Ethics in the Age of AI
How do we know if our shiny AI tools are being used in ethical ways? When someone asks me this question, my response goes something like this: We can’t make those decisions yet, because most of us are “playing catch-up” with the technology. If we don’t get smarter real soon, other people will make those decisions for us. And we may not like the world they make.
So what is “ethical AI?”
Conservationist Aldo Leopold wrote that “ethics is doing the right thing, even when no one else is watching.” His definition is more fitting than ever in the world of AI—because sometimes, no one really is watching.
The Danger of Black Box Thinking
In science and engineering, the term black box refers to any complex device for which we know the inputs and outputs, but not the inner workings. For example, to many of us, our streaming service is a black box. We push the buttons and select the program we wish to watch. We neither know nor care how it works.
To behaviorists, the human mind is a black box. It is the most complex and mostly unknown object we know. And yet, being the remarkable creatures that we are, we have the audacity to develop machines that are intended to think the way we do. Never mind that we haven’t quite figured how we think yet.
B.F. Skinner, put it like this: “The real question is not whether machines think but whether men do. The mystery which surrounds a thinking machine already surrounds a thinking man.”
In Rebooting AI, cognitive scientist Gary Marcus summarizes the problem:
“There are known knowns and known unknowns, but what we should be worried about most is the unknown unknowns.”
As talent development professionals, we have a responsibility to advocate now for the ethical deployment of these powerful tools. Here’s how.
Building transparent AI will not be easy and we need to be the ones to lead the way. If your company is implementing AI to recruit talent, train employees, evaluate performance, or anything else related to talent development, you are the customer. Insist on a thorough explanation of how the AI model will make decisions. (The answer, or lack of one, might surprise you.)
Be Skeptical
As we enter the age of AI, we must be skeptical about any shiny new AI program, even if it is already in use in many other companies. It is possible that you will be the only one asking the hard questions, so you need to be up to the task. Here are a few ideas to get you started.
This might sound obvious, but many buyers of AI solutions focus only on the desired result, such as “Find the best candidate for the job,” or “Process routine customer complaints.” True definition involves describing what decisions the AI will make to identify next steps, what criteria are used to make those decisions, and the source of the underlying data.
Request a “next best decision” test.
AIs are programmed to make the best decision every time. Human beings seldom have that luxury. Comparing the “best” decision with the “next best” is one way to identify flaws in the underlying logic, or biases that have been accidentally built into the machines by fallible humans.
Run the same queries or tasks repeatedly.
The purpose of the test is not merely to see if the machine returns a consistent result; it’s to determine if the result remains true to the original intent.
Ethical Behavior is Up to Us
While the age of AI presents new challenges, the basics remain the same. Ethical behavior is always up to us – all of us and each of us. So, let us begin the way we humans have always learned about new tools.
What is this?
How can I use it?
What could go wrong?
What can I do to protect myself and others?
Where do I begin?
