AI Disclosure in Newsrooms: A Trust Gap

Newsrooms are increasingly integrating generative AI into their workflows, prompting a critical question: how should they disclose AI's role to maintain reader trust? The AI-AI Venn diagram is getting thicker, yet current disclosure practices struggle to foster transparency.

The Transparency Dilemma

Journalists typically have two methods at their disposal. They can opt for brief one-liners or detailed disclosures. However, both approaches have significant drawbacks. A controlled experiment with 34 news readers revealed that detailed disclosures often lead to a transparency dilemma. Instead of enhancing trust, such disclosures can actually diminish it. Readers encounter what essentially becomes a dark pattern, scrolling past with a false sense of transparency.

On the flip side, the one-line approach avoids the pitfall of overwhelming readers but creates an information gap. Readers may expend unnecessary cognitive effort searching for signs of AI involvement, which the brief disclosure hints at but doesn't clarify. This leaves us with a fundamental question: How can we design AI disclosures that genuinely build trust without overwhelming or deceiving the reader?

Reimagining Disclosure

Interestingly, the readers aren't dismissing the notion of transparency altogether. They crave user-centric designs. Suggestions like detail-on-demand interactions, proportional AI-ratio visualizations, outlet-level signals, and explicit 'no AI' labels show a proactive stance from users in wanting to understand AI's role at their discretion.

This reveals a significant disconnect between what newsrooms believe constitutes responsible disclosure and what users actually require. It's a design problem that the Human-Computer Interaction (HCI) community must address. If agents have wallets, who holds the keys? In this case, the readers should arguably hold the metaphorical keys to how AI involvement is disclosed.

The Path Forward

It's time for the HCI community and newsrooms to rethink their approach. Disclosures should empower readers rather than create unnecessary hurdles. The compute layer needs a payment rail, and in this context, transparency is the currency. A user-centric model of disclosure could redefine how AI's role is perceived in journalism and beyond.

At its core, the issue isn't just about disclosure. It's about trust, autonomy, and respecting the reader's right to understand the extent of AI involvement. In a world that's increasingly autonomous, giving readers control over the information they consume is more than just a nice-to-have. It's essential.