Human-in-the-loop

how you involve, oversee and delegate to ai systems

This card covers how people stay involved in what the AI does. It is about deciding where a human checks, approves, or overrides the tool, and where the tool may act on its own. For journalism, these lines matter, since some decisions should never be left to a machine alone.

Questions to explore

// use these as prompts in a workshop or on your own. There are no right answers.

At which points must a person review or approve what the AI produced?
Where is it safe for the tool to act without someone checking first?
How do you keep review from becoming a rubber stamp people stop reading?
Who is accountable when the AI is wrong and a person signed off anyway?
What decisions should never be handed to the tool, no matter how good it gets?

Expert voices

// notes from the journalists and AI experts who helped shape this kit

“Is it okay to automate parts of content production without final human verification? Decide which steps in producing video, audio, and text can run without a human check.”

Anastasiia Korinovska, Suspilne

“Have a human in the loop, but decide deliberately at what point in the process: before, during, or after the output.”

Martin Schori, Aftonbladet

“Humans must remain in control of journalistic decisions. The question is where to draw the line between AI assistance and AI autonomy.”

Lynn Khellaf, DW

“Frame AI as assistance, not replacement, and keep emphasizing human oversight.”

Bahia Albrecht, DW Akademie

Things to consider

Oversight only works if the reviewer has time and reason to look closely.
Decide which calls stay with a human before you automate anything.
A person approving without reading is not real oversight.

using this card

Pull Human-in-the-loop when it is relevant and set it aside when it is not. Pair it with the other AI Solutions cards, lay them out on a table, and use the questions above to get everyone on the same page. Capture what you discuss on sticky notes or in a shared doc.