Poor Proximity Logic for "Nearby" Recommendations
Summary
When asked for "nearby" attractions in Mexico, Pampanga, the Arcsphere Navigator suggests locations that are 20–40 km away (Mt. Arayat and San Fernando). It fails to recommend attractions in Angeles City, which is significantly closer to the user's current location than the suggested points, resulting in a poor user experience for local discovery.
Steps to Reproduce
- Be physically located in Mexico, Pampanga.
- Command: "Recommend nearby attractions based on my current location"
- Observe the distances/travel times of the suggestions.
Expected Behavior
The AI should prioritize attractions within a 5–15 km radius. Even if the user is in Mexico, the Navigator should suggest the major hubs in Angeles City (e.g., Clark, Marquee Mall area, etc.) as they are closer and more numerous than a 40km drive to Mt. Arayat.
Actual Behavior
The AI bypasses closer, high-density attraction areas (Angeles City) and instead suggests:
- Mt. Arayat National Park: ~25–40 min drive (20–30 km)
- San Fernando Heritage District: ~20–30 min drive
The Problem
In a local context, 20–30 km is not "nearby." The AI is likely picking "famous" landmarks in the province rather than using actual spatial distancing to find the truly closest points of interest.
Suggested Fix
- Distance Thresholding: Implement a hard limit for "Nearby" queries (e.g., prioritize results within 10km first).
- Cross-Boundary Search: Ensure the AI looks at neighboring cities (like Angeles City) for closer attractions rather than staying strictly within the detected municipality's "top hits."
- Travel Time Optimization: Prioritize "Nature & Outdoors" that are within a 15-minute drive rather than a 40-minute drive.