In Good Faith

At one point in my life I was a strong junior tennis player. When I was fifteen, I reached the quarter-final of a highly competitive tournament. Normally, those matches had an umpire, but none was available when ours was due to start. My opponent and I agreed to keep score ourselves.

The match was a best of three sets. I won the first set, and was up a break in the second at 5–2. At the change of ends, as we crossed the net, I noted the score. My opponent disagreed. 4–3, he said.

It wasn’t. The match had spectators, and more than one confirmed my call of the score. But none of the people watching were tournament officials, so their agreement didn’t settle the disagreement.

I don’t think my opponent set out to lie. My suspicion is that he became confused and, once he had publicly disagreed with me, backing down became difficult. Admitting in front of an audience that he had claimed a game he hadn’t won might itself have looked like cheating. He may have felt stuck defending the score even after he began to doubt it.

The tournament director got involved. I thought the spectators would help my case, but instead he determined that he could not know which of us was right. Under those circumstances, he ruled that the match must start again. From the beginning. First set, love-all, and this time with an umpire. I lost in straight sets.

I felt robbed that day. It changed my approach to competitive tennis. For years, I ruminated on whom to blame. As I look back now, I don’t think anyone needed to act in bad faith for it to happen. My opponent could have made an honest mistake, and the director had to resolve a disagreement he hadn’t witnessed. Whatever each of us should have done differently, the score I had played to was no longer the score we could carry on from.

I think conversations depend on something similar. We don’t need to agree, but we do need some understanding of where each person stands. Without it, we can find ourselves responding to positions that nobody actually holds.

When I ask for or receive feedback from someone, I do so with the hope that I will learn what they are thinking. That sentence may feel like the most blindingly obvious thing you will read this month, but stick with me. I take feedback as something that we can build on together. If a person hasn’t actually understood what they have given to me as feedback, building on it is a substantially more difficult task.

Imagine that I send a draft document to a colleague requesting their feedback and they run it through AI before handing me the unreviewed feedback. That feedback might appear fluent enough, but perhaps it also misses a key point and suggests that we include a section that already appears. I tell my colleague that I’m very glad we agree that the section belongs in the document.

I don’t raise the use of AI with my colleague because there simply isn’t enough for me to gain by doing so. They have sent it as their feedback, so I answer it as theirs. I do the work and use the suggestions to the extent that I can. Whatever effort went into producing the feedback, the work of processing it is now mine.

In my example, I can at least see the problem. If AI feedback is good, I can use it, but I’m still bothered by where my reply goes.

Imagine now that the feedback makes a good suggestion, but I don’t agree that we should take it. The change would make the piece easier to read, but it would cost detail that matters to the readers I had in mind. I want to discuss which compromise my colleague thinks we should make, and why. If they haven’t fully considered their own suggestion, I may be the first of us to notice that there is even a choice to be made.

My colleague might look at the document, consider it, and have a great answer. That’s fine, but I am bound by courtesy to assume that they had already formed a view. Since AI arrived in the work I do, I have participated in projects where I’ve spent considerable time working through problems caused by suggestions that hadn’t been considered by the people who passed them on.

The problem is not people putting their name to words that someone else drafted. People I support frequently send emails that I drafted. For my whole career, I have received documents forwarded to me unread. Every one of these essays has had feedback from an AI assistant before publication. The external input is ordinary.

But there is a difference between sending me something to consider and sending me what appears to be the result of their own consideration. AI makes it easy to produce the second before that consideration has happened. The output can be as compelling to my colleague as they hope it will be to me. They may agree with every word without having worked through the choice those words propose.

When I reply, I am trying to understand what they think we should do, and why. Instead, I may spend a round of correspondence finding out whether they have formed a view at all.

A person using AI to help themselves understand a piece of work is one thing. If they keep using it to answer without reaching that understanding, each exchange can move the document forward while leaving them further behind.

I’ve contemplated trying to find out how much a colleague or counterparty has understood about their comments. I’ve never actually done it because adding an oral examination to the process of interacting with me just seems cruel. But I have asked how a particular suggestion fits with something already in the document. In my experience, that has brought the conversation back to the person I’m talking to.

I can’t always know how much of that understanding is there when a reply arrives. I’ll sometimes spend a full round of effort answering a position that the person who sent it hasn’t yet worked through. Even when the feedback is useful, that uncertainty adds work.

On the court, it wasn’t my choice to start again. I was told to act as if the score had never happened. With my work, I have a choice about how to respond when I’m unsure how much of a person’s judgement is in the words they’ve sent. I choose to answer in good faith, in the hope that we can build an understanding we both actually hold.