Keep the Human in the Loop

One of the most useful resources I have found for designing with AI did not come from a book. It came from Google.

The People + AI Guidebook, created by Google’s People + AI Research team, known as PAIR, is a practical collection of guidance for building AI products that people actually want to use. What I appreciate most is that it does not treat automation as the automatic goal.

In its chapter on feedback and control, the guidebook makes a point that sounds simple but changes how you plan a product. “Not every task is a chore.”

Some tasks people want off their plate. Others they enjoy. Others they need to own because the outcome is personal or high stakes. The guidebook encourages teams to decide carefully whether to automate a task or augment the person doing it.

That decision is where keeping the human in the loop begins.

Here is what that looks like in practice.

First, ask what people want to keep. Before automating anything, research which parts of the task bring satisfaction, judgment or responsibility. Those are the parts to support, not remove.

Second, give people real control over the output. The guidebook recommends designs that “allow users to adapt the output to their needs, edit it, or turn it off.” Control is not a setting buried in a menu. It is part of the main experience.

Third, make feedback effortless. A quick rating, a correction or a dismissal tells the system something valuable. PAIR describes user feedback as “the communication channel between your users, your product, and your team.”

Fourth, let people change their minds. Preferences shift. Let users update or erase earlier choices without starting over.

Fifth, respect a no. If someone chooses not to use an AI feature, that choice deserves the same respect as a yes.

The goal is not less human involvement. It is better human involvement.

Fei-Fei Li, founding co-director of Stanford’s Institute for Human-Centered AI, has said she wants to replace the word replace. As she put it, “I think AI is here to ‘augment’ humans.”

That one word changes the design question. We stop asking what the machine can take over. We start asking what the person can now do better.

Keep the human in the loop. Not as a safety net at the end, but as the reason the product exists.