Good Judgment, Bad Judgment, and the Problem with AI

The path to knowing passes through doing. (Photo of hands engaged in traditional weaving process).

“Good judgment comes from experience, and experience comes from bad judgment” is an old aphorism attributed online to any number of the internet’s favorite providers of wisdom (Mark Twain? Will Rogers? Abe Lincoln?)

I heard it as a child, so I’m going to credit it to one of the Jesuits who taught my brother and whose fondness for the expression was responsible for it making its way onto the bulletin board in our kitchen.

I’ve been thinking about the phrase a lot lately, because good judgment is vital - and I sometimes wonder whether it is in declining supply.

When I think about the downsides of AI - the ethical concerns over the use of copyrighted material, the potential elimination of vast numbers of jobs, and the environmental costs of the enormous amount of power it requires - as well as the potential upsides - speed, efficiency, translation, accessibility, and making repetitive tasks less onerous - I keep coming back to it’s potential effect on judgment.

We know that using AI responsibly requires a human being to review the results: to check for hallucinations, invented citations, factual errors, and a tone that would sound appalling to everyone except the robot’s tin ear.

But that review requires a human who knows the situation at least as well as the AI, and preferably better. In other words, it requires an experienced human.

And how did that human get that experience?

Trial and error. Mistakes and screwups, large and small, over years and years and years. Lessons learned through observation, or through moments of mortifying embarrassment. Watching someone else’s shoddy work being held up to ridicule. Saying to yourself, “I will never do that again” - and then, ideally, never doing it again.

"Failure - the greatest teacher." (I know that’s a cliché - does Claude?)

It may not have been fun at the time, but we generally believe that our experiences make us better judges of character, tone, and quality. We learn to recognize what works because we have seen what doesn’t. We develop instincts because we have spent years making decisions, seeing the consequences, and adjusting accordingly.

But what happens when we outsource more and more of that process?

When we use AI, we are often reaching an outcome. But are we learning?

I say to myself - and anyone who will listen: use AI when you want to do something, but not when you want to know something. After all, you can’t shortcut your way around the process of gaining knowledge. You can’t parachute into wisdom without passing through the fires of experience. 

But what if some of the doing is what leads to the judgment?

At this point in my career, I have years of experience that allow me to quickly analyze pros and cons, spot issues, identify problems, and anticipate potential solutions. I wasn’t born with that ability. No one is.

I earned it by doing things badly, doing them better, watching other people do them badly, making mistakes, learning from them, and gradually developing the judgment to recognize the difference.

I’m glad that judgment is still of value in the world. But where will the judgment of the future come from?

If we increasingly ask machines to do the work through which human beings have traditionally gained knowledge, experience, and judgment, what happens when the next generation is asked to decide whether the machine is right?

Who will have the experience to know?