NASA’s new dark energy space telescope can also detect killer asteroids
technologyreview.com·2h ago
AI Research
arXiv CS.AI·Jun 15·5 min read
SpheriCity: Designing Trustworthy Conversational AI for Sustainability Decision Support
TL;DR
Sustainability reports are often lengthy and complex, making it hard for users to extract useful insights. SpheriCity, a conversational AI prototype, was developed to enhance knowledge sensemaking by providing transparent sourcing and structured synthesis of information.
✦ Why It Matters
Engineers can leverage SpheriCity's design principles to enhance trust in AI applications across various domains.
Key Takeaways
- Implement transparent sourcing in AI systems to build user trust.
- Focus on structured information synthesis for complex data.
- Conduct expert reviews to refine AI tools for specific domains.
- Integrate AI solutions into existing workflows for better usability.
Full Summary
Sustainability reports contain valuable information about materials and policies but are often too complex for effective analysis. SpheriCity is a conversational AI prototype designed to assist users in navigating these reports by emphasizing evidence traceability and structured information synthesis.
The development involved a formative expert review with six sustainability experts who tested the tool using various queries related to cross-city comparisons and policy recommendations. Results showed that features like transparent sourcing and contextual explanations significantly influenced experts' trust in the AI's responses.
Experts noted that the tool's design aligned well with their workflows, enhancing its perceived usefulness. This research contributes a novel framework for evaluating AI in high-stakes knowledge domains and offers insights into how to improve user trust in AI systems.
Overall, SpheriCity represents a step forward in integrating AI into sustainability decision support.