Farsight: Fostering Responsible AI Awareness During AI Application Prototyping

Farsight improves harm identification: After using the tool, AI prototypers were better able to independently identify potential harms associated with their prompts compared to baseline, demonstrating that in-situ support during the prototyping phase can meaningfully enhance responsible AI practices.

Real incidents provide concrete learning: The tool highlights news articles about relevant AI incidents alongside user prompts, helping prototypers connect abstract risks to real-world examples—a design choice that appears to make potential harms more tangible and memorable.

Broader perspective on stakeholders and impacts: User feedback indicated that Farsight encouraged prototypers to think beyond immediate harms and consider end-users more carefully, suggesting the tool helps shift thinking from narrow technical concerns to wider societal impacts.

Existing resources fall short: The study found that Farsight was perceived as more useful and usable than existing resources available to AI prototypers, indicating a significant gap in current tools designed to support responsible AI development during the early prototyping stage.

Accessibility matters for adoption: By making Farsight publicly available and integrating it directly into the prototyping workflow (rather than requiring separate steps), the research demonstrates that responsible AI practices are more likely to be adopted when support tools are easy to access and use in real time.