure that the principles outlined in said policy impact business decisions, from procurement through product design and risk management.
“Responsible AI means using AI in ways that create value without creating hidden harm, unfairness or loss of control for the people it affects,” says Nicola Flannery, partner, technology and transformation, and digital trust and privacy lead at Deloitte.
It’s important to understand that not following these principles can lead to biases, bad decisions and breaks in the chain of accountability.
“The biggest risks of AI are not technical. They are organisational and human,” says Sara Riso, senior research manager, working life unit at Eurofound.
The code in AI doesn’t have views beyond those input by humans. The users in a business are the ones that shape it, so it’s vital they understand that the control and responsibility must rest with them.
“Responsible AI becomes real when it is built into everyday operations through employee training, clear guardrails, regular reviews of how AI is being used and clear accountability for outcomes,” says Riordan.
That means companies need to put together clear guidelines for both the executive branch and frontline workers to follow when it comes to using AI responsibly. This should cover matters such as the limits on what data is entered, clear transparency on where and how AI is being used, and a regular review of outputs.
“Organisations need to know where AI is being used and classify each use according to its risk,” says Lefebvre.
No two AI tools are alike. A digital assistant shouldn’t require the same level of restrictions as a tool shaping financial decisions. Yet the core principles for each should be the same.
“Responsible AI should be embedded into existing governance rather than treated as a stand-alone initiative,” says Power.
This is why responsible AI is far more than a matter for the IT staff in a business. The policy should include significant contributions from the organisation’s legal, compliance and HR staff.
“Buying AI does not transfer responsibility. Organisations remain accountable for how these systems are used and the decisions they support,” says Riso.
This is important for SMEs to bear in mind as buying a system doesn’t shield you from responsibility. Once you have it operating in your company, it’s on you to ensure it’s used correctly.
“Human oversight must be meaningful. The person reviewing an AI output needs the expertise, information, authority and time to challenge it. [It’s] not merely a box to tick,” says Lefebvre.
So it’s vital that those within a business with responsibility for the use of any AI tool understand how it is used and what guardrails need to be in place.
“Trust by design means building AI so that accountability, transparency, data protection and human oversight are part of the system from the start, not bolted on at the end,” says Flannery.
“Confidence will come from showcasing evidence,” says Riordan. This isn’t about slowing down the roll-out of AI across your business. It’s about ensuring the roll-out doesn’t keep hitting walls due to human error.
A smart and planned-out approach can ensure the tools you use deliver what you need and when you need it, responsibly.
