nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

Disclosure when AI Assistant is used for Metadata Creation

RichardDaniels

In ArcGIS Online there is now an AI Assistant that can auto-populate metadata for a feature or image layer by executing a deep dive into the underlying structure of the feature class. This results in 'suggested' metadata that then can be Saved to the Document. When this tool is used a Disclaimer should be automatically added to the appropriate place in the metadata. For example:

Terms of Use (Use Limitation): "Metadata shown here was partially generated using an AI-Assistant. The outputs generated by AI may not always be accurate. We encourage you to review, fact-check, and adjust outputs as needed before relying on them. You are responsible for your use of the content generated using this AI feature."

Najděte více příspěvků označených jako

Status: Otevřít

Komentáře

furme15

I like this as an optional setting. It would be convenient for anyone working for an agency with guidelines on the use of AI calling for transparency and attribution. I like that it would standardize the language used to meet such guidelines while remaining independent of the Terms of Use. As a data consumer I want to know if any dataset, but especially those marked as authoritative, displays metadata authored or partially informed by AI. Otherwise, in my mind, that authoritative stamp is somewhat undermined; metadata is an extension of the data and if it didn't originate with the authoritative source (which AI itself is not) then it should be knowable.

ReedHunter

I'm seeing the need behind this feature as one of trust. We're all experiencing how AI output can be unreliable. I'm liking the idea of potentially granular sections of the metadata being flagged as ai output until manually confirmed accurate by a steward. There will be incidents of inaccurate or incomplete metadata when it's AI generated, and letting there be the option of warning users (or other AI agents) that this metadata may not be trustworthy is a need that this feature would meet. If your organization does not want to use this feature, then they don't have to.

Vandal_WSU

This is a great idea. Having a citation like this in there would be so helpful. It could be as simple as a check box that can be unchecked after an actual person reviews/approves the metadata.

JonathanMcD

Not sure I agree with this, it should be a suggested inclusion as, ultimately, it's the user's decision whether to fully accept the suggestions or not. These suggestions then could be further edited without any assistant.

Additionally, you could use an external AI tool to create the metadata for you, It'd then be fully up to the user to include a disclaimer, which I doubt they would.

At the end of the day the tool does not auto-populate anything, it only offers suggestions if a user starts the tool. I think it'd be wholly different if, as a user published a dataset, an assistant created summary-description-terms during this workflow.

JamesBedortha

this could be a good idea

Připojte se k diskusi

Přihlaste se pro psaní příspěvků, sledování obsahu a další. Jste tu noví? Zaregistrujte se zdarma.